Optimal predictions in everyday cognition

Human perception and memory are often explained as optimal statistical inferences that are informed by accurate prior probabilities. In contrast, cognitive judgments are usually viewed as following error-prone heuristics that are insensitive to priors. We examined the optimality of human cognition in a more realistic context than typical laboratory studies, asking people to make predictions about the duration or extent of everyday phenomena such as human life spans and the box-office take of movies. Our results suggest that everyday cognitive judgments follow the same optimal statistical principles as perception and memory, and reveal a close correspondence between people's implicit probabilistic models and the statistics of the world.     

Identifying expectations about the strength of causal relationships

When we try to identify causal relationships, how strong do we expect that relationship to be? Bayesian models of causal induction rely on assumptions regarding people’s a priori beliefs about causal systems, with recent research focusing on people’s expectations about the strength of causes. These expectations are expressed in terms of prior probability distributions. While proposals about the form of such prior distributions have been made previously, many different distributions are possible, making it difficult to test such proposals exhaustively. In Experiment 1 we used iterated learning—a method in which participants make inferences about data generated based on their own responses in previous trials—to estimate participants’ prior beliefs about the strengths of causes. This method produced estimated prior distributions that were quite different from those previously proposed in the literature. Experiment 2 collected a large set of human judgments on the strength of causal relationships to be used as a benchmark for evaluating different models, using stimuli that cover a wider and more systematic set of contingencies than previous research. Using these judgments, we evaluated the predictions of various Bayesian models. The Bayesian model with priors estimated via iterated learning compared favorably against the others. Experiment 3 estimated participants’ prior beliefs concerning different causal systems, revealing key similarities in their expectations across diverse scenarios.     

Feature frequency in concept learning: What is counted?

Frequency theories of concept learning assume that people count how often features occur among instances of a concept, but different versions make various assumptions about what features they count. According to the basic feature model, only basic features are counted, whereas according to the configural model, basic features and configural features (all combinations of basic features) are counted. Two experiments assessed the predictions of both versions of frequency theory. Subjects viewed schematic human faces, which included both positive and negative instances of the concept to be learned, and then provided typicality ratings, classification responses, and frequency estimates of configural features, basic features, and whole exemplars. Because both models assume that basic features are counted, they make the same predictions in many situations. Here, the basic feature estimation and whole exemplar tests were designed such that both models make the same predictions, whereas the typicality rating, classification, and confignral feature estimation tests were designed to distinguish between the models. The pattern of results clearly supported the basic feature version of frequency theory.     

Predicting similarity judgments in intertemporal choice with machine learning

Similarity models of intertemporal choice are heuristics that choose based on similarity judgments of the reward amounts and time delays. Yet, we do not know how these judgments are made. Here, we use machine-learning algorithms to assess what factors predict similarity judgments and whether decision trees capture the judgment outcomes and process. We find that combining small and large values into numerical differences and ratios and arranging them in tree-like structures can predict both similarity judgments and response times. Our results suggest that we can use machine learning to not only model decision outcomes but also model how decisions are made. Revealing how people make these important judgments may be useful in developing interventions to help them make better decisions.     

Thinking about memories for everyday and shocking events: Do people use ease-of-retrieval cues in memory judgments?

Extant research shows that people use retrieval ease, a feeling-based cue, to judge how well they remember life periods. Extending this approach, we investigated the role of retrieval ease in memory judgments for single events. In Experiment 1, participants who were asked to recall many memories of an everyday event (New Year's Eve) rated retrieval as more difficult and judged their memory as worse than did participants asked to recall only a few memories. In Experiment 2, this ease-of-retrieval effect was found to interact with the shocking character of the remembered event: There was no effect when the event was highly shocking (i.e., learning about the attacks of September 11, 2001), whereas an effect was found when the event was experienced as less shocking (due either to increased distance to "9/11" or to the nonshocking nature of the event itself). Memory vividness accounted for additional variance in memory judgments, indicating an independent contribution of content-based cues in judgments of event memories.     

Toward a more pragmatic approach to morality: a critical evaluation of Kohlberg's model

In this article, the authors evaluate L. Kohlberg's (1984) cognitive- developmental approach to morality, find it wanting, and introduce a more pragmatic approach. They review research designed to evaluate Kohlberg's model, describe how they revised the model to accommodate discrepant findings, and explain why they concluded that it is poorly equipped to account for the ways in which people make moral decisions in their everyday lives. The authors outline in 11 propositions a framework for a new approach that is more attentive to the purposes that people use morality to achieve. People make moral judgments and engage in moral behaviors to induce themselves and others to uphold systems of cooperative exchange that help them achieve their goals and advance their interests.     

