Generalized Associative Model: A model for individual and averaged learning curves in Pavlovian conditioning

Abstract

The gap between individual curves and their averaged curve has been discussed for a long time in Pavlovian conditioning literature. To solve this problem, this study proposes a new mathematical model called the Generalized Associative Model (GAM). The GAM integrates the Weibull model with an interpretation of the traditional associative model. This interprets the Weibull model's shape parameter as a parameter representing individual differences in individual curves. Numerical simulations replicated the individual and average curves reported in previous studies. Analysis fitting the GAM to the four fear conditioning datasets further supported the validity of the GAM. The GAM has implications in both basic and clinical research.

Multiple Propositional Theory for Understanding Rule and Belief Updating: An Integrative Perspective to Improve Clinical Practice

“…Second, to represent the satisficing principle, individual differences in learning, such as Weibull curves (Gallistel et al, 2004), need to be encompassed. This has been well documented in the recently proposed GAM (Nihei et al, 2023). Third, the semantic similarities between propositions need to be considered.…”

“…Early theories have suggested that cognitive changes underpin the occurrence of sudden gain (Tang & DeRubeis, 1999). Interestingly, the increase in beliefs due to repeated presentations can occur rapidly at certain times, similar to changes in US-expectancy (Nihei et al, 2023). Verifying the propositions and Multiple Propositional Theory the satisficing principle may be involved in this process.…”

“…The population-level means for the process of belief increase show a RW Multiple Propositional Theory model-like curve (i.e., a striking increase is observed in the first trial with a large prediction error; in subsequent trials, the increase in belief decreases as the prediction error decreases). Nevertheless, in the trajectory of individual-level belief updating, some participants who do not follow this rule show a large increase in belief in trials after their second trial (Nihei et al, 2023 June 5). Of particular importance is the reason why a striking increase in belief is observed at a certain point, rather than in the first trial.…”

“…In summary, based on learning, cognitive science, computation, and social psychology, Bayesian and belief-updating models with a core assumption of prediction error have been imported into psychopathology (Jin et al, 2023;Kube & Rosenkrantz, 2021;Moscovitch et al 2023). However, several phenomena relevant to belief and rule updating are not explained by these types of models: rapid changes in US expectancy after the second trial (Nihei et al, 2023 June 5), backward blocking (Shanks, 1985), belief invariance due to strong prediction errors (Hird et al, 2019;, belief increases due to repeated exposure to belief information (Fazio et al, 2022), and probabilistic judgments in reasoning (Oaksford & Chater, 2020). For example, used the same procedure as the experiment on emotional intelligence described earlier (Kube et al, 2019), with three feedback conditions on performance: top 50%, top 10%, and top 1%.…”

“…Here, we propose that people repeatedly test hypotheses deduced from propositions, and at a certain moment when they are satisfied with a hypothesis, the confidence of the proposition increases rapidly. This principle combines the satisficing principle of hypothetical thinking , the "Aha" experience in insight problem solving, and the generalized associative model (GAM) or modified Weibull model of learning (Gallistel et al, 2004;Nihei et al, 2023 June 5). The satisficing principle states that once people find a sufficient hypothesis (model) in light of their current goal, they tend to terminate their search for better models .…”