Approximating Bayes? On the role of approximations in Bayesian cognitive science

2025/11/24 by Arnon Levy · 1 voice
Arts and Humanities · Computer Science · Neuroscience · #Bayesian Modeling and Causal Inference #Embodied and Extended Cognition #Philosophy and History of Science

paper · doi:10.1080/09515089.2025.2584488

openalex publication_date 2025/11/24 · openalex created_date 2025/11/25 · openalex updated_date 2026/06/15

Abstract

Approximations have come to occupy a central role within Bayesian cognitive science. Many cognitive scientists view current approximation-based models as showing that the mind approximates Bayesian inference. In this paper I interrogate this idea, asking what it means to claim that the mind approximates one computation by executing another. I look at three possible interpretations of such claims, finding problems with each. I argue that this poses challenges to one central rationale for the Bayesian approach, and to its top-down methodology.

Citations

• Bayesian Intractability Is Not an Ailment That Approximation Can Cure
• Human-level concept learning through probabilistic program induction
• Resource-rational analysis: Understanding human cognition as the optimal use of limited computational resources

Discussions

• Latest papers: Arnon Levy interrogates the idea that the mind approximates Bayesian inference, asking what is means to claim that the mind approximates one computation by executing another doi.org/10. [bsky, 10 points, 1 comments]

Related

• Bayesian Intractability Is Not an Ailment That Approximation Can Cure
• Bayes in the age of intelligent machines
• Bayesian Analogical Cybernetics
• The "Bayesian" brain, with a bit less Bayes
• Generalized Posteriors in Approximate Bayesian Computation