Mechanistic explanation has an impressive track record of advancing our understanding of complex, hierarchically organized physical systems, particularly biological and neural systems. But not every complex system can be understood mechanistically. Psychological capacities are often understood by providing cognitive models of the systems that underlie them. I argue that these models, while superficially similar to mechanistic models, in fact have a substantially more complex relation to the real underlying system. They are typically constructed using a range of techniques for abstracting (...) the functional properties of the system, which may not coincide with its mechanistic organization. I describe these techniques and show that despite being non-mechanistic, these cognitive models can satisfy the normative constraints on good explanations. (shrink)

A central controversy in cognitive science concerns the roles of rules versus similarity. To gain some leverage on this problem, we propose that rule‐ versus similarity‐based processes can be characterized as extremes in a multidimensional space that is composed of at least two dimensions: the number of features (Pothos, 2005) and the physical presence of features. The transition of similarity‐ to rule‐based processing is conceptualized as a transition in this space. To illustrate this, we show how a neural network model (...) uses input features (and in this sense produces similarity‐based responses) when it has a low learning rate or in the early phases of training, but it switches to using self‐generated, more abstract features (and in this sense produces rule‐based responses) when it has a higher learning rate or is in the later phases of training. Relations with categorization and the psychology of learning are pointed out. (shrink)

In this prcis we focus on phenomena central to the reaction against similarity-based theories that arose in the 1980s and that subsequently motivated the approach to semantic knowledge. Specifically, we consider (1) how concepts differentiate in early development, (2) why some groupings of items seem to form or coherent categories while others do not, (3) why different properties seem central or important to different concepts, (4) why children and adults sometimes attest to beliefs that seem to contradict their direct experience, (...) (5) how concepts reorganize between the ages of 4 and 10, and (6) the relationship between causal knowledge and semantic knowledge. The explanations our theory offers for these phenomena are illustrated with reference to a simple feed-forward connectionist model. The relationships between this simple model, the broader theory, and more general issues in cognitive science are discussed. (shrink)

Acquiring perceptual expertise is slow and effortful. However, untrained novices can accurately make difficult classification decisions by reformulating the task as similarity judgment. Given a query image and a set of reference images, individuals are asked to select the best matching reference. When references are suitably chosen, the procedure yields an implicit classification of the query image. To optimize reference selection, we develop and evaluate a predictive model of similarity-based choice. The model builds on existing psychological literature and accommodates stochastic, (...) dynamic shifts of attention among visual feature dimensions. We perform a series of human experiments with two stimulus types and nine classification tasks to validate the model and to demonstrate the model's potential to boost performance. Our system achieves high accuracy for participants who are naive as to the classification task, even when the classification task switches from trial to trial. (shrink)

Mechanistic explanations satisfy widely held norms of explanation: the ability to manipulate and answer counterfactual questions about the explanandum phenomenon. A currently debated issue is whether any nonmechanistic explanations can satisfy these explanatory norms. Weiskopf argues that the models of object recognition and categorization, JIM, SUSTAIN, and ALCOVE, are not mechanistic yet satisfy these norms of explanation. In this article I argue that these models are mechanism sketches. My argument applies recent research using model-based functional magnetic resonance imaging, a novel (...) neuroimaging method whose significance for current debates on psychological models and mechanistic explanation has yet to be explored. (shrink)

Inductive generalization, where people go beyond the data provided, is a basic cognitive capability, and it underpins theoretical accounts of learning, categorization, and decision making. To complete the inductive leap needed for generalization, people must make a key ‘‘sampling’’ assumption about how the available data were generated. Previous models have considered two extreme possibilities, known as strong and weak sampling. In strong sampling, data are assumed to have been deliberately generated as positive examples of a concept, whereas in weak sampling, (...) data are assumed to have been generated without any restrictions. We develop a more general account of sampling that allows for an intermediate mixture of these two extremes, and we test its usefulness. In two experiments, we show that most people complete simple one-dimensional generalization tasks in a way that is consistent with their believing in some mixture of strong and weak sampling, but that there are large individual differences in the relative emphasis different people give to each type of sampling. We also show experimentally that the relative emphasis of the mixture is influenced by the structure of the available information. We discuss the psychological meaning of mixing strong and weak sampling, and possible extensions of our modeling approach to richer problems of inductive generalization. (shrink)