延迟学习判断研究的发展与展望

延迟学习判断是学习判断的一种形式,是指在材料学习完以后间隔一段时间才发生的学习判断。在与即时学习判断的对比研究中发现,延迟学习判断具有较高的相对准确性,这种现象被称为延迟学习判断效应。研究者进行了大量的研究并提出了多种理论来解释这种延迟学习判断效应。随着研究的不断深入,延迟学习判断的研究从研究指标、研究方法甚至是研究的理论基础都在不断更新。延迟学习判断的研究进展,包括主要理论和相关实验,以及最新研究成果将被介绍。最后,文章梳理了延迟JOL的研究进程,并指出了未来的研究方向。     

SUN: Top-down saliency using natural statistics

When people try to find particular objects in natural scenes they make extensive use of knowledge about how and where objects tend to appear in a scene. Although many forms of such "top-down" knowledge have been incorporated into saliency map models of visual search, surprisingly, the role of object appearance has been infrequently investigated. Here we present an appearance-based saliency model derived in a Bayesian framework. We compare our approach with both bottom-up saliency algorithms as well as the state-of-the-art Contextual Guidance model of Torralba et al. (2006) at predicting human fixations. Although both top-down approaches use very different types of information, they achieve similar performance; each substantially better than the purely bottom-up models. Our experiments reveal that a simple model of object appearance can predict human fixations quite well, even making the same mistakes as people.     

Learning the form of causal relationships using hierarchical bayesian models

People learn quickly when reasoning about causal relationships, making inferences from limited data and avoiding spurious inferences. Efficient learning depends on abstract knowledge, which is often domain or context specific, and much of it must be learned. While such knowledge effects are well documented, little is known about exactly how we acquire knowledge that constrains learning. This work focuses on knowledge of the functional form of causal relationships; there are many kinds of relationships that can apply between causes and their effects, and knowledge of the form such a relationship takes is important in order to quickly identify the real causes of an observed effect. We developed a hierarchical Bayesian model of the acquisition of knowledge of the functional form of causal relationships and tested it in five experimental studies, considering disjunctive and conjunctive relationships, failure rates, and cross-domain effects. The Bayesian model accurately predicted human judgments and outperformed several alternative models.     

Similarity and causality in the use of social analogies

The present paper examines whether causal relations are necessary for the use of social analogies. Although causal relations increase the use of social analogies, it is not known whether they are necessary. Establishing this is important for understanding both analogical reasoning and learning in novel situations where people lack knowledge of the causal relations. Study 1 demonstrated that, in the absence of an explicit causal relation, as the similarity between a target individual and a previously encountered individual (the base) increased, people thought the target was increasingly likely to perform the same behavior as the base. Thus, a causal relation is not necessary. Two additional studies used asymmetries in similarity judgments to provide additional evidence that when reasoning analogically people relied on similarity. Manipulating whether subjects focused primarily on the target or the base completely reversed asymmetries both in judgments of how similar the target was to the base and in predictions of how likely the target was to perform the same behavior as the base. The asymmetries for similarity and prediction were completely parallel. Thus, in the absence of an explicit causal relation, use of the analogy was based on judgments of global similarity. The implications of these asymmetries in similarity judgments and predictions for other judgments, such as stereotyping, are also discussed.     

The dilution effect: Nondiagnostic information weakens the implications of diagnostic information

In a series of studies, subjects were asked to make predictions about target individuals. Some subjects were given information about the target which pretest subjects had judged to be “diagnostic”—that is, had judged to be usefully predictive of the outcome. Other subjects were given a mix of information judged to be diagnostic and information judged to be “nondiagnostic” by pretest subjects—that is, judged to be of little value for predicting the outcome. Subjects given mixed information made much less extreme predictions than did subjects given only diagnostic information. It was argued that this “dilution effect” occurs because people make predictions by making simple similarity judgments. That is, they compare the information they have about the target with their conception of outcome categories. The presence of individuating but nondiagnostic information about the target reduces the similarity between the target and those outcomes that are suggested by the diagnostic information. One of the major implications is that stereotypes and other “social knowledge structures” may be applied primarily to abstract, undifferentiated individuals and groups and may be largely set aside when judgments are made about concrete, individuated people.     

