nips nips2011 nips2011-280 knowledge-graph by maker-knowledge-mining

280 nips-2011-Testing a Bayesian Measure of Representativeness Using a Large Image Database


Source: pdf

Author: Joshua T. Abbott, Katherine A. Heller, Zoubin Ghahramani, Thomas L. Griffiths

Abstract: How do people determine which elements of a set are most representative of that set? We extend an existing Bayesian measure of representativeness, which indicates the representativeness of a sample from a distribution, to define a measure of the representativeness of an item to a set. We show that this measure is formally related to a machine learning method known as Bayesian Sets. Building on this connection, we derive an analytic expression for the representativeness of objects described by a sparse vector of binary features. We then apply this measure to a large database of images, using it to determine which images are the most representative members of different sets. Comparing the resulting predictions to human judgments of representativeness provides a test of this measure with naturalistic stimuli, and illustrates how databases that are more commonly used in computer vision and machine learning can be used to evaluate psychological theories. 1

Reference: text


Summary: the most important sentenses genereted by tfidf model

sentIndex sentText sentNum sentScore

1 We extend an existing Bayesian measure of representativeness, which indicates the representativeness of a sample from a distribution, to define a measure of the representativeness of an item to a set. [sent-14, score-1.711]

2 Building on this connection, we derive an analytic expression for the representativeness of objects described by a sparse vector of binary features. [sent-16, score-0.812]

3 We then apply this measure to a large database of images, using it to determine which images are the most representative members of different sets. [sent-17, score-0.321]

4 Comparing the resulting predictions to human judgments of representativeness provides a test of this measure with naturalistic stimuli, and illustrates how databases that are more commonly used in computer vision and machine learning can be used to evaluate psychological theories. [sent-18, score-1.22]

5 1 Introduction The notion of “representativeness” appeared in cognitive psychology as a proposal for a heuristic that people might use in the place of performing a probabilistic computation [1, 2]. [sent-19, score-0.194]

6 For example, we might explain why people believe that the sequence of heads and tails HHTHT is more likely than HHHHH to be produced by a fair coin by saying that the former is more representative of the output of a fair coin than the latter. [sent-20, score-0.283]

7 This proposal seems intuitive, but raises a new problem: How is representativeness itself defined? [sent-21, score-0.783]

8 Various proposals have been made, connecting representativeness to existing quantities such as similarity [1] (itself an ill-defined concept [3]), or likelihood [2]. [sent-22, score-0.911]

9 Tenenbaum and Griffiths [4] took a different approach to this question, providing a “rational analysis” of representativeness by trying to identify the problem that such a quantity solves. [sent-23, score-0.783]

10 They proposed that one sense of representativeness is being a good example of a concept, and then showed how this could be quantified via Bayesian inference. [sent-24, score-0.783]

11 The resulting model outperformed similarity and likelihood in predicting human representativeness judgments for two kinds of simple stimuli. [sent-25, score-1.092]

12 The question of what makes a good example of a concept is of direct relevance to computer scientists as well as cognitive scientists, providing a way to build better systems for retrieving images or documents relevant to a user’s query. [sent-27, score-0.266]

13 In addition, we show how measuring the representativeness of items in sets can also provide a novel method of finding outliers in sets. [sent-30, score-0.996]

14 By extending the Bayesian measure of representativeness to apply to sets of objects and testing it with a large image database, we are taking the first steps towards a closer integration of the methods of cognitive science and machine learning. [sent-31, score-1.044]

15 Cognitive science experiments typically use a small set of artificial stimuli, and evaluate different models by comparing them to human judgments about those stimuli. [sent-32, score-0.234]

16 We combine these methods by soliciting human judgments to test cognitive models with a large set of naturalistic stimuli. [sent-34, score-0.372]

17 This provides the first experimental comparison of the Bayesian Sets algorithm to human judgments, and the first evaluation of the Bayesian measure of representativeness in a realistic applied setting. [sent-35, score-0.908]

18 Section 2 provides relevant background information, including psychological theories of representativeness and the definition of Bayesian Sets. [sent-37, score-0.923]

19 Section 4 describes the dataset derived from a large image database that we use for evaluating this measure, together with the other psychological models we use for comparison. [sent-39, score-0.258]

20 Section 5 presents the results of an experiment soliciting human judgments about the representativeness of different images. [sent-40, score-1.013]

21 2 Background To approach our main question of which elements of a set are most representative of that set, we first review previous psychological models of representativeness with a particular focus on the rational model proposed by Tenenbaum and Griffiths [4]. [sent-43, score-1.072]

22 1 Representativeness While the notion of representativeness has been most prominent in the literature on judgment and decision-making, having been introduced by Kahneman and Tversky [1], similar ideas have been explored in accounts of human categorization and inductive inference [6, 7]. [sent-46, score-0.915]

23 In these accounts, representativeness is typically viewed as a form of similarity between an outcome and a process or an object and a concept. [sent-47, score-0.837]

24 Assume some data d has been observed, and we want to evaluate its representativeness of a hypothesized process or concept h. [sent-48, score-0.844]

25 In the case where we want to evaluate the representativeness of an outcome to a set, we might use metrics of the kind that are common in categorization models: an exemplar model defines similarity in terms of the sum of the similarities to the other objects in the set (e. [sent-51, score-1.015]

26 , [8, 9]), while a prototype model defines similarity in terms of the similarity to a prototype that captures the characteristics of the set (e. [sent-53, score-0.344]

27 An alternative to similarity is the idea that representativeness might track the likelihood function P (d|h) [11]. [sent-56, score-0.888]

28 However, people’s judgments from the coin flip example with which we started the paper go against this idea of equivalence, since both flips have equal likelihood yet people tend to judge HHTHT as more representative of a fair coin. [sent-58, score-0.367]

29 Analyses of typicality have also argued against the adequacy of frequency for capturing people’s judgments about what makes a good example of a category [6]. [sent-59, score-0.235]

30 Tenenbaum and Griffiths [4] took a different approach to this question, asking what problem representativeness might be solving, and then deriving an optimal solution to that problem. [sent-60, score-0.802]

31 The resulting rational model of representativeness takes the problem to be one of selecting a good example, where the best example is the one that best provides evidence for the target process or concept relative to possible alternatives. [sent-62, score-0.903]

32 This task is an example of clustering on-demand, in which the original set of items represents some concept or cluster such as “condiment” and we are to find other items that would fit appropriately into this set. [sent-67, score-0.236]

33 Bayesian Sets is a formalization of this process in which items are ranked by a modelbased probabilistic scoring criterion, measuring how well they fit into the original cluster [5]. [sent-68, score-0.251]

34 Under these assumptions, the scoring criterion for Bayesian Sets reduces to p(x∗ , Ds ) = Bscore(x ) = p(x∗ )p(Ds ) ∗ j αj + βj αj + βj + N N ˜ where αj = αj + n=1 xnj and βj = βj + N − ˜ in x and can be computed efficiently as N n=1 αj ˜ αj x∗j ˜ βj βj 1−x∗j (3) xnj . [sent-79, score-0.191]

35 The logarithm of this score is linear log Bscore(x∗ ) = c + sj x∗j (4) j ˜ ˜ ˜ where c = j log(αj + βj ) − log(αj + βj + N ) + log βj − log βj , sj = log αj − log αj − log βj + ∗ log βj , and x∗j is the jth component of x . [sent-80, score-0.241]

36 Motivated by this method, we now turn to extending the previous measure of representativeness for a sample from a distribution, to define a measure of representativeness for an item to a set. [sent-82, score-1.711]

37 3 3 A Bayesian Measure of Representativeness for Sets of Objects The Bayesian measure of representativeness introduced by Tenenbaum and Griffiths [4] indicated the representativeness of data d for a hypothesis h. [sent-83, score-1.608]

38 For instance, in an image retrieval problem, we might just have a set of images that are all assigned to the same category, without a clear idea of the distribution that characterizes that category. [sent-85, score-0.251]

39 In this section, we show how to extend the Bayesian measure of representativeness to indicate the representativeness of an element of a set, and how this relates to the Bayesian Sets method summarized above. [sent-86, score-1.608]

40 We can perform an analysis similar to that given for the representativeness of d to a hypothesis, and obtain the expression R(d, Ds ) = P (d|Ds ) D =Ds P (d|D )P (D ) (5) which is simply Equation 1 with hypotheses replaced by datasets. [sent-88, score-0.802]

41 4 Evaluating Models of Representativeness Using Image Databases Having developed a measure of the representativeness of an item in a set of objects, we now focus on the problem of evaluating this measure. [sent-94, score-0.886]

42 In the case of representativeness, the stimuli previously used by Tenenbaum and Griffiths [4] to evaluate different representativeness models consisted of 4 coin flip sequences and 45 arguments based on predicates applied to a set of 10 mammals. [sent-96, score-0.903]

43 Using an existing image database of naturalistic scenes, we can better test the predictions of different representativeness theories with stimuli much more in common with the environment humans naturally confront. [sent-100, score-1.062]

44 In the rest of this section, we present the dataset used for evaluation and outline the implementations of existing models of representativeness we compare our rational Bayesian model against. [sent-101, score-0.927]

45 1 Dataset We use the dataset presented in [14], a subset of images taken from the Corel database commonly used in content-based image retrieval systems. [sent-103, score-0.314]

46 The images in the dataset are partitioned into 50 labeled sets depicting unique categories, with varying numbers of images in each set (the mean is 264). [sent-104, score-0.338]

47 The labels have been criticized for not always being of high quality [16], which provides an additional (realistic) challenge for the models of representativeness that we aim to evaluate. [sent-109, score-0.854]

48 The images in this dataset are represented as 240-dimensional feature vectors, composed of 48 Gabor texture features, 27 Tamura texture features, and 165 color histogram features. [sent-110, score-0.164]

49 The images were additionally preprocessed through a binarization stage, transforming the entire dataset into a sparse binary matrix that represents the features which most distinguish each image from the rest of the dataset. [sent-111, score-0.225]

50 2 Models of Representativeness We compare our Bayesian model against a likelihood model and two similarity models: a prototype model and an exemplar model. [sent-114, score-0.289]

51 We build upon a simple leave-one-out framework to allow a fair comparison of these different representativeness models. [sent-115, score-0.81]

52 Given a set of images with a particular category label, we iterate through each image in the set and compute a score for how well this image represents the rest of the set (see Algorithm 1). [sent-116, score-0.389]

53 An example of using this measure on the set of 299 images for category label coast is presented in Figure 1. [sent-122, score-0.314]

54 Panels (a) and (b) of this figure show the top nine and bottom nine ranked images, respectively, where it is quite apparent that the top ranked images depict a better set of coast examples than the bottom rankings. [sent-123, score-0.668]

55 It also becomes clear how poorly this label applies to some of the images in the bottom rankings, which is an important issue if using the labels provided with the Corel database as part of a training set for learning algorithms. [sent-124, score-0.296]

56 This model treats representative judgments of an item x∗ as p(x∗ |Ds ) for a set ∗ Ds = {x1 , . [sent-126, score-0.272]

57 The likelihood model scoring criterion is Lscore(x∗ ) = p(x∗ , Ds ) = p(Ds ) j 1 x ˜ (˜ j ) ∗j βj α αj + βj + N N ˜ where αj = αj + n=1 xnj and βj = βj + N − ˜ linear in x and can be computed efficiently as N n=1 log Lscore(x∗ ) = c + 1−x∗j (6) xnj . [sent-131, score-0.249]

58 In this model we define a prototype vector xproto to be the modal features for a set of items Ds . [sent-135, score-0.246]

59 We define the exemplar model using a similar scoring metric to the prototype model, except rather than computing the distance of x∗ to a single prototype, we compute a distance for each item in the set Ds . [sent-139, score-0.31]

60 5 Modeling Human Ratings of Representativeness Given a set of images provided with a category label, how do people determine which images are good or bad examples of that category? [sent-142, score-0.458]

61 In this section we present an experiment which evaluates our models through comparison with human judgments of the representativeness of images. [sent-143, score-0.998]

62 The stimuli were created by identifying the top 10 and bottom 10 ranked images for each of the 50 categories for the Bayesian, likelihood, and prototype models and then taking the union of these sets for each category. [sent-147, score-0.586]

63 The exemplar model was excluded in this process as it required optimization of its λ parameter, meaning that the best and worst images could not be determined in advance. [sent-148, score-0.22]

64 Participants were shown a series of images and asked to rate how good an example each image was of the assigned category label. [sent-150, score-0.294]

65 Image quality ratings were made on a scale of 1-7, with a rating of 1 meaning the image is a very bad example and a rating of 7 meaning the image is a very good example. [sent-152, score-0.421]

66 2 Results Once the human ratings were collected, we computed the mean ratings for each image and the mean of the top 10 and bottom 10 results for each algorithm used to create the stimuli. [sent-154, score-0.567]

67 We also computed 6 6 Bayes Likelihood Prototype Mean quality ratings 5. [sent-155, score-0.195]

68 5 3 Top 10 Rankings Bottom 10 Rankings Figure 2: Mean quality ratings of the top 10 and bottom 10 rankings of the different representativeness models over 50 categories. [sent-158, score-1.167]

69 The vertical axis is bounded by the best possible top 10 ratings and the worst possible bottom 10 ratings across categories. [sent-160, score-0.444]

70 bounds for the ratings based on the optimal set of top 10 and bottom 10 images per category. [sent-161, score-0.418]

71 These are the images which participants rated highest and lowest, regardless of which algorithm was used to create the stimuli. [sent-162, score-0.193]

72 The mean ratings for the optimal top 10 images was slightly less than the highest possible rating allowed (m = 6. [sent-163, score-0.407]

73 074), while the mean ratings for the optimal bottom 10 images was significantly higher than the lowest possible rating allowed (m = 2. [sent-165, score-0.433]

74 The Bayesian model had the overall highest ratings for its top 10 rankings (m = 5. [sent-169, score-0.287]

75 026) and the overall lowest ratings for its bottom 10 rankings (m = 3. [sent-171, score-0.313]

76 031), and prototype having the lowest top 10 (m = 4. [sent-178, score-0.199]

77 We tested for statistical significance via pairwise t-tests on the mean differences of the top and bottom 10 ratings over all 50 categories, for each pair of models. [sent-183, score-0.283]

78 Although we did not explicitly ask participants to rank images, their quality ratings implicitly provide an ordering on the images that can be compared against the models. [sent-187, score-0.364]

79 This also gives us an opportunity to evaluate the exemplar model, optimizing its λ parameter to maximize the fit to the human data. [sent-188, score-0.166]

80 To perform this correlation we recorded the model scores over all images for each category, and then computed the correlation of each model with the human judgments within that category. [sent-189, score-0.37]

81 3 Discussion Overall, the Bayesian model of representativeness provided the best account of people’s judgments of which images were good and bad examples of the different categories. [sent-200, score-1.087]

82 The mean ratings over the entire dataset were best predicted by our model, indicating that on average, the model predictions for images in the top 10 results were deemed of high quality and the predictions for images in the bottom 10 results were deemed of low quality. [sent-201, score-0.692]

83 Since the images from the Corel database come with labels given by human judges, few images are actually very bad examples of their prescribed labels. [sent-202, score-0.438]

84 This explains why the ratings for the bottom 10 images are not much lower. [sent-203, score-0.367]

85 This explains why the ratings for the top 10 images are not much higher, and thus why the difference between top and bottom 10 on average is not larger. [sent-205, score-0.469]

86 016 scores from the different models against the ranked order of human quality ratings, the Bayesian account was also significantly more accurate than the other models. [sent-216, score-0.203]

87 While the actual correlation value was less than 1, the dataset was rather varied in terms of quality for each category and thus it was not expected to be a perfect correlation. [sent-217, score-0.179]

88 6 Finding Outliers in Sets Measuring the representativeness of items in sets can also provide a novel method of finding outliers in sets. [sent-219, score-0.979]

89 Since models of representativeness can be used to rank items in a set by how good an example they are of the entire set, outliers should receive low rankings. [sent-221, score-0.956]

90 To empirically test this idea we can take an image from a particular category and inject it into all other categories, and see whether the different measures can identify it as an outlier. [sent-223, score-0.175]

91 To find a good candidate image we used the top ranking image per category as ranked by the Bayesian model. [sent-224, score-0.388]

92 Thus, the top ranked image for a particular category is assumed to be a bad example of the other categories. [sent-226, score-0.333]

93 We evaluated how low this outlier was ranked by each of the representativeness measures 50 times, testing the models with a single injected outlier from each category to get a more robust measure. [sent-227, score-1.136]

94 The final evaluation was based on the normalized outlier ranking for each category (position of outlier divided by total number of images in the category), averaged over the 50 injections. [sent-228, score-0.39]

95 We interpret this as another indication of the noisiness of the original image labels in the dataset since there were a number of images in each category that were ranked lower than the outlier. [sent-232, score-0.435]

96 7 Conclusions We have extended an existing Bayesian model of representativeness to handle sets of items and showed how it closely approximates a method of clustering on-demand – Bayesian Sets – that had been developed in machine learning. [sent-233, score-0.919]

97 We exploited this relationship to allow us to evaluate a set of psychological models of representativeness using a large database of naturalistic images. [sent-234, score-1.028]

98 Our Bayesian measure of representativeness significantly outperformed other proposed accounts in predicting human judgments of how representative images were of different categories. [sent-235, score-1.278]

99 We also introduced a novel method of detecting outliers in sets of data using our representativeness measure, and showed that it outperformed other measures. [sent-237, score-0.926]

100 We hope that the combination of methods from cognitive science and computer science that we used to obtain these results is the first step towards closer integration between these disciplines, linking psychological theories and behavioral methods to sophisticated algorithms and large databases. [sent-238, score-0.208]


similar papers computed by tfidf model

tfidf for this paper:

wordName wordTfidf (topN-words)

