Adaptive Integration of Perceptual and Reward Information in an Uncertain World

Published: Nov. 7, 2024

Perceptual uncertainty and salience both impact decision-making, but how these factors precisely trial-and-error reinforcement learning is not well understood. Here, we test the hypotheses that (H1) perceptual modulates reward-based (H2) economic decision-making driven by value of sensory information. For this, combined computational modeling with a uncertainty-augmented reward-learning task in human behavioral experiment ( N = 98). In line our hypotheses, found subjects regulated behavior response to which they could distinguish choice options based on information (belief state), addition errors made predicting outcomes. Moreover, considered combination expected values for decision-making. Taken together, this shows are closely intertwined share common basis real world.

Language: Английский

Approximate planning in spatial search

PLoS Computational Biology, Journal Year: 2024, Volume and Issue: 20(11), P. e1012582 - e1012582

Published: Nov. 12, 2024

How people plan is an active area of research in cognitive science, neuroscience, and artificial intelligence. However, tasks traditionally used to study planning the laboratory tend be constrained environments, such as Chess bandit problems. To date there still no agreed-on model how realistic contexts, navigation search, where values intuitively derive from interactions between perception cognition. address this gap move towards a more naturalistic planning, we present novel spatial Maze Search Task (MST) costs rewards are physically situated distances locations. We task two behavioral experiments evaluate contrast multiple distinct computational models including optimal expected utility several one-step heuristics inspired by studies information family planners that deviate which action estimated found people’s deviations best explained with limited horizon, however our results do not exclude possibility human may also affected mechanisms numerosity probability perception. This result makes theoretical contribution showing horizon generalizes demonstrates value multi-model approach for understanding

Language: Английский

Increasing the Construct Validity of Computational Phenotypes of Mental Illness Through Active Inference and Brain Imaging

Brain Sciences, Journal Year: 2024, Volume and Issue: 14(12), P. 1278 - 1278

Published: Dec. 19, 2024

After more than 30 years since its inception, the utility of brain imaging for understanding and diagnosing mental illnesses is in doubt, receiving well-grounded criticisms from clinical practitioners. Symptom-based correlational approaches have struggled to provide psychiatry with reliable brain-imaging metrics. However, emergence computational has paved a new path not only psychopathology illness but also practical tools practice terms metrics, specifically phenotypes. these phenotypes still lack sufficient test–retest reliability. In this review, we describe recent works revealing that mind brain-related show structural (not random) variation over time, longitudinal changes. Furthermore, findings suggest causes changes will improve construct validity an ensuing increase We propose active inference framework offers general-purpose approach causally by incorporating as observations within partially observable Markov decision processes.

Language: Английский