Interacting with volatile environments stabilizes hidden-state inference and its brain signatures.

Area of Science:

• Cognitive neuroscience
• Decision-making under uncertainty
• Bayesian inference

Background:

• Accurate decisions in uncertain environments require understanding cue origins and action outcomes.
• Both processes are Bayesian inference but studied separately, hindering direct comparison.
• Existing research often uses distinct experimental setups for cue-based vs. outcome-based inference.

Purpose of the Study:

• To formally compare cue-based and outcome-based inference within a unified framework.
• To investigate how the control over environmental sampling influences learning and decision-making under uncertainty.
• To explore the neural underpinnings of these distinct inference strategies.

Main Methods:

• Framed a reversal learning task as either cue-based or outcome-based inference.
• Measured human behavior and magnetoencephalographic (MEG) activity.
• Analyzed multivariate patterns of neural activity related to belief updating and evidence integration.

Main Results:

• Humans perceived the same volatile environment as more stable when inferring via uncertain outcomes compared to uncertain cues.
• Neural activity patterns in associative temporal lobe regions reflected this behavioral difference.
• The interaction between inferred beliefs and incoming evidence differed based on the inference strategy.

Conclusions:

• The degree of control over environmental sampling significantly shapes learning and decision-making under uncertainty.
• Perception of environmental stability is modulated by whether inference is driven by observing cues or interacting with outcomes.
• Neural mechanisms in associative regions differentiate learning strategies based on inference type.