[('representativeness', 0.783), ('ds', 0.261), ('ratings', 0.161), ('judgments', 0.137), ('images', 0.135), ('bayesian', 0.126), ('prototype', 0.118), ('psychological', 0.099), ('category', 0.098), ('items', 0.097), ('ranked', 0.091), ('exemplar', 0.085), ('rational', 0.078), ('se', 0.078), ('bscore', 0.077), ('corel', 0.077), ('representative', 0.074), ('bottom', 0.071), ('cognitive', 0.068), ('tenenbaum', 0.065), ('human', 0.062), ('xnj', 0.062), ('image', 0.061), ('ths', 0.061), ('item', 0.061), ('grif', 0.061), ('outliers', 0.06), ('naturalistic', 0.058), ('people', 0.058), ('outlier', 0.055), ('similarity', 0.054), ('database', 0.053), ('rankings', 0.051), ('top', 0.051), ('dw', 0.051), ('psychology', 0.049), ('scoring', 0.046), ('stimuli', 0.046), ('concept', 0.042), ('measure', 0.042), ('nine', 0.042), ('dist', 0.042), ('kahneman', 0.041), ('theories', 0.041), ('sets', 0.039), ('coin', 0.039), ('heller', 0.037), ('dwi', 0.037), ('retrieval', 0.036), ('rating', 0.036), ('participants', 0.034), ('quality', 0.034), ('score', 0.034), ('likelihood', 0.032), ('bad', 0.032), ('hhtht', 0.031), ('lscore', 0.031), ('soliciting', 0.031), ('tversky', 0.031), ('xproto', 0.031), ('lowest', 0.03), ('dataset', 0.029), ('objects', 0.029), ('analogical', 0.027), ('fair', 0.027), ('categorization', 0.026), ('ranking', 0.026), ('log', 0.026), ('logarithm', 0.025), ('berkeley', 0.025), ('xij', 0.024), ('highest', 0.024), ('outperformed', 0.024), ('coast', 0.023), ('judgment', 0.023), ('denominator', 0.023), ('review', 0.022), ('testing', 0.022), ('criterion', 0.021), ('scientists', 0.021), ('evaluation', 0.021), ('labels', 0.021), ('accounts', 0.021), ('predictions', 0.02), ('detecting', 0.02), ('zoubin', 0.02), ('categories', 0.019), ('hypotheses', 0.019), ('might', 0.019), ('evaluate', 0.019), ('correlation', 0.018), ('deemed', 0.018), ('turk', 0.018), ('measuring', 0.017), ('members', 0.017), ('reasoning', 0.016), ('models', 0.016), ('label', 0.016), ('ip', 0.016), ('measures', 0.016)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.99999988 280 nips-2011-Testing a Bayesian Measure of Representativeness Using a Large Image Database

Author: Joshua T. Abbott, Katherine A. Heller, Zoubin Ghahramani, Thomas L. Griffiths

Abstract: How do people determine which elements of a set are most representative of that set? We extend an existing Bayesian measure of representativeness, which indicates the representativeness of a sample from a distribution, to define a measure of the representativeness of an item to a set. We show that this measure is formally related to a machine learning method known as Bayesian Sets. Building on this connection, we derive an analytic expression for the representativeness of objects described by a sparse vector of binary features. We then apply this measure to a large database of images, using it to determine which images are the most representative members of different sets. Comparing the resulting predictions to human judgments of representativeness provides a test of this measure with naturalistic stimuli, and illustrates how databases that are more commonly used in computer vision and machine learning can be used to evaluate psychological theories. 1

2 0.15662009 90 nips-2011-Evaluating the inverse decision-making approach to preference learning

Author: Alan Jern, Christopher G. Lucas, Charles Kemp

Abstract: Psychologists have recently begun to develop computational accounts of how people infer others’ preferences from their behavior. The inverse decision-making approach proposes that people infer preferences by inverting a generative model of decision-making. Existing data sets, however, do not provide sufficient resolution to thoroughly evaluate this approach. We introduce a new preference learning task that provides a benchmark for evaluating computational accounts and use it to compare the inverse decision-making approach to a feature-based approach, which relies on a discriminative combination of decision features. Our data support the inverse decision-making approach to preference learning. A basic principle of decision-making is that knowing people’s preferences allows us to predict how they will behave: if you know your friend likes comedies and hates horror films, you can probably guess which of these options she will choose when she goes to the theater. Often, however, we do not know what other people like and we can only infer their preferences from their behavior. If you know that a different friend saw a comedy today, does that mean that he likes comedies in general? The conclusion you draw will likely depend on what else was playing and what movie choices he has made in the past. A goal for social cognition research is to develop a computational account of people’s ability to infer others’ preferences. One computational approach is based on inverse decision-making. This approach begins with a model of how someone’s preferences lead to a decision. Then, this model is inverted to determine the most likely preferences that motivated an observed decision. An alternative approach might simply learn a functional mapping between features of an observed decision and the preferences that motivated it. For instance, in your friend’s decision to see a comedy, perhaps the more movie options he turned down, the more likely it is that he has a true preference for comedies. The difference between the inverse decision-making approach and the feature-based approach maps onto the standard dichotomy between generative and discriminative models. Economists have developed an instance of the inverse decision-making approach known as the multinomial logit model [1] that has been widely used to infer consumer’s preferences from their choices. This model has recently been explored as a psychological model [2, 3, 4], but there are few behavioral data sets for evaluating it as a model of how people learn others’ preferences. Additionally, the data sets that do exist tend to be drawn from the developmental literature, which focuses on simple tasks that collect only one or two judgments from children [5, 6, 7]. The limitations of these data sets make it difficult to evaluate the multinomial logit model with respect to alternative accounts of preference learning like the feature-based approach. In this paper, we use data from a new experimental task that elicits a detailed set of preference judgments from a single participant in order to evaluate the predictions of several preference learning models from both the inverse decision-making and feature-based classes. Our task requires each participant to sort a large number of observed decisions on the basis of how strongly they indicate 1 (a) (b) (c) d c c (d) b b a a x d x d c b a x 1. Number of chosen effects (−/+) 2. Number of forgone effects (+/+) 3. Number of forgone options (+/+) 4. Number of forgone options containing x (−/−) 5. Max/min number of effects in a forgone option (+/−) 6. Is x in every option? (−/−) 7. Chose only option with x? (+/+) 8. Is x the only difference between options? (+/+) 9. Do all options have same number of effects? (+/+) 10. Chose option with max/min number of effects? (−/−) Figure 1: (a)–(c) Examples of the decisions used in the experiments. Each column represents one option and the boxes represent different effects. The chosen option is indicated by the black rectangle. (d) Features used by the weighted feature and ranked feature models. Features 5 and 10 involved maxima in Experiment 1, which focused on all positive effects, and minima in Experiment 2, which focused on all negative effects. The signs in parentheses indicate the direction of the feature that suggests a stronger preference in Experiment 1 / Experiment 2. a preference for a chosen item. Because the number of decisions is large and these decisions vary on multiple dimensions, predicting how people will order them offers a challenging benchmark on which to compare computational models of preference learning. Data sets from these sorts of detailed tasks have proved fruitful in other domains. For example, data reported by Shepard, Hovland, and Jenkins [8]; Osherson, Smith, Wilkie, L´ pez, and Shafir [9]; and Wasserman, Elek, Chatlosh, o and Baker [10] have motivated much subsequent research on category learning, inductive reasoning, and causal reasoning, respectively. We first describe our preference learning task in detail. We then present several inverse decisionmaking and feature-based models of preference learning and compare these models’ predictions to people’s judgments in two experiments. The data are well predicted by models that follow the inverse decision-making approach, suggesting that this computational approach may help explain how people learn others’ preferences. 1 Multi-attribute decisions and revealed preferences We designed a task that can be used to elicit a large number of preference judgments from a single participant. The task involves a set of observed multi-attribute decisions, some examples of which are represented visually in Figure 1. Each decision is among a set of options and each option produces a set of effects. Figure 1 shows several decisions involving a total of five effects distributed among up to five options. The differently colored boxes represent different effects and the chosen option is marked by a black rectangle. For example, 1a shows a choice between an option with four effects and an option with a single effect; here, the decision maker chose the second option. In our task, people are asked to rank a large number of these decisions by how strongly they suggest that the decision maker had a preference for a particular effect (e.g., effect x in Figure 1). By imposing some minimal constraints, the space of unique multi-attribute decisions is finite and we can obtain rankings for every decision in the space. For example, Figure 2c shows a complete list of 47 unique decisions involving up to five effects, subject to several constraints described later. Three of these decisions are shown in Figure 1. If all the effects are positive—pieces of candy, for example—the first decision (1a) suggests a strong preference for candy x, because the decision maker turned down four pieces in favor of one. The second decision (1b), however, offers much weaker evidence because nearly everyone would choose four pieces of candy over one, even without a specific preference for x. The third decision (1c) provides evidence that is strong but perhaps not quite as strong as the first decision. When all effects are negative—like electric shocks at different body locations—decision makers may still find some effects more tolerable than others, but different inferences are sometimes supported. For example, for negative effects, 1a provides weak evidence that x is relatively tolerable because nearly everyone would choose one shock over four. 2 A computational account of preference learning We now describe a simple computational model for learning a person’s preferences after observing that person make a decision like the ones in Figure 1. We assume that there are n available options 2 {o1 , . . . , on }, each of which produces one or more effects from the set {f1 , f2 , ..., fm }. For simplicity, we assume that effects are binary. Let ui denote the utility the decision maker assigns to effect fi . We begin by specifying a model of decision-making that makes the standard assumptions that decision makers tend to choose things with greater utility and that utilities are additive. That is, if fj is a binary vector indicating the effects produced by option oj and u is a vector of utilities assigned to each of the m effects, then the total utility associated with option oj can be expressed as Uj = fj T u. We complete the specification of the model by applying the Luce choice rule [11], a common psychological model of choice behavior, as the function that chooses among the options: p(c = oj |u, f ) = exp(Uj ) = exp(Uk ) n k=1 exp(fj T u) n T k=1 exp(fk u) (1) where c denotes the choice made. This model can predict the choice someone will make among a specified set of options, given the utilities that person assigns to the effects in each option. To obtain estimates of someone’s utilities, we invert this model by applying Bayes’ rule: p(u|c, F) = p(c|u, F)p(u) p(c|F) (2) where F = {f1 , . . . , fn } specifies the available options and their corresponding effects. This is the multinomial logit model [1], a standard econometric model. In order to apply Equation 2 we must specify a prior p(u) on the utilities. We adopt a standard approach that places independent Gaussian priors on the utilities: ui ∼ N (µ, σ 2 ). For decisions where effects are positive—like candies—we set µ = 2σ, which corresponds to a prior distribution that places approximately 2% of the probability mass below zero. Similarly, for negative effects—like electric shocks—we set µ = −2σ. 2.1 Ordering a set of observed decisions Equation 2 specifies a posterior probability distribution over utilities for a single observed decision but does not provide a way to compare the inferences drawn from multiple decisions for the purposes of ordering them. Suppose we are interested in a decision maker’s preference for effect x and we wish to order a set of decisions by how strongly they support this preference. Two criteria for ordering the decisions are as follows: Absolute utility Relative utility p(c|ux , F)p(ux ) p(c|F) p(c|∀j ux ≥ uj , F)p(∀j ux ≥ uj ) p(∀j ux ≥ uj |c, F) = p(c|F) E(ux |c, F) = Eux The absolute utility model orders decisions by the mean posterior utility for effect x. This criterion is perhaps the most natural way to assess how much a decision indicates a preference for x, but it requires an inference about the utility of x in isolation, and research suggests that people often think about the utility of an effect only in relation to other salient possibilities [12]. The relative utility model applies this idea to preference learning by ordering decisions based on how strongly they suggest that x has a greater utility than all other effects. The decisions in Figures 1b and 1c are cases where the two models lead to different predictions. If the effects are all negative (e.g., electric shocks), the absolute utility model predicts that 1b provides stronger evidence for a tolerance for x because the decision maker chose to receive four shocks instead of just one. The relative utility model predicts that 1c provides stronger evidence because 1b offers no way to determine the relative tolerance of the four chosen effects with respect to one another. Like all generative models, the absolute and relative models incorporate three qualitatively different components: the likelihood term p(c|u, F), the prior p(u), and the reciprocal of the marginal likelihood 1/p(c|F). We assume that the total number of effects is fixed in advance and, as a result, the prior term will be the same for all decisions that we consider. The two other components, however, will vary across decisions. The inverse decision-making approach predicts that both components should influence preference judgments, and we will test this prediction by comparing our 3 two inverse decision-making models to two alternatives that rely only one of these components as an ordering criterion: p(c|∀j ux ≥ uj , F) 1/p(c|F) Representativeness Surprise The representativeness model captures how likely the observed decision would be if the utility for x were high, and previous research has shown that people sometimes rely on a representativeness computation of this kind [13]. The surprise model captures how unexpected the observed decision is overall; surprising decisions may be best explained in terms of a strong preference for x, but unsurprising decisions provide little information about x in particular. 2.2 Feature-based models We also consider a class of feature-based models that use surface features to order decisions. The ten features that we consider are shown in Figure 1d, where x is the effect of interest. As an example, the first feature specifies the number of effects chosen; because x is always among the chosen effects, decisions where few or no other effects belong to the chosen option suggest the strongest preference for x (when all effects are positive). This and the second feature were previously identified by Newtson [14]; we included the eight additional features shown in Figure 1d in an attempt to include all possible features that seemed both simple and relevant. We consider two methods for combining this set of features to order a set of decisions by how strongly they suggest a preference for x. The first model is a standard linear regression model, which we refer to as the weighted feature model. The model learns a weight for each feature, and the rank of a given decision is determined by a weighted sum of its features. The second model is a ranked feature model that sorts the observed decisions with respect to a strict ranking of the features. The top-ranked feature corresponds to the primary sort key, the second-ranked feature to the secondary sort key, and so on. For example, suppose that the top-ranked feature is the number of chosen effects and the second-ranked feature is the number of forgone options. Sorting the three decisions in Figure 1 according to this criterion produces the following ordering: 1a,1c,1b. This notion of sorting items on the basis of ranked features has been applied before to decision-making [15, 16] and other domains of psychology [17], but we are not aware of any previous applications to preference learning. Although our inverse decision-making and feature-based models represent two very different approaches, both may turn out to be valuable. An inverse decision-making approach may be the appropriate account of preference learning at Marr’s [18] computational level, and a feature-based approach may capture the psychological processes by which the computational-level account is implemented. Our goal, therefore, is not necessarily to accept one of these approaches and dismiss the other. Instead, we entertain three distinct possibilities. First, both approaches may account well for the data, which would support the idea that they are valid accounts operating at different levels of analysis. Second, the inverse decision-making approach may offer a better account, suggesting that process-level accounts other than the feature-based approach should be explored. Finally, the feature-based approach may offer a better account, suggesting that inverse decision-making does not constitute an appropriate computational-level account of preference learning. 