A decision-theoretic model of multistability: perceptual switches as internal actions DOI Creative Commons
Shervin Safavi, Peter Dayan

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 11, 2024

Abstract Perceptual multistability has been studied for centuries using a diverse collection of approaches. Insights derived from this phenomenon range core principles information processing, such as perceptual inference, to high-level concerns, visual awareness. The dominant computational explanations are based on the Helmholtzian view perception inverse inference. However, these approaches struggle account crucial role played by value, e.g., with percepts paired reward dominating longer periods than unpaired ones. In study, we formulate in terms dynamic, value-based, choice, employing formalism partially observable Markov decision process (POMDP). We use binocular rivalry an example, considering different explicit and implicit sources (and punishment) each percept. resulting values time-dependent influenced novelty form exploration. solution POMDP is optimal policy, show that can replicate explain several characteristics rivalry, ranging classic hallmarks apparently spontaneous random switches approximately gamma-distributed dominance more subtle aspects rich temporal dynamics switching rates. Overall, our decision-theoretic perspective not only accounts wealth unexplained data, but also opens up modern conceptions internal reinforcement learning service understanding phenomena, sensory processing generally.

Language: Английский

Bayesian p-curve mixture models as a tool to dissociate effect size and effect prevalence DOI Creative Commons
John P. Veillette, Howard C. Nusbaum

Communications Psychology, Journal Year: 2025, Volume and Issue: 3(1)

Published: Jan. 23, 2025

Much research in the behavioral sciences aims to characterize "typical" person. A statistically significant group-averaged effect size is often interpreted as evidence that typical person shows an effect, but only true under certain distributional assumptions for which explicit rarely presented. Mean varies with both within-participant and population prevalence (proportion of showing effect). Few studies consider how affects mean estimates existing estimators are, conversely, confounded by uncertainty about size. We introduce a widely applicable Bayesian method, p-curve mixture model, jointly probabilistically clustering participant-level data based on their likelihood null distribution. Our approach, we provide software tool, outperforms estimation methods when uncertain sensitive differences or across groups conditions. Statistically group-level effects are misinterpreted imply population. This Resource provides method tool infer experimental directly.

Language: Английский

Citations

0
What the Average Really Means: Dissociating Effect Size and Effect Prevalence usingp-curve Mixtures DOI Creative Commons
John P. Veillette, Howard C. Nusbaum

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Aug. 1, 2024

Abstract Much research in the behavioral sciences aims to characterize “typical” person. A statistically significant group-averaged effect size is often interpreted as evidence that typical person shows an effect, but only true under certain distributional assumptions for which explicit rarely presented. Mean varies with both within-participant and population prevalence (proportion of showing effect). Few studies consider how affects mean estimates existing estimators are, conversely, confounded by uncertainty about size. We introduce a widely applicable Bayesian method, p -curve mixture model, jointly probabilistically clustering participant-level data based on their likelihood null distribution. Our approach, we provide software tool, outperforms estimation methods when uncertain sensitive differences or across groups conditions.

Language: Английский

Citations

0
A decision-theoretic model of multistability: perceptual switches as internal actions DOI Creative Commons
Shervin Safavi, Peter Dayan

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 11, 2024

Abstract Perceptual multistability has been studied for centuries using a diverse collection of approaches. Insights derived from this phenomenon range core principles information processing, such as perceptual inference, to high-level concerns, visual awareness. The dominant computational explanations are based on the Helmholtzian view perception inverse inference. However, these approaches struggle account crucial role played by value, e.g., with percepts paired reward dominating longer periods than unpaired ones. In study, we formulate in terms dynamic, value-based, choice, employing formalism partially observable Markov decision process (POMDP). We use binocular rivalry an example, considering different explicit and implicit sources (and punishment) each percept. resulting values time-dependent influenced novelty form exploration. solution POMDP is optimal policy, show that can replicate explain several characteristics rivalry, ranging classic hallmarks apparently spontaneous random switches approximately gamma-distributed dominance more subtle aspects rich temporal dynamics switching rates. Overall, our decision-theoretic perspective not only accounts wealth unexplained data, but also opens up modern conceptions internal reinforcement learning service understanding phenomena, sensory processing generally.

Language: Английский

Citations

0