The influence of reward magnitude on stimulus memory and stimulus generalization in categorization decisions.

Abstract

Reward magnitude is a central concept in most theories of preferential decision making and learning. However, it is unknown whether variable rewards also influence cognitive processes when learning how to make accurate decisions (e.g., sorting healthy and unhealthy food differing in appeal). To test this, we conducted 3 studies. Participants learned to classify objects with 3 feature dimensions into two categories before solving a transfer task with novel objects. During learning, we rewarded all correct decisions, but specific category exemplars yielded a 10 times higher reward (high vs. low). Counterintuitively, categorization performance did not increase for high-reward stimuli, compared with an equal-reward baseline condition. Instead, performance decreased reliably for low-reward stimuli. To analyze the influence of reward magnitude on category generalization, we implemented an exemplar-categorization model and a cue-weighting model using a Bayesian modeling approach. We tested whether reward magnitude affects (a) the availability of exemplars in memory, (b) their psychological similarity to the stimulus, or (c) attention to stimulus features. In all studies, the evidence favored the hypothesis that reward magnitude affects the similarity gradients of high-reward exemplars compared with the equal-reward baseline. The results from additional reward-judgment tasks (Studies 2 and 3) strongly suggest that the cognitive processes of reward-value generalization parallel those of category generalization. Overall, the studies provide insights highlighting the need for integrating reward-and category-learning theories. AbstractReward magnitude is a central concept in most theories of preferential decision making and learning. However, it is unknown whether variable rewards also influence cognitive processes when learning how to make accurate decisions (e.g., sorting healthy and unhealthy food differing in appeal). To test this, we conducted three studies.Participants learned to classify objects with three feature dimensions into two categories before solving a transfer task with novel objects. During learning, we rewarded correct decisions, but specific category exemplars yielded a 10 times higher reward (high vs.low). Counterintuitively, categorization performance did not increase for high-reward stimuli, compared to an equal-reward baseline condition. Instead, performance decreased reliably for low-reward stimuli. To analyze the influence of reward magnitude on category generalization, we implemented an exemplar-categorization model and a cue-weighting model using a Bayesian modeling approach. We tested whether reward magnitude affects (a) the availability of exemplars in memory, (b) their psychological similarity to the stimulus, or (c) attention to stimulus features. In all studies, the evidence favored the hypothesis that reward magnitude affects the similarity gradients of high-reward exemplars compared to the equal-reward baseline. The results from additional reward-judgment tasks (Studies 2 and 3) strongly su...

A cognitive category-learning model of rule abstraction, attention learning, and contextual modulation.

“…This theoretically commits CAL to the idea that abstraction is mainly driven by the rule-learning network, and strong memorization is more akin to stimulus identification. In other words, in exemplar models (e.g., GCM; Nosofsky, 1986), if the memory strength parameter of an exemplar becomes stronger, an increase in its recall accuracy is predicted, while a decrease in accuracy for exemplars from other categories is also predicted (see also Hendrickson et al, 2019; Homa et al, 2019; Schlegelmilch & von Helversen, 2020), similar to a recall bias. In CAL, increasing the memory strength of a stored instance increases its recall accuracy and decreases its interfering influence on category inferences for dissimilar instances.…”

“…CATEGORY ABSTRACTION LEARNING 29 in its recall accuracy is predicted, while a decrease in accuracy for exemplars from is also predicted (see also Hendrickson et al, Homa et al, 2019;Schlegelmilch & von Helversen, 2020), similar to a In CAL, increasing the memory strength of a stored instance increases its recall accuracy and decreases its interfering influence on category inferences for dissimilar instances.…”

A Cognitive Category-Learning Model of Rule Abstraction, Attention Learning, and Contextual Modulation

“…However, the formal concepts of exemplar theory define memory strength as the degree by which each exemplar is integrated into overall similarity. In this sense, if the memory strength of an exemplar becomes stronger, an increase in its recall accuracy is predicted, while a decrease in accuracy of recalling exemplars from other categories is predicted, as well (see also Hendrickson et al, 2019;Homa et al, 2019;Schlegelmilch & von Helversen, 2020), similar to a recall bias. In the current version of CAL the increase in memory strength means an increase in the association between a stored instance and a category label, which does not affect how likely the retrieval of that instance or other instances is.…”

Measuring the mixture of rule-based and exemplar-based processes in judgment: A hierarchical Bayesian approach.

“…In this article, we introduce a hierarchical Bayesian version of the RulEx-J model since the hierarchical Bayesian modeling framework offers many advantages and has therefore become a very popular tool for estimating latent parameters of cognitive models (e.g., Bott et al, 2020; Mattes et al, 2020; Schlegelmilch & von Helversen, 2020; Schubert et al, 2019; for general introductions, see Lee, 2018; McElreath, 2020; Rouder et al, 2018). For instance, the hierarchical structure of the model naturally reflects the hierarchical data structure of many experiments, where several participants perform multiple trials of the same task and it is the aim of the researcher to draw conclusions on the group level (e.g., Steingroever et al, 2018).…”