3 Experiment 1: Positive effects Our first experiment focuses on decisions involving only positive effects. The full set of 47 decisions we used is shown in Figure 2c. This set includes every possible unique decision with up to five different effects, subject to the following constraints: (1) one of the effects (effect x) must always appear in the chosen option, (2) there are no repeated options, (3) each effect may appear in an option at most once, (4) only effects in the chosen option may be repeated in other options, and (5) when effects appear in multiple options, the number of effects is held constant across options. The first constraint is necessary for the sorting task, the second two constraints create a finite space of decisions, and the final two constraints limit attention to what we deemed the most interesting cases. Method 43 Carnegie Mellon undergraduates participated for course credit. Each participant was given a set of cards, with one decision printed on each card. The decisions were represented visually 4 (a) (c) Decisions 42 40 45 Mean human rankings 38 30 23 20 22 17 13 12 11 10 9 8 7 6 19 18 31 34 28 21 26 36 35 33 37 27 29 32 25 24 16 15 14 5 4 3 2 1 Absolute utility model rankings (b) Mean human rankings (Experiment 1) 47 43 44 46 45 38 37 36 34 35 30 32 33 31 29 28 24 26 27 25 21 19 22 20 18 16 17 12 13 7 6 11 5 9 4 10 8 1 2 3 42 40 41 39 47 46 44 41 43 39 23 15 14 Mean human rankings (Experiment 2) 1. dcbax 2. cbax 3. bax 4. ax 5. x 6. dcax | bcax 7. dx | cx | bx | ax 8. cax | bax 9. bdx | bcx | bax 10. dcx | bax 11. bx | ax 12. bdx | cax | bax 13. cx | bx | ax 14. d | cbax 15. c | bax 16. b | ax 17. d | c | bax 18. dc | bax 19. c | b | ax 20. dc | bx | ax 21. bdc | bax 22. ad | cx | bx | ax 23. d | c | b | ax 24. bad | bcx | bax 25. ac | bx | ax 26. cb | ax 27. cbad | cbax 28. dc | b | ax 29. ad | ac | bx | ax 30. ab | ax 31. bad | bax 32. dc | ab | ax 33. dcb | ax 34. a | x 35. bad | bac | bax 36. ac | ab | ax 37. ad | ac | ab | ax 38. b | a | x 39. ba | x 40. c | b | a | x 41. cb | a | x 42. d | c | b | a | x 43. cba | x 44. dc | ba | x 45. dc | b | a | x 46. dcb | a | x 47. dcba | x Figure 2: (a) Comparison between the absolute utility model rankings and the mean human rankings for Experiment 1. Each point represents one decision, numbered with respect to the list in panel c. (b) Comparison between the mean human rankings in Experiments 1 and 2. In both scatter plots, the solid diagonal lines indicate a perfect correspondence between the two sets of rankings. (c) The complete set of decisions, ordered by the mean human rankings from Experiment 1. Options are separated by vertical bars and the chosen option is always at the far right. Participants were always asked about a preference for effect x. as in Figure 1 but without the letter labels. Participants were told that the effects were different types of candy and each option was a bag containing one or more pieces of candy. They were asked to sort the cards by how strongly each decision suggested that the decision maker liked a particular target candy, labeled x in Figure 2c. They sorted the cards freely on a table but reported their final rankings by writing them on a sheet of paper, from weakest to strongest evidence. They were instructed to order the cards as completely as possible, but were told that they could assign the same ranking to a set of cards if they believed those cards provided equal evidence. 3.1 Results Two participants were excluded as outliers based on the criterion that their rankings for at least five decisions were at least three standard deviations from the mean rankings. We performed a hierarchical clustering analysis of the remaining 41 participants’ rankings using rank correlation as a similarity metric. Participants’ rankings were highly correlated: cutting the resulting dendrogram at 0.2 resulted in one cluster that included 33 participants and the second largest cluster included 5 Surprise MAE = 17.8 MAE = 7.0 MAE = 4.3 MAE = 17.3 MAE = 9.5 Human rankings Experiment 2 Negative effects Representativeness MAE = 2.3 MAE = 6.7 Experiment 1 Positive effects Relative utility MAE = 2.3 Human rankings Absolute utility Model rankings Model rankings Model rankings Model rankings Figure 3: Comparison between human rankings in both experiments and predicted rankings from four models. The solid diagonal lines indicate a perfect correspondence between human and model rankings. only 3 participants. Thus, we grouped all participants together and analyzed their mean rankings. The 0.2 threshold was chosen because it produced the most informative clustering in Experiment 2. Inverse decision-making models We implemented the inverse decision-making models using importance sampling with 5 million samples drawn from the prior distribution p(u). Because all the effects were positive, we used a prior on utilities that placed nearly all probability mass above zero (µ = 4, σ = 2). The mean human rankings are compared with the absolute utility model rankings in Figure 2a, and the mean human rankings are listed in order in 2c. Fractional rankings were used for both the human data and the model predictions. The human rankings in the figure are the means of participants’ fractional rankings. The first row of Figure 3 contains similar plots that allow comparison of the four models we considered. In these plots, the solid diagonal lines indicate a perfect correspondence between model and human rankings. Thus, the largest deviations from this line represent the largest deviations in the data from the model’s predictions. Figure 3 shows that the absolute and relative utility models make virtually identical predictions and both models provide a strong account of the human rankings as measured by mean absolute error (MAE = 2.3 in both cases). Moreover, both models correctly predict the highest ranked decision and the set of lowest ranked decisions. The only clear discrepancy between the model predictions and the data is the cluster of points at the lower left, labeled as Decisions 6–13 in Figure 2a. These are all cases in which effect x appears in all options and therefore these decisions provide no information about a decision maker’s preference for x. Consequently, the models assign the same ranking to this group as to the group of decisions in which there is only a single option (Decisions 1–5). Although people appeared to treat these groups somewhat differently, the models still correctly predict that the entire group of decisions 1–13 is ranked lower than all other decisions. The surprise and representativeness models do not perform nearly as well (MAE = 7.0 and 17.8, respectively). Although the surprise model captures some of the general trends in the human rankings, it makes several major errors. For example, consider Decision 7: dx|cx|bx|ax. This decision provides no information about a preference for x because it appears in every option. The decision is surprising, however, because a decision maker choosing at random from these options would make the observed choice only 1/4 of the time. The representativeness model performs even worse, primarily because it does not take into account alternative explanations for why an option was chosen, such as the fact that no other options were available (e.g., Decision 1 in Figure 2c). The failure of these models to adequately account for the data suggests that both the likelihood p(c|u, F) and marginal likelihood p(c|F) are important components of the absolute and relative utility models. Feature-based models We compared the performance of the absolute and relative utility models to our two feature-based models: the weighted feature and ranked feature models. For each participant, 6 (b) Ranked feature 10 10 5 Figure 4: Results of the feature-based model analysis from Experiment 1 for (a) the weighted feature models and (b) the ranked feature models. The histograms show the minimum number of features needed to match the accuracy (measured by MAE) of the absolute utility model for each participant. 15 5 1 2 3 4 5 6 >6 15 1 2 3 4 5 6 7 8 9 10 >10 Number of participants (a) Weighted feature Number of features needed we considered every subset of features1 in Figure 1d in order to determine the minimum number of features needed by the two models to achieve the same level of accuracy as the absolute utility model, as measured by mean absolute error. The results of these analyses are shown in Figure 4. For the majority of participants, at least four features were needed by both models to match the accuracy of the absolute utility model. For the weighted feature model, 14 participants could not be fit as well as the absolute utility model even when all ten features were considered. These results indicate that a feature-based account of people’s inferences in our task must be supplied with a relatively large number of features. By contrast, the inverse decision-making approach provides a relatively parsimonious account of the data. 4 Experiment 2: Negative effects Experiment 2 focused on a setting in which all effects are negative, motivated by the fact that the inverse decision-making models predict several major differences in orderings when effects are negative rather than positive. For instance, the absolute utility model’s relative rankings of the decisions in Figures 1a and 1b are reversed when all effects are negative rather than positive. Method 42 Carnegie Mellon undergraduates participated for course credit. The experimental design was identical to Experiment 1 except that participants were told that the effects were electric shocks at different body locations. They were asked to sort the cards on the basis of how strongly each decision suggested that the decision maker finds shocks at the target location relatively tolerable. The model predictions were derived in the same way as for Experiment 1, but with a prior distribution on utilities that placed nearly all probability mass below zero (µ = −4, σ = 2) to reflect the fact that effects were all negative. 4.1 Results Three participants were excluded as outliers by the same criterion applied in Experiment 1. The resulting mean rankings are compared with the corresponding rankings from Experiment 1 in Figure 2b. The figure shows that responses based on positive and negative effects were substantially different in a number of cases. Figure 3 shows how the mean rankings compare to the predictions of the four models we considered. Although the relative utility model is fairly accurate, no model achieves the same level of accuracy as the absolute and relative utility models in Experiment 1. In addition, the relative utility model provides a poor account of the responses of many individual participants. To better understand responses at the individual level, we repeated the hierarchical clustering analysis described in Experiment 1, which revealed that 29 participants could be grouped into one of four clusters, with the remaining participants each in their own clusters. We analyzed these four clusters independently, excluding the 10 participants that could not be naturally grouped. We compared the mean rankings of each cluster to the absolute and relative utility models, as well as all one- and two-feature weighted feature and ranked feature models. Figure 5 shows that the mean rankings of participants in Cluster 1 (N = 8) were best fit by the absolute utility model, the mean rankings of participants in Cluster 2 (N = 12) were best fit by the relative utility model, and the mean rankings of participants in Clusters 3 (N = 3) and 4 (N = 6) were better fit by feature-based models than by either the absolute or relative utility models. 1 A maximum of six features was considered for the ranked feature model because considering more features was computationally intractable. 7 Cluster 4 N =6 MAE = 4.9 MAE = 14.0 MAE = 7.9 MAE = 5.3 MAE = 2.6 MAE = 13.0 MAE = 6.2 Human rankings Relative utility Cluster 3 N =3 MAE = 2.6 Absolute utility Cluster 2 N = 12 Human rankings Cluster 1 N =8 Factors: 1,3 Factors: 1,8 MAE = 2.3 MAE = 5.2 Model rankings Best−fitting weighted feature Factors: 6,7 MAE = 4.0 Model rankings Model rankings Model rankings Human rankings Factors: 3,8 MAE = 4.8 Figure 5: Comparison between human rankings for four clusters of participants identified in Experiment 2 and predicted rankings from three models. Each point in the plots corresponds to one decision and the solid diagonal lines indicate a perfect correspondence between human and model rankings. The third row shows the predictions of the best-fitting two-factor weighted feature model for each cluster. The two factors listed refer to Figure 1d. To examine how well the models accounted for individuals’ rankings within each cluster, we compared the predictions of the inverse decision-making models to the best-fitting two-factor featurebased model for each participant. In Cluster 1, 7 out of 8 participants were best fit by the absolute utility model; in Cluster 2, 8 out of 12 participants were best fit by the relative utility model; in Clusters 3 and 4, all participants were better fit by feature-based models. No single feature-based model provided the best fit for more than two participants, suggesting that participants not fit well by the inverse decision-making models were not using a single alternative strategy. Applying the feature-based model analysis from Experiment 1 to the current results revealed that the weighted feature model required an average of 6.0 features to match the performance of the absolute utility model for participants in Cluster 1, and an average of 3.9 features to match the performance of the relative utility model for participants in Cluster 2. Thus, although a single model did not fit all participants well in the current experiment, many participants were fit well by one of the two inverse decision-making models, suggesting that this general approach is useful for explaining how people reason about negative effects as well as positive effects. 5 Conclusion In two experiments, we found that an inverse decision-making approach offered a good computational account of how people make judgments about others’ preferences. Although this approach is conceptually simple, our analyses indicated that it captures the influence of a fairly large number of relevant decision features. Indeed, the feature-based models that we considered as potential process models of preference learning could only match the performance of the inverse decision-making approach when supplied with a relatively large number of features. We feel that this result rules out the feature-based approach as psychologically implausible, meaning that alternative process-level accounts will need to be explored. One possibility is sampling, which has been proposed as a psychological mechanism for approximating probabilistic inferences [19, 20]. However, even if process models that use large numbers of features are considered plausible, the inverse decision-making approach provides a valuable computational-level account that helps to explain which decision features are informative. Acknowledgments This work was supported in part by the Pittsburgh Life Sciences Greenhouse Opportunity Fund and by NSF grant CDI-0835797. 8 References [1] D. McFadden. Conditional logit analysis of qualitative choice behavior. In P. Zarembka, editor, Frontiers in Econometrics. Amademic Press, New York, 1973. [2] C. G. Lucas, T. L. Griffiths, F. Xu, and C. Fawcett. A rational model of preference learning and choice prediction by children. In Proceedings of Neural Information Processing Systems 21, 2009. [3] L. Bergen, O. R. Evans, and J. B. Tenenbaum. Learning structured preferences. In Proceedings of the 32nd Annual Conference of the Cognitive Science Society, 2010. [4] A. Jern and C. Kemp. Decision factors that support preference learning. In Proceedings of the 33rd Annual Conference of the Cognitive Science Society, 2011. [5] T. Kushnir, F. Xu, and H. M. Wellman. Young children use statistical sampling to infer the preferences of other people. Psychological Science, 21(8):1134–1140, 2010. [6] L. Ma and F. Xu. Young children’s use of statistical sampling evidence to infer the subjectivity of preferences. Cognition, in press. [7] M. J. Doherty. Theory of Mind: How Children Understand Others’ Thoughts and Feelings. Psychology Press, New York, 2009. [8] R. N. Shepard, C. I. Hovland, and H. M. Jenkins. Learning and memorization of classifications. Psychological Monographs, 75, Whole No. 517, 1961. [9] D. N. Osherson, E. E. Smith, O. Wilkie, A. L´ pez, and E. Shafir. Category-based induction. Psychological o Review, 97(2):185–200, 1990. [10] E. A. Wasserman, S. M. Elek, D. L. Chatlosh, and A. G. Baker. Rating causal relations: Role of probability in judgments of response-outcome contingency. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19(1):174–188, 1993. [11] R. D. Luce. Individual choice behavior. John Wiley, 1959. [12] D. Ariely, G. Loewenstein, and D. Prelec. Tom Sawyer and the construction of value. Journal of Economic Behavior & Organization, 60:1–10, 2006. [13] D. Kahneman and A. Tversky. Subjective probability: A judgment of representativeness. Cognitive Psychology, 3(3):430–454, 1972. [14] D. Newtson. Dispositional inference from effects of actions: Effects chosen and effects forgone. Journal of Experimental Social Psychology, 10:489–496, 1974. [15] P. C. Fishburn. Lexicographic orders, utilities and decision rules: A survey. Management Science, 20(11):1442–1471, 1974. [16] G. Gigerenzer and P. M. Todd. Fast and frugal heuristics: The adaptive toolbox. Oxford University Press, New York, 1999. [17] A. Prince and P. Smolensky. Optimality Theory: Constraint Interaction in Generative Grammar. WileyBlackwell, 2004. [18] D. Marr. Vision. W. H. Freeman, San Francisco, 1982. [19] A. N. Sanborn, T. L. Griffiths, and D. J. Navarro. Rational approximations to rational models: Alternative algorithms for category learning. Psychological Review, 117:1144–1167, 2010. [20] L. Shi and T. L. Griffiths. Neural implementation of Bayesian inference by importance sampling. In Proceedings of Neural Information Processing Systems 22, 2009. 9