In his enormously influential The Modularity of Mind, Jerry Fodor (1983) proposed that the mind was divided into input modules and central processes. Much subsequent research focused on the modules and whether processes like speech perception or spatial vision are truly modular. Much less attention has been given to Fodor's writing on the central processes, what would today be called higher‐level cognition. In “Fodor's First Law of the Nonexistence of Cognitive Science,” he argued that central processes are “bad candidates for (...) scientific study” and would resist attempts at empirical analysis. This essay evaluates his argument for this remarkable claim, concluding that although central processes may well be “messier” than input modules, this does not mean that they cannot be studied and understood. The article briefly reviews the scientific progress made in understanding central processes in the 35 years since the book was published, showing that Fodor's prediction is clearly falsified by massive advances in topics like decision making and analogy. The essay concludes that Fodor's Law was not based on a clear argument for why the complexities of central systems could not be studied but was likely based on intuitions and preferences that were common in psychology at the time. (shrink)

Linguistic and non-linguistic pattern learning have been studied separately, but we argue for a comparative approach. Analogous inductive problems arise in phonological and visual pattern learning. Evidence from three experiments shows that human learners can solve them in analogous ways, and that human performance in both cases can be captured by the same models. We test GMECCS, an implementation of the Configural Cue Model in a Maximum Entropy phonotactic-learning framework with a single free parameter, against the alternative hypothesis that learners (...) seek featurally simple algebraic rules. We study the full typology of patterns introduced by Shepard, Hovland, and Jenkins (), instantiated as both phonotactic patterns and visual analogs, using unsupervised training. Unlike SHJ, Experiments 1 and 2 found that both phonotactic and visual patterns that depended on fewer features could be more difficult than those that depended on more features, as predicted by GMECCS but not by rule-seeking. GMECCS also correctly predicted performance differences between stimulus subclasses within each pattern. A third experiment tried supervised training to elicit simple rule-seeking phonotactic learning, but cue-based behavior persisted. We conclude that similar cue-based cognitive processes are available for phonological and visual concept learning, and hence that studying either kind of learning can lead to significant insights about the other. (shrink)

Symbols enable people to organize and communicate about the world. However, the ways in which symbolic knowledge is learned and then represented in the mind are poorly understood. We present a formal analysis of symbolic learning—in particular, word learning—in terms of prediction and cue competition, and we consider two possible ways in which symbols might be learned: by learning to predict a label from the features of objects and events in the world, and by learning to predict features from a (...) label. This analysis predicts significant differences in symbolic learning depending on the sequencing of objects and labels. We report a computational simulation and two human experiments that confirm these differences, revealing the existence of Feature-Label-Ordering effects in learning. Discrimination learning is facilitated when objects predict labels, but not when labels predict objects. Our results and analysis suggest that the semantic categories people use to understand and communicate about the world can only be learned if labels are predicted from objects. We discuss the implications of this for our understanding of the nature of language and symbolic thought, and in particular, for theories of reference. (shrink)

Rogers & McClelland's (R&M's) précis represents an important effort to address key issues in concepts and categorization, but few of the simulations deliver what is promised. We argue that the models are seriously underconstrained, importantly incomplete, and psychologically implausible; more broadly, R&M dwell too heavily on the apparent successes without comparable concern for limitations already noted in the literature.

Every scientist chooses a preferred level of analysis and this choice shapes the research program, even determining what counts as evidence. This contribution revisits Marr's three levels of analysis and evaluates the prospect of making progress at each individual level. After reviewing limitations of theorizing within a level, two strategies for integration across levels are considered. One is top–down in that it attempts to build a bridge from the computational to algorithmic level. Limitations of this approach include insufficient theoretical constraint (...) at the computation level to provide a foundation for integration, and that people are suboptimal for reasons other than capacity limitations. Instead, an inside-out approach is forwarded in which all three levels of analysis are integrated via the algorithmic level. This approach maximally leverages mutual data constraints at all levels. For example, algorithmic models can be used to interpret brain imaging data, and brain imaging data can be used to select among competing models. Examples of this approach to integration are provided. This merging of levels raises questions about the relevance of Marr's tripartite view. (shrink)

In his book Doing Without Concepts, Edouard Machery argues that cognitive scientists should reject the concept of “concept” as a natural, psychological kind. I review and critique several of Machery’s arguments, focusing on his definition of “concept” and on claims against the possibility and utility of a unified account of concepts. In particular, I suggest ways in which prototype, exemplar, and theory-theory approaches to concepts might be integrated.