A hierarchical bayesian model of human decision-making on an optimal stopping problem

We consider human performance on an optimal stopping problem where people are presented with a list of numbers independently chosen from a uniform distribution. People are told how many numbers are in the list, and how they were chosen. People are then shown the numbers one at a time, and are instructed to choose the maximum, subject to the constraint that they must choose a number at the time it is presented, and any choice below the maximum is incorrect. We present empirical evidence that suggests people use threshold-based models to make decisions, choosing the first currently maximal number that exceeds a fixed threshold for that position in the list. We then develop a hierarchical generative account of this model family, and use Bayesian methods to learn about the parameters of the generative process, making inferences about the threshold decision models people use. We discuss the interesting aspects of human performance on the task, including the lack of learning, and the presence of large individual differences, and consider the possibility of extending the modeling framework to account for individual differences. We also use the modeling results to discuss the merits of hierarchical, generative and Bayesian models of cognitive processes more generally.     

Insight and strategy in multiple-cue learning

In multiple-cue learning (also known as probabilistic category learning) people acquire information about cue-outcome relations and combine these into predictions or judgments. Previous researchers claimed that people can achieve high levels of performance without explicit knowledge of the task structure or insight into their own judgment policies. It has also been argued that people use a variety of suboptimal strategies to solve such tasks. In three experiments the authors reexamined these conclusions by introducing novel measures of task knowledge and self-insight and using "rolling regression" methods to analyze individual learning. Participants successfully learned a four-cue probabilistic environment and showed accurate knowledge of both the task structure and their own judgment processes. Learning analyses suggested that the apparent use of suboptimal strategies emerges from the incremental tracking of statistical contingencies in the environment.     

An empirical and computational investigation of perceiving and remembering event temporal relations

Events have beginnings, ends, and often overlap in time. A major question is how perceivers come to parse a stream of multimodal information into meaningful units and how different event boundaries may vary event processing. This work investigates the roles of these three types of event boundaries in constructing event temporal relations. Predictions were made based on how people would err according to the beginning state, end state, and overlap heuristic hypotheses. Participants viewed animated events that include all the logical possibilities of event temporal relations, and then made temporal relation judgments. The results showed that people make use of the overlap between events and take into account the ends and beginnings, but they weight ends more than beginnings. Neural network simulations showed a self-organized distinction when learning temporal relations between events with overlap versus those without.     

A rational model of the effects of distributional information on feature learning

Most psychological theories treat the features of objects as being fixed and immediately available to observers. However, novel objects have an infinite array of properties that could potentially be encoded as features, raising the question of how people learn which features to use in representing those objects. We focus on the effects of distributional information on feature learning, considering how a rational agent should use statistical information about the properties of objects in identifying features. Inspired by previous behavioral results on human feature learning, we present an ideal observer model based on nonparametric Bayesian statistics. This model balances the idea that objects have potentially infinitely many features with the goal of using a relatively small number of features to represent any finite set of objects. We then explore the predictions of this ideal observer model. In particular, we investigate whether people are sensitive to how parts co-vary over objects they observe. In a series of four behavioral experiments (three using visual stimuli, one using conceptual stimuli), we demonstrate that people infer different features to represent the same four objects depending on the distribution of parts over the objects they observe. Additionally in all four experiments, the features people infer have consequences for how they generalize properties to novel objects. We also show that simple models that use the raw sensory data as inputs and standard dimensionality reduction techniques (principal component analysis and independent component analysis) are insufficient to explain our results.     

A rational model of the effects of distributional information on feature learning

Most psychological theories treat the features of objects as being fixed and immediately available to observers. However, novel objects have an infinite array of properties that could potentially be encoded as features, raising the question of how people learn which features to use in representing those objects. We focus on the effects of distributional information on feature learning, considering how a rational agent should use statistical information about the properties of objects in identifying features. Inspired by previous behavioral results on human feature learning, we present an ideal observer model based on nonparametric Bayesian statistics. This model balances the idea that objects have potentially infinitely many features with the goal of using a relatively small number of features to represent any finite set of objects. We then explore the predictions of this ideal observer model. In particular, we investigate whether people are sensitive to how parts co-vary over objects they observe. In a series of four behavioral experiments (three using visual stimuli, one using conceptual stimuli), we demonstrate that people infer different features to represent the same four objects depending on the distribution of parts over the objects they observe. Additionally in all four experiments, the features people infer have consequences for how they generalize properties to novel objects. We also show that simple models that use the raw sensory data as inputs and standard dimensionality reduction techniques (principal component analysis and independent component analysis) are insufficient to explain our results.     