3 0.097895995 34 nips-2011-An Unsupervised Decontamination Procedure For Improving The Reliability Of Human Judgments

Author: Michael C. Mozer, Benjamin Link, Harold Pashler

Abstract: Psychologists have long been struck by individuals’ limitations in expressing their internal sensations, impressions, and evaluations via rating scales. Instead of using an absolute scale, individuals rely on reference points from recent experience. This relativity of judgment limits the informativeness of responses on surveys, questionnaires, and evaluation forms. Fortunately, the cognitive processes that map stimuli to responses are not simply noisy, but rather are influenced by recent experience in a lawful manner. We explore techniques to remove sequential dependencies, and thereby decontaminate a series of ratings to obtain more meaningful human judgments. In our formulation, the problem is to infer latent (subjective) impressions from a sequence of stimulus labels (e.g., movie names) and responses. We describe an unsupervised approach that simultaneously recovers the impressions and parameters of a contamination model that predicts how recent judgments affect the current response. We test our iterated impression inference, or I3 , algorithm in three domains: rating the gap between dots, the desirability of a movie based on an advertisement, and the morality of an action. We demonstrate significant objective improvements in the quality of the recovered impressions. 1

4 0.089425541 130 nips-2011-Inductive reasoning about chimeric creatures

Author: Charles Kemp

Abstract: Given one feature of a novel animal, humans readily make inferences about other features of the animal. For example, winged creatures often fly, and creatures that eat fish often live in the water. We explore the knowledge that supports these inferences and compare two approaches. The first approach proposes that humans rely on abstract representations of dependency relationships between features, and is formalized here as a graphical model. The second approach proposes that humans rely on specific knowledge of previously encountered animals, and is formalized here as a family of exemplar models. We evaluate these models using a task where participants reason about chimeras, or animals with pairs of features that have not previously been observed to co-occur. The results support the hypothesis that humans rely on explicit representations of relationships between features. Suppose that an eighteenth-century naturalist learns about a new kind of animal that has fur and a duck’s bill. Even though the naturalist has never encountered an animal with this pair of features, he should be able to make predictions about other features of the animal—for example, the animal could well live in water but probably does not have feathers. Although the platypus exists in reality, from a eighteenth-century perspective it qualifies as a chimera, or an animal that combines two or more features that have not previously been observed to co-occur. Here we describe a probabilistic account of inductive reasoning and use it to account for human inferences about chimeras. The inductive problems we consider are special cases of the more general problem in Figure 1a where a reasoner is given a partially observed matrix of animals by features then asked to infer the values of the missing entries. This general problem has been previously studied and is addressed by computational models of property induction, categorization, and generalization [1–7]. A challenge faced by all of these models is to capture the background knowledge that guides inductive inferences. Some accounts rely on similarity relationships between animals [6, 8], others rely on causal relationships between features [9, 10], and others incorporate relationships between animals and relationships between features [11]. We will evaluate graphical models that capture both kinds of relationships (Figure 1a), but will focus in particular on relationships between features. Psychologists have previously suggested that humans rely on explicit mental representations of relationships between features [12–16]. Often these representations are described as theories—for example, theories that specify a causal relationship between having wings and flying, or living in the sea and eating fish. Relationships between features may take several forms: for example, one feature may cause, enable, prevent, be inconsistent with, or be a special case of another feature. For simplicity, we will treat all of these relationships as instances of dependency relationships between features, and will capture them using an undirected graphical model. Previous studies have used graphical models to account for human inferences about features but typically these studies consider toy problems involving a handful of novel features such as “has gene X14” or “has enzyme Y132” [9, 11]. Participants might be told, for example, that gene X14 leads to the production of enzyme Y132, then asked to use this information when reasoning about novel animals. Here we explore whether a graphical model approach can account for inferences 1 (a) slow heavy flies (b) wings hippo 1 1 0 0 rhino 1 1 0 0 sparrow 0 0 1 1 robin 0 0 1 1 new ? ? 1 ? o Figure 1: Inductive reasoning about animals and features. (a) Inferences about the features of a new animal onew that flies may draw on similarity relationships between animals (the new animal is similar to sparrows and robins but not hippos and rhinos), and on dependency relationships between features (flying and having wings are linked). (b) A graph product produced by combining the two graph structures in (a). about familiar features. Working with familiar features raises a methodological challenge since participants have a substantial amount of knowledge about these features and can reason about them in multiple ways. Suppose, for example, that you learn that a novel animal can fly (Figure 1a). To conclude that the animal probably has wings, you might consult a mental representation similar to the graph at the top of Figure 1a that specifies a dependency relationship between flying and having wings. On the other hand, you might reach the same conclusion by thinking about flying creatures that you have previously encountered (e.g. sparrows and robins) and noticing that these creatures have wings. Since the same conclusion can be reached in two different ways, judgments about arguments of this kind provide little evidence about the mental representations involved. The challenge of working with familiar features directly motivates our focus on chimeras. Inferences about chimeras draw on rich background knowledge but require the reasoner to go beyond past experience in a fundamental way. For example, if you learn that an animal flies and has no legs, you cannot make predictions about the animal by thinking of flying, no-legged creatures that you have previously encountered. You may, however, still be able to infer that the novel animal has wings if you understand the relationship between flying and having wings. We propose that graphical models over features can help to explain how humans make inferences of this kind, and evaluate our approach by comparing it to a family of exemplar models. The next section introduces these models, and we then describe two experiments designed to distinguish between the models. 1 Reasoning about objects and features Our models make use of a binary matrix D where the rows {o1 , . . . , o129 } correspond to objects, and the columns {f 1 , . . . , f 56 } correspond to features. A subset of the objects is shown in Figure 2a, and the full set of features is shown in Figure 2b and its caption. Matrix D was extracted from the Leuven natural concept database [17], which includes 129 animals and 757 features in total. We chose a subset of these features that includes a mix of perceptual and behavioral features, and that includes many pairs of features that depend on each other. For example, animals that “live in water” typically “can swim,” and animals that have “no legs” cannot “jump far.” Matrix D can be used to formulate problems where a reasoner observes one or two features of a new object (i.e. animal o130 ) and must make inferences about the remaining features of the animal. The next two sections describe graphical models that can be used to address this problem. The first graphical model O captures relationships between objects, and the second model F captures relationships between features. We then discuss how these models can be combined, and introduce a family of exemplar-style models that will be compared with our graphical models. A graphical model over objects Many accounts of inductive reasoning focus on similarity relationships between objects [6, 8]. Here we describe a tree-structured graphical model O that captures these relationships. The tree was constructed from matrix D using average linkage clustering and the Jaccard similarity measure, and part of the resulting structure is shown in Figure 2a. The subtree in Figure 2a includes clusters 2 alligator caiman crocodile monitor lizard dinosaur blindworm boa cobra python snake viper chameleon iguana gecko lizard salamander frog toad tortoise turtle anchovy herring sardine cod sole salmon trout carp pike stickleback eel flatfish ray plaice piranha sperm whale squid swordfish goldfish dolphin orca whale shark bat fox wolf beaver hedgehog hamster squirrel mouse rabbit bison elephant hippopotamus rhinoceros lion tiger polar bear deer dromedary llama giraffe zebra kangaroo monkey cat dog cow horse donkey pig sheep (a) (b) can swim lives in water eats fish eats nuts eats grain eats grass has gills can jump far has two legs has no legs has six legs has four legs can fly can be ridden has sharp teeth nocturnal has wings strong predator can see in dark eats berries lives in the sea lives in the desert crawls lives in the woods has mane lives in trees can climb well lives underground has feathers has scales slow has fur heavy Figure 2: Graph structures used to define graphical models O and F. (a) A tree that captures similarity relationships between animals. The full tree includes 129 animals, and only part of the tree is shown here. The grey points along the branches indicate locations where a novel animal o130 could be attached to the tree. (b) A network capturing pairwise dependency relationships between features. The edges capture both positive and negative dependencies. All edges in the network are shown, and the network also includes 20 isolated nodes for the following features: is black, is blue, is green, is grey, is pink, is red, is white, is yellow, is a pet, has a beak, stings, stinks, has a long neck, has feelers, sucks blood, lays eggs, makes a web, has a hump, has a trunk, and is cold-blooded. corresponding to amphibians and reptiles, aquatic creatures, and land mammals, and the subtree omitted for space includes clusters for insects and birds. We assume that the features in matrix D (i.e. the columns) are generated independently over O: P (f i |O, π i , λi ). P (D|O, π, λ) = i i i i The distribution P (f |O, π , λ ) is based on the intuition that nearby nodes in O tend to have the same value of f i . Previous researchers [8, 18] have used a directed graphical model where the distribution at the root node is based on the baserate π i , and any other node v with parent u has the following conditional probability distribution: i P (v = 1|u) = π i + (1 − π i )e−λ l , if u = 1 i π i − π i e−λ l , if u = 0 (1) where l is the length of the branch joining node u to node v. The variability parameter λi captures the extent to which feature f i is expected to vary over the tree. Note, for example, that any node v must take the same value as its parent u when λ = 0. To avoid free parameters, the feature baserates π i and variability parameters λi are set to their maximum likelihood values given the observed values of the features {f i } in the data matrix D. The conditional distributions in Equation 1 induce a joint distribution over all of the nodes in graph O, and the distribution P (f i |O, π i , λi ) is computed by marginalizing out the values of the internal nodes. Although we described O as a directed graphical model, the model can be converted into an equivalent undirected model with a potential for each edge in the tree and a potential for the root node. Here we use the undirected version of the model, which is a natural counterpart to the undirected model F described in the next section. The full version of structure O in Figure 2a includes 129 familiar animals, and our task requires inferences about a novel animal o130 that must be slotted into the structure. Let D′ be an expanded version of D that includes a row for o130 , and let O′ be an expanded version of O that includes a node for o130 . The edges in Figure 2a are marked with evenly spaced gray points, and we use a 3 uniform prior P (O′ ) over all trees that can be created by attaching o130 to one of these points. Some of these trees have identical topologies, since some edges in Figure 2a have multiple gray points. Predictions about o130 can be computed using: P (D′ |D) = P (D′ |O′ , D)P (O′ |D) ∝ O′ P (D′ |O′ , D)P (D|O′ )P (O′ ). (2) O′ Equation 2 captures the basic intuition that the distribution of features for o130 is expected to be consistent with the distribution observed for previous animals. For example, if o130 is known to fly then the trees with high posterior probability P (O′ |D) will be those where o130 is near other flying creatures (Figure 1a), and since these creatures have wings Equation 2 predicts that o130 probably also has wings. As this example suggests, model O captures dependency relationships between features implicitly, and therefore stands in contrast to models like F that rely on explicit representations of relationships between features. A graphical model over features Model F is an undirected graphical model defined over features. The graph shown in Figure 2b was created by identifying pairs where one feature depends directly on another. The author and a research assistant both independently identified candidate sets of pairwise dependencies, and Figure 2b was created by merging these sets and reaching agreement about how to handle any discrepancies. As previous researchers have suggested [13, 15], feature dependencies can capture several kinds of relationships. For example, wings enable flying, living in the sea leads to eating fish, and having no legs rules out jumping far. We work with an undirected graph because some pairs of features depend on each other but there is no clear direction of causal influence. For example, there is clearly a dependency relationship between being nocturnal and seeing in the dark, but no obvious sense in which one of these features causes the other. We assume that the rows of the object-feature matrix D are generated independently from an undirected graphical model F defined over the feature structure in Figure 2b: P (oi |F). P (D|F) = i Model F includes potential functions for each node and for each edge in the graph. These potentials were learned from matrix D using the UGM toolbox for undirected graphical models [19]. The learned potentials capture both positive and negative relationships: for example, animals that live in the sea tend to eat fish, and tend not to eat berries. Some pairs of feature values never occur together in matrix D (there are no creatures that fly but do not have wings). We therefore chose to compute maximum a posteriori values of the potential functions rather than maximum likelihood values, and used a diffuse Gaussian prior with a variance of 100 on the entries in each potential. After learning the potentials for model F, we can make predictions about a new object o130 using the distribution P (o130 |F). For example, if o130 is known to fly (Figure 1a), model F predicts that o130 probably has wings because the learned potentials capture a positive dependency between flying and having wings. Combining object and feature relationships There are two simple ways to combine models O and F in order to develop an approach that incorporates both relationships between features and relationships between objects. The output combination model computes the predictions of both models in isolation, then combines these predictions using a weighted sum. The resulting model is similar to a mixture-of-experts model, and to avoid free parameters we use a mixing weight of 0.5. The structure combination model combines the graph structures used by the two models and relies on a set of potentials defined over the resulting graph product. An example of a graph product is shown in Figure 1b, and the potential functions for this graph are inherited from the component models in the natural way. Kemp et al. [11] use a similar approach to combine a functional causal model with an object model O, but note that our structure combination model uses an undirected model F rather than a functional causal model over features. Both combination models capture the intuition that inductive inferences rely on relationships between features and relationships between objects. The output combination model has the virtue of 4 simplicity, and the structure combination model is appealing because it relies on a single integrated representation that captures both relationships between features and relationships between objects. To preview our results, our data suggest that the combination models perform better overall than either O or F in isolation, and that both combination models perform about equally well. Exemplar models We will compare the family of graphical models already described with a family of exemplar models. The key difference between these model families is that the exemplar models do not rely on explicit representations of relationships between objects and relationships between features. Comparing the model families can therefore help to establish whether human inferences rely on representations of this sort. Consider first a problem where a reasoner must predict whether object o130 has feature k after observing that it has feature i. An exemplar model addresses the problem by retrieving all previouslyobserved objects with feature i and computing the proportion that have feature k: P (ok = 1|oi = 1) = |f k & f i | |f i | (3) where |f k | is the number of objects in matrix D that have feature k, and |f k & f i | is the number that have both feature k and feature i. Note that we have streamlined our notation by using ok instead of o130 to refer to the kth feature value for object o130 . k Suppose now that the reasoner observes that object o130 has features i and j. The natural generalization of Equation 3 is: P (ok = 1|oi = 1, oj = 1) = |f k & f i & f j | |f i & f j | (4) Because we focus on chimeras, |f i & f j | = 0 and Equation 4 is not well defined. We therefore evaluate an exemplar model that computes predictions for the two observed features separately then computes the weighted sum of these predictions: P (ok = 1|oi = 1, oj = 1) = wi |f k & f i | |f k & f j | + wj . i| |f |f j | (5) where the weights wi and wj must sum to one. We consider four ways in which the weights could be set. The first strategy sets wi = wj = 0.5. The second strategy sets wi ∝ |f i |, and is consistent with an approach where the reasoner retrieves all exemplars in D that are most similar to the novel animal and reports the proportion of these exemplars that have feature k. The third strategy sets wi ∝ |f1i | , and captures the idea that features should be weighted by their distinctiveness [20]. The final strategy sets weights according to the coherence of each feature [21]. A feature is coherent if objects with that feature tend to resemble each other overall, and we define the coherence of feature i as the expected Jaccard similarity between two randomly chosen objects from matrix D that both have feature i. Note that the final three strategies are all consistent with previous proposals from the psychological literature, and each one might be expected to perform well. Because exemplar models and prototype models are often compared, it is natural to consider a prototype model [22] as an additional baseline. A standard prototype model would partition the 129 animals into categories and would use summary statistics for these categories to make predictions about the novel animal o130 . We will not evaluate this model because it corresponds to a coarser version of model O, which organizes the animals into a hierarchy of categories. The key characteristic shared by both models is that they explicitly capture relationships between objects but not features. 2 Experiment 1: Chimeras Our first experiment explores how people make inferences about chimeras, or novel animals with features that have not previously been observed to co-occur. Inferences about chimeras raise challenges for exemplar models, and therefore help to establish whether humans rely on explicit representations of relationships between features. Each argument can be represented as f i , f j → f k 5 exemplar r = 0.42 7 feature F exemplar (wi = |f i |) (wi = 0.5) r = 0.44 7 object O r = 0.69 7 output combination r = 0.31 7 structure combination r = 0.59 7 r = 0.60 7 5 5 5 5 5 3 3 3 3 3 3 all 5 1 1 0 1 r = 0.06 7 conflict 0.5 1 1 0 0.5 1 r = 0.71 7 1 0 0.5 1 r = −0.02 7 1 0 0.5 1 r = 0.49 7 0 5 5 5 5 3 3 3 3 1 5 3 0.5 r = 0.57 7 5 3 1 0 0.5 1 r = 0.51 7 edge 0.5 r = 0.17 7 1 1 0 0.5 1 r = 0.64 7 1 0 0.5 1 r = 0.83 7 1 0 0.5 1 r = 0.45 7 1 0 0.5 1 r = 0.76 7 0 5 5 5 5 3 3 3 3 1 5 3 0.5 r = 0.79 7 5 3 1 1 0 0.5 1 r = 0.26 7 other 1 0 1 0 0.5 1 r = 0.25 7 1 0 0.5 1 r = 0.19 7 1 0 0.5 1 r = 0.25 7 1 0 0.5 1 r = 0.24 7 0 7 5 5 5 5 5 3 3 3 3 1 5 3 0.5 r = 0.33 3 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 0 0.5 1 Figure 3: Argument ratings for Experiment 1 plotted against the predictions of six models. The y-axis in each panel shows human ratings on a seven point scale, and the x-axis shows probabilities according to one of the models. Correlation coefficients are shown for each plot. where f i and f k are the premises (e.g. “has no legs” and “can fly”) and f k is the conclusion (e.g. “has wings”). We are especially interested in conflict cases where the premises f i and f j lead to opposite conclusions when taken individually: for example, most animals with no legs do not have wings, but most animals that fly do have wings. Our models that incorporate feature structure F can resolve this conflict since F includes a dependency between “wings” and “can fly” but not between “wings” and “has no legs.” Our models that do not include F cannot resolve the conflict and predict that humans will be uncertain about whether the novel animal has wings. Materials. The object-feature matrix D includes 447 feature pairs {f i , f j } such that none of the 129 animals has both f i and f j . We selected 40 pairs (see the supporting material) and created 400 arguments in total by choosing 10 conclusion features for each pair. The arguments can be assigned to three categories. Conflict cases are arguments f i , f j → f k such that the single-premise arguments f i → f k and f j → f k lead to incompatible predictions. For our purposes, two singlepremise arguments with the same conclusion are deemed incompatible if one leads to a probability greater than 0.9 according to Equation 3, and the other leads to a probability less than 0.1. Edge cases are arguments f i , f j → f k such that the feature network in Figure 2b includes an edge between f k and either f i or f j . Note that some arguments are both conflict cases and edge cases. All arguments that do not fall into either one of these categories will be referred to as other cases. The 400 arguments for the experiment include 154 conflict cases, 153 edge cases, and 120 other cases. 34 arguments are both conflict cases and edge cases. We chose these arguments based on three criteria. First, we avoided premise pairs that did not co-occur in matrix D but that co-occur in familiar animals that do not belong to D. For example, “is pink” and “has wings” do not co-occur in D but “flamingo” is a familiar animal that has both features. Second, we avoided premise pairs that specified two different numbers of legs—for example, {“has four legs,” “has six legs”}. Finally, we aimed to include roughly equal numbers of conflict cases, edge cases, and other cases. Method. 16 undergraduates participated for course credit. The experiment was carried out using a custom-built computer interface, and one argument was presented on screen at a time. Participants 6 rated the probability of the conclusion on seven point scale where the endpoints were labeled “very unlikely” and “very likely.” The ten arguments for each pair of premises were presented in a block, but the order of these blocks and the order of the arguments within these blocks were randomized across participants. Results. Figure 3 shows average human judgments plotted against the predictions of six models. The plots in the first row include all 400 arguments in the experiment, and the remaining rows show results for conflict cases, edge cases, and other cases. The previous section described four exemplar models, and the two shown in Figure 3 are the best performers overall. Even though the graphical models include more numerical parameters than the exemplar models, recall that these parameters are learned from matrix D rather than fit to the experimental data. Matrix D also serves as the basis for the exemplar models, which means that all of the models can be compared on equal terms. The first row of Figure 3 suggests that the three models which include feature structure F perform better than the alternatives. The output combination model is the worst of the three models that incorporate F, and the correlation achieved by this model is significantly greater than the correlation achieved by the best exemplar model (p < 0.001, using the Fisher transformation to convert correlation coefficients to z scores). Our data therefore suggest that explicit representations of relationships between features are needed to account for inductive inferences about chimeras. The model that includes the feature structure F alone performs better than the two models that combine F with the object structure O, which may not be surprising since Experiment 1 focuses specifically on novel animals that do not slot naturally into structure O. Rows two through four suggest that the conflict arguments in particular raise challenges for the models which do not include feature structure F. Since these conflict cases are arguments f i , f j → f k where f i → f k has strength greater than 0.9 and f j → f k has strength less than 0.1, the first exemplar model averages these strengths and assigns an overall strength of around 0.5 to each argument. The second exemplar model is better able to differentiate between the conflict arguments, but still performs substantially worse than the three models that include structure F. The exemplar models perform better on the edge arguments, but are outperformed by the models that include F. Finally, all models achieve roughly the same level of performance on the other arguments. Although the feature model F performs best overall, the predictions of this model still leave room for improvement. The two most obvious outliers in the third plot in the top row represent the arguments {is blue, lives in desert → lives in woods} and {is pink, lives in desert → lives in woods}. Our participants sensibly infer that any animal which lives in the desert cannot simultaneously live in the woods. In contrast, the Leuven database indicates that eight of the twelve animals that live in the desert also live in the woods, and the edge in Figure 2b between “lives in the desert” and “lives in the woods” therefore represents a positive dependency relationship according to model F. This discrepancy between model and participants reflects the fact that participants made inferences about individual animals but the Leuven database is based on features of animal categories. Note, for example, that any individual animal is unlikely to live in the desert and the woods, but that some animal categories (including snakes, salamanders, and lizards) are found in both environments. 3 Experiment 2: Single-premise arguments Our results so far suggest that inferences about chimeras rely on explicit representations of relationships between features but provide no evidence that relationships between objects are important. It would be a mistake, however, to conclude that relationships between objects play no role in inductive reasoning. Previous studies have used object structures like the example in Figure 2a to account for inferences about novel features [11]—for example, given that alligators have enzyme Y132 in their blood, it seems likely that crocodiles also have this enzyme. Inferences about novel objects can also draw on relationships between objects rather than relationships between features. For example, given that a novel animal has a beak you will probably predict that it has feathers, not because there is any direct dependency between these two features, but because the beaked animals that you know tend to have feathers. Our second experiment explores inferences of this kind. Materials and Method. 32 undergraduates participated for course credit. The task was identical to Experiment 1 with the following exceptions. Each two-premise argument f i , f j → f k from Experiment 1 was converted into two one-premise arguments f i → f k and f j → f k , and these 7 feature F exemplar r = 0.78 7 object O r = 0.54 7 output combination r = 0.75 7 structure combination r = 0.75 7 all 5 5 5 5 5 3 3 3 3 3 1 1 0 edge 0.5 1 r = 0.87 7 1 0 0.5 1 r = 0.87 7 1 0 0.5 1 r = 0.84 7 1 0 0.5 1 r = 0.86 7 0 5 5 5 3 3 3 1 5 3 0.5 r = 0.85 7 5 3 1 1 0 0.5 1 r = 0.79 7 other r = 0.77 7 1 0 0.5 1 r = 0.21 7 1 0 0.5 1 r = 0.74 7 1 0 0.5 1 r = 0.66 7 0 5 5 5 5 3 3 3 3 1 r = 0.73 7 5 0.5 3 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 0 0.5 1 Figure 4: Argument ratings and model predictions for Experiment 2. one-premise arguments were randomly assigned to two sets. 16 participants rated the 400 arguments in the first set, and the other 16 rated the 400 arguments in the second set. Results. Figure 4 shows average human ratings for the 800 arguments plotted against the predictions of five models. Unlike Figure 3, Figure 4 includes a single exemplar model since there is no need to consider different feature weightings in this case. Unlike Experiment 1, the feature model F performs worse than the other alternatives (p < 0.001 in all cases). Not surprisingly, this model performs relatively well for edge cases f j → f k where f j and f k are linked in Figure 2b, but the final row shows that the model performs poorly across the remaining set of arguments. Taken together, Experiments 1 and 2 suggest that relationships between objects and relationships between features are both needed to account for human inferences. Experiment 1 rules out an exemplar approach but models that combine graph structures over objects and features perform relatively well in both experiments. We considered two methods for combining these structures and both performed equally well. Combining the knowledge captured by these structures appears to be important, and future studies can explore in detail how humans achieve this combination. 4 Conclusion This paper proposed that graphical models are useful for capturing knowledge about animals and their features and showed that a graphical model over features can account for human inferences about chimeras. A family of exemplar models and a graphical model defined over objects were unable to account for our data, which suggests that humans rely on mental representations that explicitly capture dependency relationships between features. Psychologists have previously used graphical models to capture relationships between features, but our work is the first to focus on chimeras and to explore models defined over a large set of familiar features. Although a simple undirected model accounted relatively well for our data, this model is only a starting point. The model incorporates dependency relationships between features, but people know about many specific kinds of dependencies, including cases where one feature causes, enables, prevents, or is inconsistent with another. An undirected graph with only one class of edges cannot capture this knowledge in full, and richer representations will ultimately be needed in order to provide a more complete account of human reasoning. Acknowledgments I thank Madeleine Clute for assisting with this research. This work was supported in part by the Pittsburgh Life Sciences Greenhouse Opportunity Fund and by NSF grant CDI-0835797. 8 References [1] R. N. Shepard. Towards a universal law of generalization for psychological science. Science, 237:1317– 1323, 1987. [2] J. R. Anderson. The adaptive nature of human categorization. Psychological Review, 98(3):409–429, 1991. [3] E. Heit. A Bayesian analysis of some forms of inductive reasoning. In M. Oaksford and N. Chater, editors, Rational models of cognition, pages 248–274. Oxford University Press, Oxford, 1998. [4] J. B. Tenenbaum and T. L. Griffiths. Generalization, similarity, and Bayesian inference. Behavioral and Brain Sciences, 24:629–641, 2001. [5] C. Kemp and J. B. Tenenbaum. Structured statistical models of inductive reasoning. Psychological Review, 116(1):20–58, 2009. [6] D. N. Osherson, E. E. Smith, O. Wilkie, A. Lopez, and E. Shafir. Category-based induction. Psychological Review, 97(2):185–200, 1990. [7] D. J. Navarro. Learning the context of a category. In J. Lafferty, C. K. I. Williams, J. Shawe-Taylor, R.S. Zemel, and A. Culotta, editors, Advances in Neural Information Processing Systems 23, pages 1795–1803. 2010. [8] C. Kemp, T. L. Griffiths, S. Stromsten, and J. B. Tenenbaum. Semi-supervised learning with trees. In Advances in Neural Information Processing Systems 16, pages 257–264. MIT Press, Cambridge, MA, 2004. [9] B. Rehder. A causal-model theory of conceptual representation and categorization. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29:1141–1159, 2003. [10] B. Rehder and R. Burnett. Feature inference and the causal structure of categories. Cognitive Psychology, 50:264–314, 2005. [11] C. Kemp, P. Shafto, and J. B. Tenenbaum. An integrated account of generalization across objects and features. Cognitive Psychology, in press. [12] S. E. Barrett, H. Abdi, G. L. Murphy, and J. McCarthy Gallagher. Theory-based correlations and their role in children’s concepts. Child Development, 64:1595–1616, 1993. [13] S. A. Sloman, B. C. Love, and W. Ahn. Feature centrality and conceptual coherence. Cognitive Science, 22(2):189–228, 1998. [14] D. Yarlett and M. Ramscar. A quantitative model of counterfactual reasoning. In T. G. Dietterich, S. Becker, and Z. Ghahramani, editors, Advances in Neural Information Processing Systems 14, pages 123–130. MIT Press, Cambridge, MA, 2002. [15] W. Ahn, J. K. Marsh, C. C. Luhmann, and K. Lee. Effect of theory-based feature correlations on typicality judgments. Memory and Cognition, 30(1):107–118, 2002. [16] D. C. Meehan C. McNorgan, R. A. Kotack and K. McRae. Feature-feature causal relations and statistical co-occurrences in object concepts. Memory and Cognition, 35(3):418–431, 2007. [17] S. De Deyne, S. Verheyen, E. Ameel, W. Vanpaemel, M. J. Dry, W. Voorspoels, and G. Storms. Exemplar by feature applicability matrices and other Dutch normative data for semantic concepts. Behavior Research Methods, 40(4):1030–1048, 2008. [18] J. P. Huelsenbeck and F. Ronquist. MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics, 17(8):754–755, 2001. [19] M. Schmidt. UGM: A Matlab toolbox for probabilistic undirected graphical models. 2007. Available at http://people.cs.ubc.ca/∼schmidtm/Software/UGM.html. [20] L. J. Nelson and D. T. Miller. The distinctiveness effect in social categorization: you are what makes you unusual. Psychological Science, 6:246–249, 1995. [21] A. L. Patalano, S. Chin-Parker, and B. H. Ross. The importance of being coherent: category coherence, cross-classification and reasoning. Journal of memory and language, 54:407–424, 2006. [22] S. K. Reed. Pattern recognition and categorization. Cognitive Psychology, 3:393–407, 1972. 9