Algorithms for approximate Bayesian inference, such as those based on sampling (i.e., Monte Carlo methods), provide a natural source of models of how people may deal with uncertainty with limited cognitive resources. Here, we consider the idea that individual differences in working memory capacity (WMC) may be usefully modeled in terms of the number of samples, or “particles,” available to perform inference. To test this idea, we focus on two recent experiments that report positive associations between WMC and two distinct (...) aspects of categorization performance: the ability to learn novel categories, and the ability to switch between different categorization strategies (“knowledge restructuring”). In favor of the idea of modeling WMC as a number of particles, we show that a single model can reproduce both experimental results by varying the number of particles—increasing the number of particles leads to both faster category learning and improved strategy‐switching. Furthermore, when we fit the model to individual participants, we found a positive association between WMC and best‐fit number of particles for strategy switching. However, no association between WMC and best‐fit number of particles was found for category learning. These results are discussed in the context of the general challenge of disentangling the contributions of different potential sources of behavioral variability. (shrink)

This article demonstrates the potential of using hierarchical Bayesian methods to relate models and data in the cognitive sciences. This is done using a worked example that considers an existing model of category representation, the Varying Abstraction Model (VAM), which attempts to infer the representations people use from their behavior in category learning tasks. The VAM allows for a wide variety of category representations to be inferred, but this article shows how a hierarchical Bayesian analysis can provide a unifying explanation (...) of the representational possibilities using 2 parameters. One parameter controls the emphasis on abstraction in category representations, and the other controls the emphasis on similarity. Using 30 previously published data sets, this work shows how inferences about these parameters, and about the category representations they generate, can be used to evaluate data in terms of the ongoing exemplar versus prototype and similarity versus rules debates in the literature. Using this concrete example, this article emphasizes the advantages of hierarchical Bayesian models in converting model selection problems to parameter estimation problems, and providing one way of specifying theoretically based priors for competing models. (shrink)

A fundamental question in the study of human cognition is how people learn to predict the category membership of an example from its properties. Leading approaches account for a wide range of data in terms of comparison to stored examples, abstractions capturing statistical regularities, or logical rules. Across three experiments, participants learned a category structure in a low‐dimension, continuous‐valued space consisting of regularly alternating regions of class membership (A B A B). The dependent measure was generalization performance for novel items (...) outside the range of the training space. Human learners often extended the alternation pattern––a finding of critical interest given that leading theories of categorization based on similarity or dimensional rules fail to predict this behavior. In addition, we provide novel theoretical interpretations of the observed phenomenon. (shrink)

Learning to understand a single causal system can be an achievement, but humans must learn about multiple causal systems over the course of a lifetime. We present a hierarchical Bayesian framework that helps to explain how learning about several causal systems can accelerate learning about systems that are subsequently encountered. Given experience with a set of objects, our framework learns a causal model for each object and a causal schema that captures commonalities among these causal models. The schema organizes the (...) objects into categories and specifies the causal powers and characteristic features of these categories and the characteristic causal interactions between categories. A schema of this kind allows causal models for subsequent objects to be rapidly learned, and we explore this accelerated learning in four experiments. Our results confirm that humans learn rapidly about the causal powers of novel objects, and we show that our framework accounts better for our data than alternative models of causal learning. (shrink)

Three experiments with 88 college-aged participants explored how unlabeled experiences—learning episodes in which people encounter objects without information about their category membership—influence beliefs about category structure. Participants performed a simple one-dimensional categorization task in a brief supervised learning phase, then made a large number of unsupervised categorization decisions about new items. In all three experiments, the unsupervised experience altered participants’ implicit and explicit mental category boundaries, their explicit beliefs about the most representative members of each category, and even their memory (...) for the items encountered during the supervised learning phase. These changes were influenced by both the range and frequency distribution of the unlabeled stimuli: mental category boundaries shifted toward the middle of the range and toward the trough of the bimodal distribution of unlabeled items, whereas beliefs about the most representative category members shifted toward the modes of the unlabeled distribution. One consequence of this shift in representations is a false-consensus effect (Experiment 3) where participants, despite receiving very disparate training experiences, show strong agreement in judgments about representativeness and boundary location following unsupervised category judgments. (shrink)