The Effect of Inference Order and Experience-Related Knowledge on Diagnostic Conjunction Probabilities

Ideally, a decision maker′s diagnostic probability judgments should not be affected by making predictive judgments before making diagnostic inferences. The purpose of this study is to investigate how experience-related knowledge and the inference presentation order affect a decision maker′s diagnostic conjunction probability judgments. Specifically, when decision makers are asked to make diagnoses in different judgment domains with which they have different levels of experience, we examine how making predictions first affects their subsequent diagnostic judgments in a standard conjunction paradigm. Professional auditors with experience in the auditing domain and MBA students with little or no auditing experience participated in the experiment. The results indicate that when the task involves a domain with which people have experience, making predictions prior to diagnoses has a significant influence on their subsequent diagnostic conjunction probabilities. When auditors made diagnoses in a familiar audit task situation, they were strongly influenced by whether or not they were asked to make predictions in advance. However, there was no influence of inference order on auditors′ diagnoses in a medical task, with which they do not have experience-related knowledge. Similarly, MBA students, having no experience-related knowledge in either audit or medical domains, were not affected by the inference order in making diagnoses. In the discussion of these exploratory results, we suggest that this inference order effect may be due to subjects′ anchoring on the predictive probability and insufficiently adjusting it to yield the diagnostic probability judgment.     

The role of learning data in causal reasoning about observations and interventions

Recent studies have shown that people have the capacity to derive interventional predictions for previously unseen actions from observational knowledge, a finding that challenges associative theories of causal learning and reasoning (e.g., Meder, Hagmayer, & Waldmann, 2008). Although some researchers have claimed that such inferences are based mainly on qualitative reasoning about the structure of a causal system (e.g., Sloman, 2005), we propose that people use both the causal structure and its parameters for their inferences. We here employ an observational trial-by-trial learning paradigm to test this prediction. In Experiment 1, the causal strength of the links within a given causal model was varied, whereas in Experiment 2, base rate information was manipulated while keeping the structure of the model constant. The results show that learners’ causal judgments were strongly affected by the observed learning data despite being presented with identical hypotheses about causal structure. The findings show furthermore that participants correctly distinguished between observations and hypothetical interventions. However, they did not adequately differentiate between hypothetical and counterfactual interventions.     

Frequency, probability, and prediction: easy solutions to cognitive illusions?

Many errors in probabilistic judgment have been attributed to people's inability to think in statistical terms when faced with information about a single case. Prior theoretical analyses and empirical results imply that the errors associated with case-specific reasoning may be reduced when people make frequentistic predictions about a set of cases. In studies of three previously identified cognitive biases, we find that frequency-based predictions are different from-but no better than-case-specific judgments of probability. First, in studies of the "planning fallacy, " we compare the accuracy of aggregate frequency and case-specific probability judgments in predictions of students' real-life projects. When aggregate and single-case predictions are collected from different respondents, there is little difference between the two: Both are overly optimistic and show little predictive validity. However, in within-subject comparisons, the aggregate judgments are significantly more conservative than the single-case predictions, though still optimistically biased. Results from studies of overconfidence in general knowledge and base rate neglect in categorical prediction underline a general conclusion. Frequentistic predictions made for sets of events are no more statistically sophisticated, nor more accurate, than predictions made for individual events using subjective probability.     

Does easily learned mean easily remembered?: it depends on your beliefs about intelligence

Because numerous studies have shown that feelings of encoding fluency are positively correlated with judgments of learning, a single dominant heuristic, easily learned = easily remembered (ELER), has been posited to explain how people interpret encoding fluency when assessing their own memory. However, the inferences people draw from feelings of encoding fluency may vary with their beliefs about why information is easy or effortful to encode. We conducted two experiments in which participants studied word lists and then predicted their future recall of those items. Results revealed that subjects who viewed intelligence as fixed, and who tended to interpret effortful encoding as indicating that they had reached the limits of their ability, used the ELER heuristic to make judgments of learning. However, subjects who viewed intelligence as malleable, and who tended to interpret effortful encoding as indicating greater engagement in learning, did not use the ELER heuristic and at times predicted greater memory for items that they found more effortful to learn.