5 0.084837317 15 nips-2011-A rational model of causal inference with continuous causes

Author: Thomas L. Griffiths, Michael James

Abstract: Rational models of causal induction have been successful in accounting for people’s judgments about causal relationships. However, these models have focused on explaining inferences from discrete data of the kind that can be summarized in a 2× 2 contingency table. This severely limits the scope of these models, since the world often provides non-binary data. We develop a new rational model of causal induction using continuous dimensions, which aims to diminish the gap between empirical and theoretical approaches and real-world causal induction. This model successfully predicts human judgments from previous studies better than models of discrete causal inference, and outperforms several other plausible models of causal induction with continuous causes in accounting for people’s inferences in a new experiment. 1

6 0.082381248 66 nips-2011-Crowdclustering

7 0.072656989 126 nips-2011-Im2Text: Describing Images Using 1 Million Captioned Photographs

8 0.071731411 141 nips-2011-Large-Scale Category Structure Aware Image Categorization

9 0.071679138 10 nips-2011-A Non-Parametric Approach to Dynamic Programming

10 0.0699993 224 nips-2011-Probabilistic Modeling of Dependencies Among Visual Short-Term Memory Representations

11 0.0630778 112 nips-2011-Heavy-tailed Distances for Gradient Based Image Descriptors

12 0.06287761 206 nips-2011-Optimal Reinforcement Learning for Gaussian Systems

13 0.05919078 22 nips-2011-Active Ranking using Pairwise Comparisons

14 0.057328604 35 nips-2011-An ideal observer model for identifying the reference frame of objects

15 0.057137474 258 nips-2011-Sparse Bayesian Multi-Task Learning

16 0.056877006 214 nips-2011-PiCoDes: Learning a Compact Code for Novel-Category Recognition

17 0.05491738 157 nips-2011-Learning to Search Efficiently in High Dimensions

18 0.048943605 219 nips-2011-Predicting response time and error rates in visual search

19 0.048916854 293 nips-2011-Understanding the Intrinsic Memorability of Images

20 0.046016172 231 nips-2011-Randomized Algorithms for Comparison-based Search


similar papers computed by lsi model

lsi for this paper:

topicId topicWeight

[(0, 0.127), (1, 0.061), (2, -0.011), (3, 0.078), (4, 0.014), (5, -0.032), (6, 0.013), (7, -0.032), (8, 0.014), (9, 0.069), (10, -0.037), (11, 0.007), (12, 0.038), (13, -0.005), (14, 0.202), (15, -0.029), (16, 0.063), (17, 0.026), (18, 0.152), (19, -0.069), (20, 0.005), (21, 0.007), (22, -0.017), (23, -0.025), (24, -0.063), (25, 0.097), (26, 0.105), (27, -0.113), (28, 0.015), (29, 0.132), (30, -0.021), (31, 0.008), (32, -0.011), (33, -0.023), (34, -0.005), (35, 0.088), (36, -0.031), (37, 0.015), (38, -0.007), (39, 0.051), (40, -0.033), (41, 0.115), (42, 0.06), (43, 0.039), (44, 0.052), (45, -0.068), (46, -0.139), (47, 0.011), (48, -0.001), (49, 0.011)]

similar papers list:

simIndex simValue paperId paperTitle

same-paper 1 0.92449218 280 nips-2011-Testing a Bayesian Measure of Representativeness Using a Large Image Database

Author: Joshua T. Abbott, Katherine A. Heller, Zoubin Ghahramani, Thomas L. Griffiths

Abstract: How do people determine which elements of a set are most representative of that set? We extend an existing Bayesian measure of representativeness, which indicates the representativeness of a sample from a distribution, to define a measure of the representativeness of an item to a set. We show that this measure is formally related to a machine learning method known as Bayesian Sets. Building on this connection, we derive an analytic expression for the representativeness of objects described by a sparse vector of binary features. We then apply this measure to a large database of images, using it to determine which images are the most representative members of different sets. Comparing the resulting predictions to human judgments of representativeness provides a test of this measure with naturalistic stimuli, and illustrates how databases that are more commonly used in computer vision and machine learning can be used to evaluate psychological theories. 1

2 0.82688552 90 nips-2011-Evaluating the inverse decision-making approach to preference learning