Research in education and cognitive development suggests that explaining plays a key role in learning and generalization: When learners provide explanations—even to themselves—they learn more effectively and generalize more readily to novel situations. This paper proposes and tests a subsumptive constraints account of this effect. Motivated by philosophical theories of explanation, this account predicts that explaining guides learners to interpret what they are learning in terms of unifying patterns or regularities, which promotes the discovery of broad generalizations. Three experiments provide (...) evidence for the subsumptive constraints account: prompting participants to explain while learning artificial categories promotes the induction of a broad generalization underlying category membership, relative to describing items (Exp. 1), thinking aloud (Exp. 2), or free study (Exp. 3). Although explaining facilitates discovery, Experiment 1 finds that description is more beneficial for learning item details. Experiment 2 additionally suggests that explaining anomalous observations may play a special role in belief revision. The findings provide insight into explanation’s role in discovery and generalization. (shrink)

Humans excel in categorization. Yet from a computational standpoint, learning a novel probabilistic classification task involves severe computational challenges. The present paper investigates one way to address these challenges: assuming class-conditional independence of features. This feature independence assumption simplifies the inference problem, allows for informed inferences about novel feature combinations, and performs robustly across different statistical environments. We designed a new Bayesian classification learning model that incorporates varying degrees of prior belief in class-conditional independence, learns whether or not independence holds, (...) and adapts its behavior accordingly. Theoretical results from two simulation studies demonstrate that classification behavior can appear to start simple, yet adapt effectively to unexpected task structures. Two experiments—designed using optimal experimental design principles—were conducted with human learners. Classification decisions of the majority of participants were best accounted for by a version of the model with very high initial prior belief in class-conditional independence, before adapting to the true environmental structure. Class-conditional independence may be a strong and useful default assumption in category learning tasks. (shrink)

Cognitive Science, Volume 46, Issue 2, February 2022.

Applying machine learning algorithms to automatically infer relationships between concepts from large-scale collections of documents presents a unique opportunity to investigate at scale how human semantic knowledge is organized, how people use it to make fundamental judgments (“How similar are cats and bears?”), and how these judgments depend on the features that describe concepts (e.g., size, furriness). However, efforts to date have exhibited a substantial discrepancy between algorithm predictions and human empirical judgments. Here, we introduce a novel approach to generating (...) embeddings for this purpose motivated by the idea that semantic context plays a critical role in human judgment. We leverage this idea by constraining the topic or domain from which documents used for generating embeddings are drawn (e.g., referring to the natural world vs. transportation apparatus). Specifically, we trained state-of-the-art machine learning algorithms using contextually-constrained text corpora (domain-specific subsets of Wikipedia articles, 50+ million words each) and showed that this procedure greatly improved predictions of empirical similarity judgments and feature ratings of contextually relevant concepts. Furthermore, we describe a novel, computationally tractable method for improving predictions of contextually-unconstrained embedding models based on dimensionality reduction of their internal representation to a small number of contextually relevant semantic features. By improving the correspondence between predictions derived automatically by machine learning methods using vast amounts of data and more limited, but direct empirical measurements of human judgments, our approach may help leverage the availability of online corpora to better understand the structure of human semantic representations and how people make judgments based on those. (shrink)

Humans and many other species selectively attend to stimuli or stimulus dimensions—but why should an animal constrain information input in this way? To investigate the adaptive functions of attention, we used a genetic algorithm to evolve simple connectionist networks that had to make categorization decisions in a variety of environmental structures. The results of these simulations show that while learned attention is not universally adaptive, its benefit is not restricted to the reduction of input complexity in order to keep it (...) within an organism's processing capacity limitations. Instead, being able to shift attention provides adaptive benefit by allowing faster learning with fewer errors in a range of ecologically plausible environments. (shrink)

We critically review key lines of evidence and theoretical argument relevant to Machery's These include interactions between different kinds of concept representations, unified approaches to explaining contextual effects on concept retrieval, and a critique of empirical dissociations as evidence for concept heterogeneity. We suggest there are good grounds for retaining the concept construct in human cognition.