Many category learning experiments use supervised learning (i.e., trial-by-trial feedback). Most of those procedures use deterministic feedback, teaching participants to classify exemplars into consistent categories (i.e., the stimulus i is always classified in category k). Though some researchers suggest that natural learning conditions are more likely to be inconsistent, the literature using probabilistic feedback in category learning experiments is sparse. Our analysis of the literature suggests that part of the reason for this sparsity is a relative lack of flexibility of current paradigms and procedures for designing probabilistic feedback experiments. The work we report here offers a novel paradigm (the Probabilistic Prototype Distortion task) which allows researchers greater flexibility when creating experiments with different p(category|feature) probabilities, and also allows parametrically manipulating the amount of randomness in an experimental task. In the current work, we offer a detailed procedure, implementation, experimental results and discussion of this novel procedure. Our results suggest that by designing experiments with our procedures, the experimental setup allows subjects to achieve the desired classification performance.

Over the last 20 years, causal-model theory has produced muchknowledge about causal-based categorization. However, per-sistent violations to the normative causal-model theory areprevalent. In particular, violations to the Markov conditionhave been repeatedly found. These violations have receiveddifferent explanations. Here, we develop a model that startsfrom generally accepted cognitive phenomena (e.g., process-ing limitations, the relevance of inference in cognitive process-ing) and assumes that people are not fully causal nor fully asso-ciative when performing causal-based categorization, offeringa new explanation for Markov violations.

Causal-based cognition is thought to be relevant for human beings because it allows inferring the unfolding of events. Theories of causal-based cognition offer researchers a way to understand inter-feature relations, above and beyond the purely associative relations posited by similarity theories. In the causal-model theory (a.k.a. the Generative Model), people are thought to categorize an exemplar depending on how likely its particular feature combination is, given the category’s causal model. This mechanism predicts the coherence effect (i.e., when people categorize, features interact). This effect has been widely reported in the literature. In the current experiment, we sought to specify conditions that modulate the coherence effect. To that end, we implemented a between-subjects manipulation where participants had to judge either category membership or category consistency. Our results show that subjects exhibit a larger coherence effect in consistency condition. We discuss our results’ relevance for causal-model theory and for the possibility of distinguishing causal-based from similarity-based processing.

An agent-based model for semantic search and retrieval in memory is proposed. The model seeks to generate verbal fluency lists with properties similar to those generated by humans in the semantic fluency task. This model is compared to a random walk in a semantic network in its ability to adjust to the results of 141 undergraduate students in the semantic fluency task in eight different outcomes. We found that the agent-based model fits participants' results better than the random walk model. The results were consistent with optimal foraging theories, and the distributions of the total number of words, similarities, and frequency values were similar to those generated by participants. The potential uses of this model as a virtual environment to experiment with the search and retrieval process in semantic memory are discussed.

Humans excel at causal reasoning, yet at the same time consistently fail to respect its basic axioms. They seemingly fail to recognize, for instance, that only the direct causes of an event can affect its probability (the Markov condition). How can one explain this paradox? Here we argue that standard normative analyses of causal reasoning mostly apply to the idealized case where the reasoner has perfect confidence in her knowledge of the underlying causal model. Given uncertainty about the correct representation of a causal system, it is not always rational for a reasoner to respect the Markov condition and other ‘normative’ principles. To test whether uncertainty can account for the apparent fallibility of human judgments, we formulate a simple computational model of a rational-but-uncertain causal reasoner. In a re-analysis of a recent causal reasoning study, the model fits the data significantly better than its standard normative counterpart.

This work provides an alternative account for deviations in human causal reasoning from normative predictions based on Causal Bayesian Networks (CBNs). We highlight violations of the Markov condition (Screening Off) and insufficient Explaining Away. Different from other accounts, our model does not assume that people fail to honor normative predictions due to reliance on heuristics, hidden nodes and links or cognitive limitations. Instead, we propose that people are rationally uncertain about the received causal model they are asked to reason with. We fitted the model to published data from two experiments where people were asked to make probability estimates on inferences of interest within a causal model. We find that the model is able to i) reproduce deviations from normative predictions, and ii) predict changes in the magnitude of these deviations across contexts. We conclude that assuming that people, in order to be rational, will always fully believe in the information they receive about a causal model may be too strong an assumption.

Over the last few decades, Causal Model Theory (CMT) has become a dominant framework for human causal-based reasoning, including categorization and inference. CMT prescribes how people should reason about probabilistic events in terms of causal models. In typical causal-based categorization experiments, subjects are provided with verbal descriptions of causally linked features, generally including probabilistic information. Another line of research focuses on perceived or experienced causal events, rather than on verbal descriptions. In this work we asked whether effects which are consistent with CMT, and that have been obtained with verbal descriptions, generalize to visually perceived events. In two experiments, we presented subjects with videos of a 3D A→B causal event rather than verbal descriptions. In Exp. 1, we found that subjects who saw the causal event did not show the coherence effect in categorization (i.e., subjects tend to rate the null ¬A¬B event as a category member). However, subjects who did see the null event during training did show the effect. In Exp. 2, we ruled out the possibility that Exp. 1’s results were simply an effect of how frequently events were experienced during training. We conclude that a one-shot perceived causal event is not sufficient for people to show causal-based reasoning as CMT predicts.

Probabilistic categorization (PC) has been used mostly to distinguish between memory systems (declarative vs. procedural). Most literature on PC has relied on the Weather Prediction Task. However, this task doesn't provide the flexibility in assigning probabilities to exemplar features that is often required in more ecological settings. Recently, Marchant and Chaigneau (2021) developed a PC task that allows flexible classification probabilities by computing p(category|feature). In this study, we implemented Marchant and Chaigneau's PC task under two feedback conditions (i.e., 70% and 90%) counterbalanced by three features' relevance conditions. In the transfer phase, subjects rated exemplars' category membership. During learning, we implemented eye-tracking to capture the number of fixations to each exemplar's features. Our work in progress shows that fixations on relevant features predict transfer responses, suggesting that people show declarative knowledge of critical features according to their relevance. Interestingly, declarative retrieval varies with the reliability of feedback.