Author: Alan Jern, Christopher G. Lucas, Charles Kemp

Abstract: Psychologists have recently begun to develop computational accounts of how people infer others’ preferences from their behavior. The inverse decision-making approach proposes that people infer preferences by inverting a generative model of decision-making. Existing data sets, however, do not provide sufficient resolution to thoroughly evaluate this approach. We introduce a new preference learning task that provides a benchmark for evaluating computational accounts and use it to compare the inverse decision-making approach to a feature-based approach, which relies on a discriminative combination of decision features. Our data support the inverse decision-making approach to preference learning. A basic principle of decision-making is that knowing people’s preferences allows us to predict how they will behave: if you know your friend likes comedies and hates horror films, you can probably guess which of these options she will choose when she goes to the theater. Often, however, we do not know what other people like and we can only infer their preferences from their behavior. If you know that a different friend saw a comedy today, does that mean that he likes comedies in general? The conclusion you draw will likely depend on what else was playing and what movie choices he has made in the past. A goal for social cognition research is to develop a computational account of people’s ability to infer others’ preferences. One computational approach is based on inverse decision-making. This approach begins with a model of how someone’s preferences lead to a decision. Then, this model is inverted to determine the most likely preferences that motivated an observed decision. An alternative approach might simply learn a functional mapping between features of an observed decision and the preferences that motivated it. For instance, in your friend’s decision to see a comedy, perhaps the more movie options he turned down, the more likely it is that he has a true preference for comedies. The difference between the inverse decision-making approach and the feature-based approach maps onto the standard dichotomy between generative and discriminative models. Economists have developed an instance of the inverse decision-making approach known as the multinomial logit model [1] that has been widely used to infer consumer’s preferences from their choices. This model has recently been explored as a psychological model [2, 3, 4], but there are few behavioral data sets for evaluating it as a model of how people learn others’ preferences. Additionally, the data sets that do exist tend to be drawn from the developmental literature, which focuses on simple tasks that collect only one or two judgments from children [5, 6, 7]. The limitations of these data sets make it difficult to evaluate the multinomial logit model with respect to alternative accounts of preference learning like the feature-based approach. In this paper, we use data from a new experimental task that elicits a detailed set of preference judgments from a single participant in order to evaluate the predictions of several preference learning models from both the inverse decision-making and feature-based classes. Our task requires each participant to sort a large number of observed decisions on the basis of how strongly they indicate 1 (a) (b) (c) d c c (d) b b a a x d x d c b a x 1. Number of chosen effects (−/+) 2. Number of forgone effects (+/+) 3. Number of forgone options (+/+) 4. Number of forgone options containing x (−/−) 5. Max/min number of effects in a forgone option (+/−) 6. Is x in every option? (−/−) 7. Chose only option with x? (+/+) 8. Is x the only difference between options? (+/+) 9. Do all options have same number of effects? (+/+) 10. Chose option with max/min number of effects? (−/−) Figure 1: (a)–(c) Examples of the decisions used in the experiments. Each column represents one option and the boxes represent different effects. The chosen option is indicated by the black rectangle. (d) Features used by the weighted feature and ranked feature models. Features 5 and 10 involved maxima in Experiment 1, which focused on all positive effects, and minima in Experiment 2, which focused on all negative effects. The signs in parentheses indicate the direction of the feature that suggests a stronger preference in Experiment 1 / Experiment 2. a preference for a chosen item. Because the number of decisions is large and these decisions vary on multiple dimensions, predicting how people will order them offers a challenging benchmark on which to compare computational models of preference learning. Data sets from these sorts of detailed tasks have proved fruitful in other domains. For example, data reported by Shepard, Hovland, and Jenkins [8]; Osherson, Smith, Wilkie, L´ pez, and Shafir [9]; and Wasserman, Elek, Chatlosh, o and Baker [10] have motivated much subsequent research on category learning, inductive reasoning, and causal reasoning, respectively. We first describe our preference learning task in detail. We then present several inverse decisionmaking and feature-based models of preference learning and compare these models’ predictions to people’s judgments in two experiments. The data are well predicted by models that follow the inverse decision-making approach, suggesting that this computational approach may help explain how people learn others’ preferences. 1 Multi-attribute decisions and revealed preferences We designed a task that can be used to elicit a large number of preference judgments from a single participant. The task involves a set of observed multi-attribute decisions, some examples of which are represented visually in Figure 1. Each decision is among a set of options and each option produces a set of effects. Figure 1 shows several decisions involving a total of five effects distributed among up to five options. The differently colored boxes represent different effects and the chosen option is marked by a black rectangle. For example, 1a shows a choice between an option with four effects and an option with a single effect; here, the decision maker chose the second option. In our task, people are asked to rank a large number of these decisions by how strongly they suggest that the decision maker had a preference for a particular effect (e.g., effect x in Figure 1). By imposing some minimal constraints, the space of unique multi-attribute decisions is finite and we can obtain rankings for every decision in the space. For example, Figure 2c shows a complete list of 47 unique decisions involving up to five effects, subject to several constraints described later. Three of these decisions are shown in Figure 1. If all the effects are positive—pieces of candy, for example—the first decision (1a) suggests a strong preference for candy x, because the decision maker turned down four pieces in favor of one. The second decision (1b), however, offers much weaker evidence because nearly everyone would choose four pieces of candy over one, even without a specific preference for x. The third decision (1c) provides evidence that is strong but perhaps not quite as strong as the first decision. When all effects are negative—like electric shocks at different body locations—decision makers may still find some effects more tolerable than others, but different inferences are sometimes supported. For example, for negative effects, 1a provides weak evidence that x is relatively tolerable because nearly everyone would choose one shock over four. 2 A computational account of preference learning We now describe a simple computational model for learning a person’s preferences after observing that person make a decision like the ones in Figure 1. We assume that there are n available options 2 {o1 , . . . , on }, each of which produces one or more effects from the set {f1 , f2 , ..., fm }. For simplicity, we assume that effects are binary. Let ui denote the utility the decision maker assigns to effect fi . We begin by specifying a model of decision-making that makes the standard assumptions that decision makers tend to choose things with greater utility and that utilities are additive. That is, if fj is a binary vector indicating the effects produced by option oj and u is a vector of utilities assigned to each of the m effects, then the total utility associated with option oj can be expressed as Uj = fj T u. We complete the specification of the model by applying the Luce choice rule [11], a common psychological model of choice behavior, as the function that chooses among the options: p(c = oj |u, f ) = exp(Uj ) = exp(Uk ) n k=1 exp(fj T u) n T k=1 exp(fk u) (1) where c denotes the choice made. This model can predict the choice someone will make among a specified set of options, given the utilities that person assigns to the effects in each option. To obtain estimates of someone’s utilities, we invert this model by applying Bayes’ rule: p(u|c, F) = p(c|u, F)p(u) p(c|F) (2) where F = {f1 , . . . , fn } specifies the available options and their corresponding effects. This is the multinomial logit model [1], a standard econometric model. In order to apply Equation 2 we must specify a prior p(u) on the utilities. We adopt a standard approach that places independent Gaussian priors on the utilities: ui ∼ N (µ, σ 2 ). For decisions where effects are positive—like candies—we set µ = 2σ, which corresponds to a prior distribution that places approximately 2% of the probability mass below zero. Similarly, for negative effects—like electric shocks—we set µ = −2σ. 2.1 Ordering a set of observed decisions Equation 2 specifies a posterior probability distribution over utilities for a single observed decision but does not provide a way to compare the inferences drawn from multiple decisions for the purposes of ordering them. Suppose we are interested in a decision maker’s preference for effect x and we wish to order a set of decisions by how strongly they support this preference. Two criteria for ordering the decisions are as follows: Absolute utility Relative utility p(c|ux , F)p(ux ) p(c|F) p(c|∀j ux ≥ uj , F)p(∀j ux ≥ uj ) p(∀j ux ≥ uj |c, F) = p(c|F) E(ux |c, F) = Eux The absolute utility model orders decisions by the mean posterior utility for effect x. This criterion is perhaps the most natural way to assess how much a decision indicates a preference for x, but it requires an inference about the utility of x in isolation, and research suggests that people often think about the utility of an effect only in relation to other salient possibilities [12]. The relative utility model applies this idea to preference learning by ordering decisions based on how strongly they suggest that x has a greater utility than all other effects. The decisions in Figures 1b and 1c are cases where the two models lead to different predictions. If the effects are all negative (e.g., electric shocks), the absolute utility model predicts that 1b provides stronger evidence for a tolerance for x because the decision maker chose to receive four shocks instead of just one. The relative utility model predicts that 1c provides stronger evidence because 1b offers no way to determine the relative tolerance of the four chosen effects with respect to one another. Like all generative models, the absolute and relative models incorporate three qualitatively different components: the likelihood term p(c|u, F), the prior p(u), and the reciprocal of the marginal likelihood 1/p(c|F). We assume that the total number of effects is fixed in advance and, as a result, the prior term will be the same for all decisions that we consider. The two other components, however, will vary across decisions. The inverse decision-making approach predicts that both components should influence preference judgments, and we will test this prediction by comparing our 3 two inverse decision-making models to two alternatives that rely only one of these components as an ordering criterion: p(c|∀j ux ≥ uj , F) 1/p(c|F) Representativeness Surprise The representativeness model captures how likely the observed decision would be if the utility for x were high, and previous research has shown that people sometimes rely on a representativeness computation of this kind [13]. The surprise model captures how unexpected the observed decision is overall; surprising decisions may be best explained in terms of a strong preference for x, but unsurprising decisions provide little information about x in particular. 2.2 Feature-based models We also consider a class of feature-based models that use surface features to order decisions. The ten features that we consider are shown in Figure 1d, where x is the effect of interest. As an example, the first feature specifies the number of effects chosen; because x is always among the chosen effects, decisions where few or no other effects belong to the chosen option suggest the strongest preference for x (when all effects are positive). This and the second feature were previously identified by Newtson [14]; we included the eight additional features shown in Figure 1d in an attempt to include all possible features that seemed both simple and relevant. We consider two methods for combining this set of features to order a set of decisions by how strongly they suggest a preference for x. The first model is a standard linear regression model, which we refer to as the weighted feature model. The model learns a weight for each feature, and the rank of a given decision is determined by a weighted sum of its features. The second model is a ranked feature model that sorts the observed decisions with respect to a strict ranking of the features. The top-ranked feature corresponds to the primary sort key, the second-ranked feature to the secondary sort key, and so on. For example, suppose that the top-ranked feature is the number of chosen effects and the second-ranked feature is the number of forgone options. Sorting the three decisions in Figure 1 according to this criterion produces the following ordering: 1a,1c,1b. This notion of sorting items on the basis of ranked features has been applied before to decision-making [15, 16] and other domains of psychology [17], but we are not aware of any previous applications to preference learning. Although our inverse decision-making and feature-based models represent two very different approaches, both may turn out to be valuable. An inverse decision-making approach may be the appropriate account of preference learning at Marr’s [18] computational level, and a feature-based approach may capture the psychological processes by which the computational-level account is implemented. Our goal, therefore, is not necessarily to accept one of these approaches and dismiss the other. Instead, we entertain three distinct possibilities. First, both approaches may account well for the data, which would support the idea that they are valid accounts operating at different levels of analysis. Second, the inverse decision-making approach may offer a better account, suggesting that process-level accounts other than the feature-based approach should be explored. Finally, the feature-based approach may offer a better account, suggesting that inverse decision-making does not constitute an appropriate computational-level account of preference learning. 3 Experiment 1: Positive effects Our first experiment focuses on decisions involving only positive effects. The full set of 47 decisions we used is shown in Figure 2c. This set includes every possible unique decision with up to five different effects, subject to the following constraints: (1) one of the effects (effect x) must always appear in the chosen option, (2) there are no repeated options, (3) each effect may appear in an option at most once, (4) only effects in the chosen option may be repeated in other options, and (5) when effects appear in multiple options, the number of effects is held constant across options. The first constraint is necessary for the sorting task, the second two constraints create a finite space of decisions, and the final two constraints limit attention to what we deemed the most interesting cases. Method 43 Carnegie Mellon undergraduates participated for course credit. Each participant was given a set of cards, with one decision printed on each card. The decisions were represented visually 4 (a) (c) Decisions 42 40 45 Mean human rankings 38 30 23 20 22 17 13 12 11 10 9 8 7 6 19 18 31 34 28 21 26 36 35 33 37 27 29 32 25 24 16 15 14 5 4 3 2 1 Absolute utility model rankings (b) Mean human rankings (Experiment 1) 47 43 44 46 45 38 37 36 34 35 30 32 33 31 29 28 24 26 27 25 21 19 22 20 18 16 17 12 13 7 6 11 5 9 4 10 8 1 2 3 42 40 41 39 47 46 44 41 43 39 23 15 14 Mean human rankings (Experiment 2) 1. dcbax 2. cbax 3. bax 4. ax 5. x 6. dcax | bcax 7. dx | cx | bx | ax 8. cax | bax 9. bdx | bcx | bax 10. dcx | bax 11. bx | ax 12. bdx | cax | bax 13. cx | bx | ax 14. d | cbax 15. c | bax 16. b | ax 17. d | c | bax 18. dc | bax 19. c | b | ax 20. dc | bx | ax 21. bdc | bax 22. ad | cx | bx | ax 23. d | c | b | ax 24. bad | bcx | bax 25. ac | bx | ax 26. cb | ax 27. cbad | cbax 28. dc | b | ax 29. ad | ac | bx | ax 30. ab | ax 31. bad | bax 32. dc | ab | ax 33. dcb | ax 34. a | x 35. bad | bac | bax 36. ac | ab | ax 37. ad | ac | ab | ax 38. b | a | x 39. ba | x 40. c | b | a | x 41. cb | a | x 42. d | c | b | a | x 43. cba | x 44. dc | ba | x 45. dc | b | a | x 46. dcb | a | x 47. dcba | x Figure 2: (a) Comparison between the absolute utility model rankings and the mean human rankings for Experiment 1. Each point represents one decision, numbered with respect to the list in panel c. (b) Comparison between the mean human rankings in Experiments 1 and 2. In both scatter plots, the solid diagonal lines indicate a perfect correspondence between the two sets of rankings. (c) The complete set of decisions, ordered by the mean human rankings from Experiment 1. Options are separated by vertical bars and the chosen option is always at the far right. Participants were always asked about a preference for effect x. as in Figure 1 but without the letter labels. Participants were told that the effects were different types of candy and each option was a bag containing one or more pieces of candy. They were asked to sort the cards by how strongly each decision suggested that the decision maker liked a particular target candy, labeled x in Figure 2c. They sorted the cards freely on a table but reported their final rankings by writing them on a sheet of paper, from weakest to strongest evidence. They were instructed to order the cards as completely as possible, but were told that they could assign the same ranking to a set of cards if they believed those cards provided equal evidence. 3.1 Results Two participants were excluded as outliers based on the criterion that their rankings for at least five decisions were at least three standard deviations from the mean rankings. We performed a hierarchical clustering analysis of the remaining 41 participants’ rankings using rank correlation as a similarity metric. Participants’ rankings were highly correlated: cutting the resulting dendrogram at 0.2 resulted in one cluster that included 33 participants and the second largest cluster included 5 Surprise MAE = 17.8 MAE = 7.0 MAE = 4.3 MAE = 17.3 MAE = 9.5 Human rankings Experiment 2 Negative effects Representativeness MAE = 2.3 MAE = 6.7 Experiment 1 Positive effects Relative utility MAE = 2.3 Human rankings Absolute utility Model rankings Model rankings Model rankings Model rankings Figure 3: Comparison between human rankings in both experiments and predicted rankings from four models. The solid diagonal lines indicate a perfect correspondence between human and model rankings. only 3 participants. Thus, we grouped all participants together and analyzed their mean rankings. The 0.2 threshold was chosen because it produced the most informative clustering in Experiment 2. Inverse decision-making models We implemented the inverse decision-making models using importance sampling with 5 million samples drawn from the prior distribution p(u). Because all the effects were positive, we used a prior on utilities that placed nearly all probability mass above zero (µ = 4, σ = 2). The mean human rankings are compared with the absolute utility model rankings in Figure 2a, and the mean human rankings are listed in order in 2c. Fractional rankings were used for both the human data and the model predictions. The human rankings in the figure are the means of participants’ fractional rankings. The first row of Figure 3 contains similar plots that allow comparison of the four models we considered. In these plots, the solid diagonal lines indicate a perfect correspondence between model and human rankings. Thus, the largest deviations from this line represent the largest deviations in the data from the model’s predictions. Figure 3 shows that the absolute and relative utility models make virtually identical predictions and both models provide a strong account of the human rankings as measured by mean absolute error (MAE = 2.3 in both cases). Moreover, both models correctly predict the highest ranked decision and the set of lowest ranked decisions. The only clear discrepancy between the model predictions and the data is the cluster of points at the lower left, labeled as Decisions 6–13 in Figure 2a. These are all cases in which effect x appears in all options and therefore these decisions provide no information about a decision maker’s preference for x. Consequently, the models assign the same ranking to this group as to the group of decisions in which there is only a single option (Decisions 1–5). Although people appeared to treat these groups somewhat differently, the models still correctly predict that the entire group of decisions 1–13 is ranked lower than all other decisions. The surprise and representativeness models do not perform nearly as well (MAE = 7.0 and 17.8, respectively). Although the surprise model captures some of the general trends in the human rankings, it makes several major errors. For example, consider Decision 7: dx|cx|bx|ax. This decision provides no information about a preference for x because it appears in every option. The decision is surprising, however, because a decision maker choosing at random from these options would make the observed choice only 1/4 of the time. The representativeness model performs even worse, primarily because it does not take into account alternative explanations for why an option was chosen, such as the fact that no other options were available (e.g., Decision 1 in Figure 2c). The failure of these models to adequately account for the data suggests that both the likelihood p(c|u, F) and marginal likelihood p(c|F) are important components of the absolute and relative utility models. Feature-based models We compared the performance of the absolute and relative utility models to our two feature-based models: the weighted feature and ranked feature models. For each participant, 6 (b) Ranked feature 10 10 5 Figure 4: Results of the feature-based model analysis from Experiment 1 for (a) the weighted feature models and (b) the ranked feature models. The histograms show the minimum number of features needed to match the accuracy (measured by MAE) of the absolute utility model for each participant. 15 5 1 2 3 4 5 6 >6 15 1 2 3 4 5 6 7 8 9 10 >10 Number of participants (a) Weighted feature Number of features needed we considered every subset of features1 in Figure 1d in order to determine the minimum number of features needed by the two models to achieve the same level of accuracy as the absolute utility model, as measured by mean absolute error. The results of these analyses are shown in Figure 4. For the majority of participants, at least four features were needed by both models to match the accuracy of the absolute utility model. For the weighted feature model, 14 participants could not be fit as well as the absolute utility model even when all ten features were considered. These results indicate that a feature-based account of people’s inferences in our task must be supplied with a relatively large number of features. By contrast, the inverse decision-making approach provides a relatively parsimonious account of the data. 4 Experiment 2: Negative effects Experiment 2 focused on a setting in which all effects are negative, motivated by the fact that the inverse decision-making models predict several major differences in orderings when effects are negative rather than positive. For instance, the absolute utility model’s relative rankings of the decisions in Figures 1a and 1b are reversed when all effects are negative rather than positive. Method 42 Carnegie Mellon undergraduates participated for course credit. The experimental design was identical to Experiment 1 except that participants were told that the effects were electric shocks at different body locations. They were asked to sort the cards on the basis of how strongly each decision suggested that the decision maker finds shocks at the target location relatively tolerable. The model predictions were derived in the same way as for Experiment 1, but with a prior distribution on utilities that placed nearly all probability mass below zero (µ = −4, σ = 2) to reflect the fact that effects were all negative. 4.1 Results Three participants were excluded as outliers by the same criterion applied in Experiment 1. The resulting mean rankings are compared with the corresponding rankings from Experiment 1 in Figure 2b. The figure shows that responses based on positive and negative effects were substantially different in a number of cases. Figure 3 shows how the mean rankings compare to the predictions of the four models we considered. Although the relative utility model is fairly accurate, no model achieves the same level of accuracy as the absolute and relative utility models in Experiment 1. In addition, the relative utility model provides a poor account of the responses of many individual participants. To better understand responses at the individual level, we repeated the hierarchical clustering analysis described in Experiment 1, which revealed that 29 participants could be grouped into one of four clusters, with the remaining participants each in their own clusters. We analyzed these four clusters independently, excluding the 10 participants that could not be naturally grouped. We compared the mean rankings of each cluster to the absolute and relative utility models, as well as all one- and two-feature weighted feature and ranked feature models. Figure 5 shows that the mean rankings of participants in Cluster 1 (N = 8) were best fit by the absolute utility model, the mean rankings of participants in Cluster 2 (N = 12) were best fit by the relative utility model, and the mean rankings of participants in Clusters 3 (N = 3) and 4 (N = 6) were better fit by feature-based models than by either the absolute or relative utility models. 1 A maximum of six features was considered for the ranked feature model because considering more features was computationally intractable. 7 Cluster 4 N =6 MAE = 4.9 MAE = 14.0 MAE = 7.9 MAE = 5.3 MAE = 2.6 MAE = 13.0 MAE = 6.2 Human rankings Relative utility Cluster 3 N =3 MAE = 2.6 Absolute utility Cluster 2 N = 12 Human rankings Cluster 1 N =8 Factors: 1,3 Factors: 1,8 MAE = 2.3 MAE = 5.2 Model rankings Best−fitting weighted feature Factors: 6,7 MAE = 4.0 Model rankings Model rankings Model rankings Human rankings Factors: 3,8 MAE = 4.8 Figure 5: Comparison between human rankings for four clusters of participants identified in Experiment 2 and predicted rankings from three models. Each point in the plots corresponds to one decision and the solid diagonal lines indicate a perfect correspondence between human and model rankings. The third row shows the predictions of the best-fitting two-factor weighted feature model for each cluster. The two factors listed refer to Figure 1d. To examine how well the models accounted for individuals’ rankings within each cluster, we compared the predictions of the inverse decision-making models to the best-fitting two-factor featurebased model for each participant. In Cluster 1, 7 out of 8 participants were best fit by the absolute utility model; in Cluster 2, 8 out of 12 participants were best fit by the relative utility model; in Clusters 3 and 4, all participants were better fit by feature-based models. No single feature-based model provided the best fit for more than two participants, suggesting that participants not fit well by the inverse decision-making models were not using a single alternative strategy. Applying the feature-based model analysis from Experiment 1 to the current results revealed that the weighted feature model required an average of 6.0 features to match the performance of the absolute utility model for participants in Cluster 1, and an average of 3.9 features to match the performance of the relative utility model for participants in Cluster 2. Thus, although a single model did not fit all participants well in the current experiment, many participants were fit well by one of the two inverse decision-making models, suggesting that this general approach is useful for explaining how people reason about negative effects as well as positive effects. 5 Conclusion In two experiments, we found that an inverse decision-making approach offered a good computational account of how people make judgments about others’ preferences. Although this approach is conceptually simple, our analyses indicated that it captures the influence of a fairly large number of relevant decision features. Indeed, the feature-based models that we considered as potential process models of preference learning could only match the performance of the inverse decision-making approach when supplied with a relatively large number of features. We feel that this result rules out the feature-based approach as psychologically implausible, meaning that alternative process-level accounts will need to be explored. One possibility is sampling, which has been proposed as a psychological mechanism for approximating probabilistic inferences [19, 20]. However, even if process models that use large numbers of features are considered plausible, the inverse decision-making approach provides a valuable computational-level account that helps to explain which decision features are informative. Acknowledgments This work was supported in part by the Pittsburgh Life Sciences Greenhouse Opportunity Fund and by NSF grant CDI-0835797. 8 References [1] D. McFadden. Conditional logit analysis of qualitative choice behavior. In P. Zarembka, editor, Frontiers in Econometrics. Amademic Press, New York, 1973. [2] C. G. Lucas, T. L. Griffiths, F. Xu, and C. Fawcett. A rational model of preference learning and choice prediction by children. In Proceedings of Neural Information Processing Systems 21, 2009. [3] L. Bergen, O. R. Evans, and J. B. Tenenbaum. Learning structured preferences. In Proceedings of the 32nd Annual Conference of the Cognitive Science Society, 2010. [4] A. Jern and C. Kemp. Decision factors that support preference learning. In Proceedings of the 33rd Annual Conference of the Cognitive Science Society, 2011. [5] T. Kushnir, F. Xu, and H. M. Wellman. Young children use statistical sampling to infer the preferences of other people. Psychological Science, 21(8):1134–1140, 2010. [6] L. Ma and F. Xu. Young children’s use of statistical sampling evidence to infer the subjectivity of preferences. Cognition, in press. [7] M. J. Doherty. Theory of Mind: How Children Understand Others’ Thoughts and Feelings. Psychology Press, New York, 2009. [8] R. N. Shepard, C. I. Hovland, and H. M. Jenkins. Learning and memorization of classifications. Psychological Monographs, 75, Whole No. 517, 1961. [9] D. N. Osherson, E. E. Smith, O. Wilkie, A. L´ pez, and E. Shafir. Category-based induction. Psychological o Review, 97(2):185–200, 1990. [10] E. A. Wasserman, S. M. Elek, D. L. Chatlosh, and A. G. Baker. Rating causal relations: Role of probability in judgments of response-outcome contingency. Journal of Experimental Psychology: Learning, Memory, and Cognition, 19(1):174–188, 1993. [11] R. D. Luce. Individual choice behavior. John Wiley, 1959. [12] D. Ariely, G. Loewenstein, and D. Prelec. Tom Sawyer and the construction of value. Journal of Economic Behavior & Organization, 60:1–10, 2006. [13] D. Kahneman and A. Tversky. Subjective probability: A judgment of representativeness. Cognitive Psychology, 3(3):430–454, 1972. [14] D. Newtson. Dispositional inference from effects of actions: Effects chosen and effects forgone. Journal of Experimental Social Psychology, 10:489–496, 1974. [15] P. C. Fishburn. Lexicographic orders, utilities and decision rules: A survey. Management Science, 20(11):1442–1471, 1974. [16] G. Gigerenzer and P. M. Todd. Fast and frugal heuristics: The adaptive toolbox. Oxford University Press, New York, 1999. [17] A. Prince and P. Smolensky. Optimality Theory: Constraint Interaction in Generative Grammar. WileyBlackwell, 2004. [18] D. Marr. Vision. W. H. Freeman, San Francisco, 1982. [19] A. N. Sanborn, T. L. Griffiths, and D. J. Navarro. Rational approximations to rational models: Alternative algorithms for category learning. Psychological Review, 117:1144–1167, 2010. [20] L. Shi and T. L. Griffiths. Neural implementation of Bayesian inference by importance sampling. In Proceedings of Neural Information Processing Systems 22, 2009. 9

3 0.72809321 34 nips-2011-An Unsupervised Decontamination Procedure For Improving The Reliability Of Human Judgments

Author: Michael C. Mozer, Benjamin Link, Harold Pashler

Abstract: Psychologists have long been struck by individuals’ limitations in expressing their internal sensations, impressions, and evaluations via rating scales. Instead of using an absolute scale, individuals rely on reference points from recent experience. This relativity of judgment limits the informativeness of responses on surveys, questionnaires, and evaluation forms. Fortunately, the cognitive processes that map stimuli to responses are not simply noisy, but rather are influenced by recent experience in a lawful manner. We explore techniques to remove sequential dependencies, and thereby decontaminate a series of ratings to obtain more meaningful human judgments. In our formulation, the problem is to infer latent (subjective) impressions from a sequence of stimulus labels (e.g., movie names) and responses. We describe an unsupervised approach that simultaneously recovers the impressions and parameters of a contamination model that predicts how recent judgments affect the current response. We test our iterated impression inference, or I3 , algorithm in three domains: rating the gap between dots, the desirability of a movie based on an advertisement, and the morality of an action. We demonstrate significant objective improvements in the quality of the recovered impressions. 1

4 0.57391083 15 nips-2011-A rational model of causal inference with continuous causes

Author: Thomas L. Griffiths, Michael James

Abstract: Rational models of causal induction have been successful in accounting for people’s judgments about causal relationships. However, these models have focused on explaining inferences from discrete data of the kind that can be summarized in a 2× 2 contingency table. This severely limits the scope of these models, since the world often provides non-binary data. We develop a new rational model of causal induction using continuous dimensions, which aims to diminish the gap between empirical and theoretical approaches and real-world causal induction. This model successfully predicts human judgments from previous studies better than models of discrete causal inference, and outperforms several other plausible models of causal induction with continuous causes in accounting for people’s inferences in a new experiment. 1

5 0.56542307 130 nips-2011-Inductive reasoning about chimeric creatures

Author: Charles Kemp

Abstract: Given one feature of a novel animal, humans readily make inferences about other features of the animal. For example, winged creatures often fly, and creatures that eat fish often live in the water. We explore the knowledge that supports these inferences and compare two approaches. The first approach proposes that humans rely on abstract representations of dependency relationships between features, and is formalized here as a graphical model. The second approach proposes that humans rely on specific knowledge of previously encountered animals, and is formalized here as a family of exemplar models. We evaluate these models using a task where participants reason about chimeras, or animals with pairs of features that have not previously been observed to co-occur. The results support the hypothesis that humans rely on explicit representations of relationships between features. Suppose that an eighteenth-century naturalist learns about a new kind of animal that has fur and a duck’s bill. Even though the naturalist has never encountered an animal with this pair of features, he should be able to make predictions about other features of the animal—for example, the animal could well live in water but probably does not have feathers. Although the platypus exists in reality, from a eighteenth-century perspective it qualifies as a chimera, or an animal that combines two or more features that have not previously been observed to co-occur. Here we describe a probabilistic account of inductive reasoning and use it to account for human inferences about chimeras. The inductive problems we consider are special cases of the more general problem in Figure 1a where a reasoner is given a partially observed matrix of animals by features then asked to infer the values of the missing entries. This general problem has been previously studied and is addressed by computational models of property induction, categorization, and generalization [1–7]. A challenge faced by all of these models is to capture the background knowledge that guides inductive inferences. Some accounts rely on similarity relationships between animals [6, 8], others rely on causal relationships between features [9, 10], and others incorporate relationships between animals and relationships between features [11]. We will evaluate graphical models that capture both kinds of relationships (Figure 1a), but will focus in particular on relationships between features. Psychologists have previously suggested that humans rely on explicit mental representations of relationships between features [12–16]. Often these representations are described as theories—for example, theories that specify a causal relationship between having wings and flying, or living in the sea and eating fish. Relationships between features may take several forms: for example, one feature may cause, enable, prevent, be inconsistent with, or be a special case of another feature. For simplicity, we will treat all of these relationships as instances of dependency relationships between features, and will capture them using an undirected graphical model. Previous studies have used graphical models to account for human inferences about features but typically these studies consider toy problems involving a handful of novel features such as “has gene X14” or “has enzyme Y132” [9, 11]. Participants might be told, for example, that gene X14 leads to the production of enzyme Y132, then asked to use this information when reasoning about novel animals. Here we explore whether a graphical model approach can account for inferences 1 (a) slow heavy flies (b) wings hippo 1 1 0 0 rhino 1 1 0 0 sparrow 0 0 1 1 robin 0 0 1 1 new ? ? 1 ? o Figure 1: Inductive reasoning about animals and features. (a) Inferences about the features of a new animal onew that flies may draw on similarity relationships between animals (the new animal is similar to sparrows and robins but not hippos and rhinos), and on dependency relationships between features (flying and having wings are linked). (b) A graph product produced by combining the two graph structures in (a). about familiar features. Working with familiar features raises a methodological challenge since participants have a substantial amount of knowledge about these features and can reason about them in multiple ways. Suppose, for example, that you learn that a novel animal can fly (Figure 1a). To conclude that the animal probably has wings, you might consult a mental representation similar to the graph at the top of Figure 1a that specifies a dependency relationship between flying and having wings. On the other hand, you might reach the same conclusion by thinking about flying creatures that you have previously encountered (e.g. sparrows and robins) and noticing that these creatures have wings. Since the same conclusion can be reached in two different ways, judgments about arguments of this kind provide little evidence about the mental representations involved. The challenge of working with familiar features directly motivates our focus on chimeras. Inferences about chimeras draw on rich background knowledge but require the reasoner to go beyond past experience in a fundamental way. For example, if you learn that an animal flies and has no legs, you cannot make predictions about the animal by thinking of flying, no-legged creatures that you have previously encountered. You may, however, still be able to infer that the novel animal has wings if you understand the relationship between flying and having wings. We propose that graphical models over features can help to explain how humans make inferences of this kind, and evaluate our approach by comparing it to a family of exemplar models. The next section introduces these models, and we then describe two experiments designed to distinguish between the models. 1 Reasoning about objects and features Our models make use of a binary matrix D where the rows {o1 , . . . , o129 } correspond to objects, and the columns {f 1 , . . . , f 56 } correspond to features. A subset of the objects is shown in Figure 2a, and the full set of features is shown in Figure 2b and its caption. Matrix D was extracted from the Leuven natural concept database [17], which includes 129 animals and 757 features in total. We chose a subset of these features that includes a mix of perceptual and behavioral features, and that includes many pairs of features that depend on each other. For example, animals that “live in water” typically “can swim,” and animals that have “no legs” cannot “jump far.” Matrix D can be used to formulate problems where a reasoner observes one or two features of a new object (i.e. animal o130 ) and must make inferences about the remaining features of the animal. The next two sections describe graphical models that can be used to address this problem. The first graphical model O captures relationships between objects, and the second model F captures relationships between features. We then discuss how these models can be combined, and introduce a family of exemplar-style models that will be compared with our graphical models. A graphical model over objects Many accounts of inductive reasoning focus on similarity relationships between objects [6, 8]. Here we describe a tree-structured graphical model O that captures these relationships. The tree was constructed from matrix D using average linkage clustering and the Jaccard similarity measure, and part of the resulting structure is shown in Figure 2a. The subtree in Figure 2a includes clusters 2 alligator caiman crocodile monitor lizard dinosaur blindworm boa cobra python snake viper chameleon iguana gecko lizard salamander frog toad tortoise turtle anchovy herring sardine cod sole salmon trout carp pike stickleback eel flatfish ray plaice piranha sperm whale squid swordfish goldfish dolphin orca whale shark bat fox wolf beaver hedgehog hamster squirrel mouse rabbit bison elephant hippopotamus rhinoceros lion tiger polar bear deer dromedary llama giraffe zebra kangaroo monkey cat dog cow horse donkey pig sheep (a) (b) can swim lives in water eats fish eats nuts eats grain eats grass has gills can jump far has two legs has no legs has six legs has four legs can fly can be ridden has sharp teeth nocturnal has wings strong predator can see in dark eats berries lives in the sea lives in the desert crawls lives in the woods has mane lives in trees can climb well lives underground has feathers has scales slow has fur heavy Figure 2: Graph structures used to define graphical models O and F. (a) A tree that captures similarity relationships between animals. The full tree includes 129 animals, and only part of the tree is shown here. The grey points along the branches indicate locations where a novel animal o130 could be attached to the tree. (b) A network capturing pairwise dependency relationships between features. The edges capture both positive and negative dependencies. All edges in the network are shown, and the network also includes 20 isolated nodes for the following features: is black, is blue, is green, is grey, is pink, is red, is white, is yellow, is a pet, has a beak, stings, stinks, has a long neck, has feelers, sucks blood, lays eggs, makes a web, has a hump, has a trunk, and is cold-blooded. corresponding to amphibians and reptiles, aquatic creatures, and land mammals, and the subtree omitted for space includes clusters for insects and birds. We assume that the features in matrix D (i.e. the columns) are generated independently over O: P (f i |O, π i , λi ). P (D|O, π, λ) = i i i i The distribution P (f |O, π , λ ) is based on the intuition that nearby nodes in O tend to have the same value of f i . Previous researchers [8, 18] have used a directed graphical model where the distribution at the root node is based on the baserate π i , and any other node v with parent u has the following conditional probability distribution: i P (v = 1|u) = π i + (1 − π i )e−λ l , if u = 1 i π i − π i e−λ l , if u = 0 (1) where l is the length of the branch joining node u to node v. The variability parameter λi captures the extent to which feature f i is expected to vary over the tree. Note, for example, that any node v must take the same value as its parent u when λ = 0. To avoid free parameters, the feature baserates π i and variability parameters λi are set to their maximum likelihood values given the observed values of the features {f i } in the data matrix D. The conditional distributions in Equation 1 induce a joint distribution over all of the nodes in graph O, and the distribution P (f i |O, π i , λi ) is computed by marginalizing out the values of the internal nodes. Although we described O as a directed graphical model, the model can be converted into an equivalent undirected model with a potential for each edge in the tree and a potential for the root node. Here we use the undirected version of the model, which is a natural counterpart to the undirected model F described in the next section. The full version of structure O in Figure 2a includes 129 familiar animals, and our task requires inferences about a novel animal o130 that must be slotted into the structure. Let D′ be an expanded version of D that includes a row for o130 , and let O′ be an expanded version of O that includes a node for o130 . The edges in Figure 2a are marked with evenly spaced gray points, and we use a 3 uniform prior P (O′ ) over all trees that can be created by attaching o130 to one of these points. Some of these trees have identical topologies, since some edges in Figure 2a have multiple gray points. Predictions about o130 can be computed using: P (D′ |D) = P (D′ |O′ , D)P (O′ |D) ∝ O′ P (D′ |O′ , D)P (D|O′ )P (O′ ). (2) O′ Equation 2 captures the basic intuition that the distribution of features for o130 is expected to be consistent with the distribution observed for previous animals. For example, if o130 is known to fly then the trees with high posterior probability P (O′ |D) will be those where o130 is near other flying creatures (Figure 1a), and since these creatures have wings Equation 2 predicts that o130 probably also has wings. As this example suggests, model O captures dependency relationships between features implicitly, and therefore stands in contrast to models like F that rely on explicit representations of relationships between features. A graphical model over features Model F is an undirected graphical model defined over features. The graph shown in Figure 2b was created by identifying pairs where one feature depends directly on another. The author and a research assistant both independently identified candidate sets of pairwise dependencies, and Figure 2b was created by merging these sets and reaching agreement about how to handle any discrepancies. As previous researchers have suggested [13, 15], feature dependencies can capture several kinds of relationships. For example, wings enable flying, living in the sea leads to eating fish, and having no legs rules out jumping far. We work with an undirected graph because some pairs of features depend on each other but there is no clear direction of causal influence. For example, there is clearly a dependency relationship between being nocturnal and seeing in the dark, but no obvious sense in which one of these features causes the other. We assume that the rows of the object-feature matrix D are generated independently from an undirected graphical model F defined over the feature structure in Figure 2b: P (oi |F). P (D|F) = i Model F includes potential functions for each node and for each edge in the graph. These potentials were learned from matrix D using the UGM toolbox for undirected graphical models [19]. The learned potentials capture both positive and negative relationships: for example, animals that live in the sea tend to eat fish, and tend not to eat berries. Some pairs of feature values never occur together in matrix D (there are no creatures that fly but do not have wings). We therefore chose to compute maximum a posteriori values of the potential functions rather than maximum likelihood values, and used a diffuse Gaussian prior with a variance of 100 on the entries in each potential. After learning the potentials for model F, we can make predictions about a new object o130 using the distribution P (o130 |F). For example, if o130 is known to fly (Figure 1a), model F predicts that o130 probably has wings because the learned potentials capture a positive dependency between flying and having wings. Combining object and feature relationships There are two simple ways to combine models O and F in order to develop an approach that incorporates both relationships between features and relationships between objects. The output combination model computes the predictions of both models in isolation, then combines these predictions using a weighted sum. The resulting model is similar to a mixture-of-experts model, and to avoid free parameters we use a mixing weight of 0.5. The structure combination model combines the graph structures used by the two models and relies on a set of potentials defined over the resulting graph product. An example of a graph product is shown in Figure 1b, and the potential functions for this graph are inherited from the component models in the natural way. Kemp et al. [11] use a similar approach to combine a functional causal model with an object model O, but note that our structure combination model uses an undirected model F rather than a functional causal model over features. Both combination models capture the intuition that inductive inferences rely on relationships between features and relationships between objects. The output combination model has the virtue of 4 simplicity, and the structure combination model is appealing because it relies on a single integrated representation that captures both relationships between features and relationships between objects. To preview our results, our data suggest that the combination models perform better overall than either O or F in isolation, and that both combination models perform about equally well. Exemplar models We will compare the family of graphical models already described with a family of exemplar models. The key difference between these model families is that the exemplar models do not rely on explicit representations of relationships between objects and relationships between features. Comparing the model families can therefore help to establish whether human inferences rely on representations of this sort. Consider first a problem where a reasoner must predict whether object o130 has feature k after observing that it has feature i. An exemplar model addresses the problem by retrieving all previouslyobserved objects with feature i and computing the proportion that have feature k: P (ok = 1|oi = 1) = |f k & f i | |f i | (3) where |f k | is the number of objects in matrix D that have feature k, and |f k & f i | is the number that have both feature k and feature i. Note that we have streamlined our notation by using ok instead of o130 to refer to the kth feature value for object o130 . k Suppose now that the reasoner observes that object o130 has features i and j. The natural generalization of Equation 3 is: P (ok = 1|oi = 1, oj = 1) = |f k & f i & f j | |f i & f j | (4) Because we focus on chimeras, |f i & f j | = 0 and Equation 4 is not well defined. We therefore evaluate an exemplar model that computes predictions for the two observed features separately then computes the weighted sum of these predictions: P (ok = 1|oi = 1, oj = 1) = wi |f k & f i | |f k & f j | + wj . i| |f |f j | (5) where the weights wi and wj must sum to one. We consider four ways in which the weights could be set. The first strategy sets wi = wj = 0.5. The second strategy sets wi ∝ |f i |, and is consistent with an approach where the reasoner retrieves all exemplars in D that are most similar to the novel animal and reports the proportion of these exemplars that have feature k. The third strategy sets wi ∝ |f1i | , and captures the idea that features should be weighted by their distinctiveness [20]. The final strategy sets weights according to the coherence of each feature [21]. A feature is coherent if objects with that feature tend to resemble each other overall, and we define the coherence of feature i as the expected Jaccard similarity between two randomly chosen objects from matrix D that both have feature i. Note that the final three strategies are all consistent with previous proposals from the psychological literature, and each one might be expected to perform well. Because exemplar models and prototype models are often compared, it is natural to consider a prototype model [22] as an additional baseline. A standard prototype model would partition the 129 animals into categories and would use summary statistics for these categories to make predictions about the novel animal o130 . We will not evaluate this model because it corresponds to a coarser version of model O, which organizes the animals into a hierarchy of categories. The key characteristic shared by both models is that they explicitly capture relationships between objects but not features. 2 Experiment 1: Chimeras Our first experiment explores how people make inferences about chimeras, or novel animals with features that have not previously been observed to co-occur. Inferences about chimeras raise challenges for exemplar models, and therefore help to establish whether humans rely on explicit representations of relationships between features. Each argument can be represented as f i , f j → f k 5 exemplar r = 0.42 7 feature F exemplar (wi = |f i |) (wi = 0.5) r = 0.44 7 object O r = 0.69 7 output combination r = 0.31 7 structure combination r = 0.59 7 r = 0.60 7 5 5 5 5 5 3 3 3 3 3 3 all 5 1 1 0 1 r = 0.06 7 conflict 0.5 1 1 0 0.5 1 r = 0.71 7 1 0 0.5 1 r = −0.02 7 1 0 0.5 1 r = 0.49 7 0 5 5 5 5 3 3 3 3 1 5 3 0.5 r = 0.57 7 5 3 1 0 0.5 1 r = 0.51 7 edge 0.5 r = 0.17 7 1 1 0 0.5 1 r = 0.64 7 1 0 0.5 1 r = 0.83 7 1 0 0.5 1 r = 0.45 7 1 0 0.5 1 r = 0.76 7 0 5 5 5 5 3 3 3 3 1 5 3 0.5 r = 0.79 7 5 3 1 1 0 0.5 1 r = 0.26 7 other 1 0 1 0 0.5 1 r = 0.25 7 1 0 0.5 1 r = 0.19 7 1 0 0.5 1 r = 0.25 7 1 0 0.5 1 r = 0.24 7 0 7 5 5 5 5 5 3 3 3 3 1 5 3 0.5 r = 0.33 3 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 0 0.5 1 Figure 3: Argument ratings for Experiment 1 plotted against the predictions of six models. The y-axis in each panel shows human ratings on a seven point scale, and the x-axis shows probabilities according to one of the models. Correlation coefficients are shown for each plot. where f i and f k are the premises (e.g. “has no legs” and “can fly”) and f k is the conclusion (e.g. “has wings”). We are especially interested in conflict cases where the premises f i and f j lead to opposite conclusions when taken individually: for example, most animals with no legs do not have wings, but most animals that fly do have wings. Our models that incorporate feature structure F can resolve this conflict since F includes a dependency between “wings” and “can fly” but not between “wings” and “has no legs.” Our models that do not include F cannot resolve the conflict and predict that humans will be uncertain about whether the novel animal has wings. Materials. The object-feature matrix D includes 447 feature pairs {f i , f j } such that none of the 129 animals has both f i and f j . We selected 40 pairs (see the supporting material) and created 400 arguments in total by choosing 10 conclusion features for each pair. The arguments can be assigned to three categories. Conflict cases are arguments f i , f j → f k such that the single-premise arguments f i → f k and f j → f k lead to incompatible predictions. For our purposes, two singlepremise arguments with the same conclusion are deemed incompatible if one leads to a probability greater than 0.9 according to Equation 3, and the other leads to a probability less than 0.1. Edge cases are arguments f i , f j → f k such that the feature network in Figure 2b includes an edge between f k and either f i or f j . Note that some arguments are both conflict cases and edge cases. All arguments that do not fall into either one of these categories will be referred to as other cases. The 400 arguments for the experiment include 154 conflict cases, 153 edge cases, and 120 other cases. 34 arguments are both conflict cases and edge cases. We chose these arguments based on three criteria. First, we avoided premise pairs that did not co-occur in matrix D but that co-occur in familiar animals that do not belong to D. For example, “is pink” and “has wings” do not co-occur in D but “flamingo” is a familiar animal that has both features. Second, we avoided premise pairs that specified two different numbers of legs—for example, {“has four legs,” “has six legs”}. Finally, we aimed to include roughly equal numbers of conflict cases, edge cases, and other cases. Method. 16 undergraduates participated for course credit. The experiment was carried out using a custom-built computer interface, and one argument was presented on screen at a time. Participants 6 rated the probability of the conclusion on seven point scale where the endpoints were labeled “very unlikely” and “very likely.” The ten arguments for each pair of premises were presented in a block, but the order of these blocks and the order of the arguments within these blocks were randomized across participants. Results. Figure 3 shows average human judgments plotted against the predictions of six models. The plots in the first row include all 400 arguments in the experiment, and the remaining rows show results for conflict cases, edge cases, and other cases. The previous section described four exemplar models, and the two shown in Figure 3 are the best performers overall. Even though the graphical models include more numerical parameters than the exemplar models, recall that these parameters are learned from matrix D rather than fit to the experimental data. Matrix D also serves as the basis for the exemplar models, which means that all of the models can be compared on equal terms. The first row of Figure 3 suggests that the three models which include feature structure F perform better than the alternatives. The output combination model is the worst of the three models that incorporate F, and the correlation achieved by this model is significantly greater than the correlation achieved by the best exemplar model (p < 0.001, using the Fisher transformation to convert correlation coefficients to z scores). Our data therefore suggest that explicit representations of relationships between features are needed to account for inductive inferences about chimeras. The model that includes the feature structure F alone performs better than the two models that combine F with the object structure O, which may not be surprising since Experiment 1 focuses specifically on novel animals that do not slot naturally into structure O. Rows two through four suggest that the conflict arguments in particular raise challenges for the models which do not include feature structure F. Since these conflict cases are arguments f i , f j → f k where f i → f k has strength greater than 0.9 and f j → f k has strength less than 0.1, the first exemplar model averages these strengths and assigns an overall strength of around 0.5 to each argument. The second exemplar model is better able to differentiate between the conflict arguments, but still performs substantially worse than the three models that include structure F. The exemplar models perform better on the edge arguments, but are outperformed by the models that include F. Finally, all models achieve roughly the same level of performance on the other arguments. Although the feature model F performs best overall, the predictions of this model still leave room for improvement. The two most obvious outliers in the third plot in the top row represent the arguments {is blue, lives in desert → lives in woods} and {is pink, lives in desert → lives in woods}. Our participants sensibly infer that any animal which lives in the desert cannot simultaneously live in the woods. In contrast, the Leuven database indicates that eight of the twelve animals that live in the desert also live in the woods, and the edge in Figure 2b between “lives in the desert” and “lives in the woods” therefore represents a positive dependency relationship according to model F. This discrepancy between model and participants reflects the fact that participants made inferences about individual animals but the Leuven database is based on features of animal categories. Note, for example, that any individual animal is unlikely to live in the desert and the woods, but that some animal categories (including snakes, salamanders, and lizards) are found in both environments. 3 Experiment 2: Single-premise arguments Our results so far suggest that inferences about chimeras rely on explicit representations of relationships between features but provide no evidence that relationships between objects are important. It would be a mistake, however, to conclude that relationships between objects play no role in inductive reasoning. Previous studies have used object structures like the example in Figure 2a to account for inferences about novel features [11]—for example, given that alligators have enzyme Y132 in their blood, it seems likely that crocodiles also have this enzyme. Inferences about novel objects can also draw on relationships between objects rather than relationships between features. For example, given that a novel animal has a beak you will probably predict that it has feathers, not because there is any direct dependency between these two features, but because the beaked animals that you know tend to have feathers. Our second experiment explores inferences of this kind. Materials and Method. 32 undergraduates participated for course credit. The task was identical to Experiment 1 with the following exceptions. Each two-premise argument f i , f j → f k from Experiment 1 was converted into two one-premise arguments f i → f k and f j → f k , and these 7 feature F exemplar r = 0.78 7 object O r = 0.54 7 output combination r = 0.75 7 structure combination r = 0.75 7 all 5 5 5 5 5 3 3 3 3 3 1 1 0 edge 0.5 1 r = 0.87 7 1 0 0.5 1 r = 0.87 7 1 0 0.5 1 r = 0.84 7 1 0 0.5 1 r = 0.86 7 0 5 5 5 3 3 3 1 5 3 0.5 r = 0.85 7 5 3 1 1 0 0.5 1 r = 0.79 7 other r = 0.77 7 1 0 0.5 1 r = 0.21 7 1 0 0.5 1 r = 0.74 7 1 0 0.5 1 r = 0.66 7 0 5 5 5 5 3 3 3 3 1 r = 0.73 7 5 0.5 3 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 1 0 0.5 1 0 0.5 1 Figure 4: Argument ratings and model predictions for Experiment 2. one-premise arguments were randomly assigned to two sets. 16 participants rated the 400 arguments in the first set, and the other 16 rated the 400 arguments in the second set. Results. Figure 4 shows average human ratings for the 800 arguments plotted against the predictions of five models. Unlike Figure 3, Figure 4 includes a single exemplar model since there is no need to consider different feature weightings in this case. Unlike Experiment 1, the feature model F performs worse than the other alternatives (p < 0.001 in all cases). Not surprisingly, this model performs relatively well for edge cases f j → f k where f j and f k are linked in Figure 2b, but the final row shows that the model performs poorly across the remaining set of arguments. Taken together, Experiments 1 and 2 suggest that relationships between objects and relationships between features are both needed to account for human inferences. Experiment 1 rules out an exemplar approach but models that combine graph structures over objects and features perform relatively well in both experiments. We considered two methods for combining these structures and both performed equally well. Combining the knowledge captured by these structures appears to be important, and future studies can explore in detail how humans achieve this combination. 4 Conclusion This paper proposed that graphical models are useful for capturing knowledge about animals and their features and showed that a graphical model over features can account for human inferences about chimeras. A family of exemplar models and a graphical model defined over objects were unable to account for our data, which suggests that humans rely on mental representations that explicitly capture dependency relationships between features. Psychologists have previously used graphical models to capture relationships between features, but our work is the first to focus on chimeras and to explore models defined over a large set of familiar features. Although a simple undirected model accounted relatively well for our data, this model is only a starting point. The model incorporates dependency relationships between features, but people know about many specific kinds of dependencies, including cases where one feature causes, enables, prevents, or is inconsistent with another. An undirected graph with only one class of edges cannot capture this knowledge in full, and richer representations will ultimately be needed in order to provide a more complete account of human reasoning. Acknowledgments I thank Madeleine Clute for assisting with this research. This work was supported in part by the Pittsburgh Life Sciences Greenhouse Opportunity Fund and by NSF grant CDI-0835797. 8 References [1] R. N. Shepard. Towards a universal law of generalization for psychological science. Science, 237:1317– 1323, 1987. [2] J. R. Anderson. The adaptive nature of human categorization. Psychological Review, 98(3):409–429, 1991. [3] E. Heit. A Bayesian analysis of some forms of inductive reasoning. In M. Oaksford and N. Chater, editors, Rational models of cognition, pages 248–274. Oxford University Press, Oxford, 1998. [4] J. B. Tenenbaum and T. L. Griffiths. Generalization, similarity, and Bayesian inference. Behavioral and Brain Sciences, 24:629–641, 2001. [5] C. Kemp and J. B. Tenenbaum. Structured statistical models of inductive reasoning. Psychological Review, 116(1):20–58, 2009. [6] D. N. Osherson, E. E. Smith, O. Wilkie, A. Lopez, and E. Shafir. Category-based induction. Psychological Review, 97(2):185–200, 1990. [7] D. J. Navarro. Learning the context of a category. In J. Lafferty, C. K. I. Williams, J. Shawe-Taylor, R.S. Zemel, and A. Culotta, editors, Advances in Neural Information Processing Systems 23, pages 1795–1803. 2010. [8] C. Kemp, T. L. Griffiths, S. Stromsten, and J. B. Tenenbaum. Semi-supervised learning with trees. In Advances in Neural Information Processing Systems 16, pages 257–264. MIT Press, Cambridge, MA, 2004. [9] B. Rehder. A causal-model theory of conceptual representation and categorization. Journal of Experimental Psychology: Learning, Memory, and Cognition, 29:1141–1159, 2003. [10] B. Rehder and R. Burnett. Feature inference and the causal structure of categories. Cognitive Psychology, 50:264–314, 2005. [11] C. Kemp, P. Shafto, and J. B. Tenenbaum. An integrated account of generalization across objects and features. Cognitive Psychology, in press. [12] S. E. Barrett, H. Abdi, G. L. Murphy, and J. McCarthy Gallagher. Theory-based correlations and their role in children’s concepts. Child Development, 64:1595–1616, 1993. [13] S. A. Sloman, B. C. Love, and W. Ahn. Feature centrality and conceptual coherence. Cognitive Science, 22(2):189–228, 1998. [14] D. Yarlett and M. Ramscar. A quantitative model of counterfactual reasoning. In T. G. Dietterich, S. Becker, and Z. Ghahramani, editors, Advances in Neural Information Processing Systems 14, pages 123–130. MIT Press, Cambridge, MA, 2002. [15] W. Ahn, J. K. Marsh, C. C. Luhmann, and K. Lee. Effect of theory-based feature correlations on typicality judgments. Memory and Cognition, 30(1):107–118, 2002. [16] D. C. Meehan C. McNorgan, R. A. Kotack and K. McRae. Feature-feature causal relations and statistical co-occurrences in object concepts. Memory and Cognition, 35(3):418–431, 2007. [17] S. De Deyne, S. Verheyen, E. Ameel, W. Vanpaemel, M. J. Dry, W. Voorspoels, and G. Storms. Exemplar by feature applicability matrices and other Dutch normative data for semantic concepts. Behavior Research Methods, 40(4):1030–1048, 2008. [18] J. P. Huelsenbeck and F. Ronquist. MRBAYES: Bayesian inference of phylogenetic trees. Bioinformatics, 17(8):754–755, 2001. [19] M. Schmidt. UGM: A Matlab toolbox for probabilistic undirected graphical models. 2007. Available at http://people.cs.ubc.ca/∼schmidtm/Software/UGM.html. [20] L. J. Nelson and D. T. Miller. The distinctiveness effect in social categorization: you are what makes you unusual. Psychological Science, 6:246–249, 1995. [21] A. L. Patalano, S. Chin-Parker, and B. H. Ross. The importance of being coherent: category coherence, cross-classification and reasoning. Journal of memory and language, 54:407–424, 2006. [22] S. K. Reed. Pattern recognition and categorization. Cognitive Psychology, 3:393–407, 1972. 9

6 0.55144322 122 nips-2011-How Do Humans Teach: On Curriculum Learning and Teaching Dimension

7 0.49476022 3 nips-2011-A Collaborative Mechanism for Crowdsourcing Prediction Problems

8 0.47471535 293 nips-2011-Understanding the Intrinsic Memorability of Images

9 0.44591507 35 nips-2011-An ideal observer model for identifying the reference frame of objects

10 0.44466203 219 nips-2011-Predicting response time and error rates in visual search

11 0.39067742 126 nips-2011-Im2Text: Describing Images Using 1 Million Captioned Photographs

12 0.38977662 235 nips-2011-Recovering Intrinsic Images with a Global Sparsity Prior on Reflectance

13 0.37430614 224 nips-2011-Probabilistic Modeling of Dependencies Among Visual Short-Term Memory Representations

14 0.37195307 184 nips-2011-Neuronal Adaptation for Sampling-Based Probabilistic Inference in Perceptual Bistability

15 0.3545717 111 nips-2011-Hashing Algorithms for Large-Scale Learning

16 0.35067096 66 nips-2011-Crowdclustering

17 0.34349465 214 nips-2011-PiCoDes: Learning a Compact Code for Novel-Category Recognition

18 0.34200519 157 nips-2011-Learning to Search Efficiently in High Dimensions

19 0.3254182 194 nips-2011-On Causal Discovery with Cyclic Additive Noise Models

20 0.32515249 216 nips-2011-Portmanteau Vocabularies for Multi-Cue Image Representation


similar papers computed by lda model

lda for this paper:

topicId topicWeight

[(0, 0.015), (4, 0.077), (20, 0.022), (26, 0.035), (27, 0.225), (31, 0.074), (33, 0.048), (43, 0.056), (45, 0.115), (56, 0.015), (57, 0.067), (74, 0.056), (83, 0.066), (97, 0.01), (99, 0.03)]

similar papers list:

simIndex simValue paperId paperTitle

1 0.8738429 6 nips-2011-A Global Structural EM Algorithm for a Model of Cancer Progression

Author: Ali Tofigh, Erik Sj̦lund, Mattias H̦glund, Jens Lagergren

Abstract: Cancer has complex patterns of progression that include converging as well as diverging progressional pathways. Vogelstein’s path model of colon cancer was a pioneering contribution to cancer research. Since then, several attempts have been made at obtaining mathematical models of cancer progression, devising learning algorithms, and applying these to cross-sectional data. Beerenwinkel et al. provided, what they coined, EM-like algorithms for Oncogenetic Trees (OTs) and mixtures of such. Given the small size of current and future data sets, it is important to minimize the number of parameters of a model. For this reason, we too focus on tree-based models and introduce Hidden-variable Oncogenetic Trees (HOTs). In contrast to OTs, HOTs allow for errors in the data and thereby provide more realistic modeling. We also design global structural EM algorithms for learning HOTs and mixtures of HOTs (HOT-mixtures). The algorithms are global in the sense that, during the M-step, they find a structure that yields a global maximum of the expected complete log-likelihood rather than merely one that improves it. The algorithm for single HOTs performs very well on reasonable-sized data sets, while that for HOT-mixtures requires data sets of sizes obtainable only with tomorrow’s more cost-efficient technologies. 1

2 0.85191113 192 nips-2011-Nonstandard Interpretations of Probabilistic Programs for Efficient Inference

Author: David Wingate, Noah Goodman, Andreas Stuhlmueller, Jeffrey M. Siskind

Abstract: Probabilistic programming languages allow modelers to specify a stochastic process using syntax that resembles modern programming languages. Because the program is in machine-readable format, a variety of techniques from compiler design and program analysis can be used to examine the structure of the distribution represented by the probabilistic program. We show how nonstandard interpretations of probabilistic programs can be used to craft efficient inference algorithms: information about the structure of a distribution (such as gradients or dependencies) is generated as a monad-like side computation while executing the program. These interpretations can be easily coded using special-purpose objects and operator overloading. We implement two examples of nonstandard interpretations in two different languages, and use them as building blocks to construct inference algorithms: automatic differentiation, which enables gradient based methods, and provenance tracking, which enables efficient construction of global proposals. 1

3 0.82613117 304 nips-2011-Why The Brain Separates Face Recognition From Object Recognition

Author: Joel Z. Leibo, Jim Mutch, Tomaso Poggio

Abstract: Many studies have uncovered evidence that visual cortex contains specialized regions involved in processing faces but not other object classes. Recent electrophysiology studies of cells in several of these specialized regions revealed that at least some of these regions are organized in a hierarchical manner with viewpointspecific cells projecting to downstream viewpoint-invariant identity-specific cells [1]. A separate computational line of reasoning leads to the claim that some transformations of visual inputs that preserve viewed object identity are class-specific. In particular, the 2D images evoked by a face undergoing a 3D rotation are not produced by the same image transformation (2D) that would produce the images evoked by an object of another class undergoing the same 3D rotation. However, within the class of faces, knowledge of the image transformation evoked by 3D rotation can be reliably transferred from previously viewed faces to help identify a novel face at a new viewpoint. We show, through computational simulations, that an architecture which applies this method of gaining invariance to class-specific transformations is effective when restricted to faces and fails spectacularly when applied to other object classes. We argue here that in order to accomplish viewpoint-invariant face identification from a single example view, visual cortex must separate the circuitry involved in discounting 3D rotations of faces from the generic circuitry involved in processing other objects. The resulting model of the ventral stream of visual cortex is consistent with the recent physiology results showing the hierarchical organization of the face processing network. 1

same-paper 4 0.78448194 280 nips-2011-Testing a Bayesian Measure of Representativeness Using a Large Image Database

Author: Joshua T. Abbott, Katherine A. Heller, Zoubin Ghahramani, Thomas L. Griffiths

Abstract: How do people determine which elements of a set are most representative of that set? We extend an existing Bayesian measure of representativeness, which indicates the representativeness of a sample from a distribution, to define a measure of the representativeness of an item to a set. We show that this measure is formally related to a machine learning method known as Bayesian Sets. Building on this connection, we derive an analytic expression for the representativeness of objects described by a sparse vector of binary features. We then apply this measure to a large database of images, using it to determine which images are the most representative members of different sets. Comparing the resulting predictions to human judgments of representativeness provides a test of this measure with naturalistic stimuli, and illustrates how databases that are more commonly used in computer vision and machine learning can be used to evaluate psychological theories. 1

5 0.63686466 127 nips-2011-Image Parsing with Stochastic Scene Grammar

Author: Yibiao Zhao, Song-chun Zhu

Abstract: This paper proposes a parsing algorithm for scene understanding which includes four aspects: computing 3D scene layout, detecting 3D objects (e.g. furniture), detecting 2D faces (windows, doors etc.), and segmenting background. In contrast to previous scene labeling work that applied discriminative classifiers to pixels (or super-pixels), we use a generative Stochastic Scene Grammar (SSG). This grammar represents the compositional structures of visual entities from scene categories, 3D foreground/background, 2D faces, to 1D lines. The grammar includes three types of production rules and two types of contextual relations. Production rules: (i) AND rules represent the decomposition of an entity into sub-parts; (ii) OR rules represent the switching among sub-types of an entity; (iii) SET rules represent an ensemble of visual entities. Contextual relations: (i) Cooperative “+” relations represent positive links between binding entities, such as hinged faces of a object or aligned boxes; (ii) Competitive “-” relations represents negative links between competing entities, such as mutually exclusive boxes. We design an efficient MCMC inference algorithm, namely Hierarchical cluster sampling, to search in the large solution space of scene configurations. The algorithm has two stages: (i) Clustering: It forms all possible higher-level structures (clusters) from lower-level entities by production rules and contextual relations. (ii) Sampling: It jumps between alternative structures (clusters) in each layer of the hierarchy to find the most probable configuration (represented by a parse tree). In our experiment, we demonstrate the superiority of our algorithm over existing methods on public dataset. In addition, our approach achieves richer structures in the parse tree. 1

6 0.636594 64 nips-2011-Convergent Bounds on the Euclidean Distance

7 0.63490301 150 nips-2011-Learning a Distance Metric from a Network

8 0.63371909 231 nips-2011-Randomized Algorithms for Comparison-based Search

9 0.63205862 219 nips-2011-Predicting response time and error rates in visual search

10 0.63063616 227 nips-2011-Pylon Model for Semantic Segmentation

11 0.62795419 1 nips-2011-$\theta$-MRF: Capturing Spatial and Semantic Structure in the Parameters for Scene Understanding

12 0.6262579 149 nips-2011-Learning Sparse Representations of High Dimensional Data on Large Scale Dictionaries

13 0.62577701 43 nips-2011-Bayesian Partitioning of Large-Scale Distance Data

14 0.62412989 303 nips-2011-Video Annotation and Tracking with Active Learning

15 0.62376434 17 nips-2011-Accelerated Adaptive Markov Chain for Partition Function Computation

16 0.62370843 253 nips-2011-Signal Estimation Under Random Time-Warpings and Nonlinear Signal Alignment

17 0.62291443 258 nips-2011-Sparse Bayesian Multi-Task Learning

18 0.62219548 180 nips-2011-Multiple Instance Filtering

19 0.62198961 168 nips-2011-Maximum Margin Multi-Instance Learning

20 0.62176043 135 nips-2011-Information Rates and Optimal Decoding in Large Neural Populations