Human Belief State-Based Exploration and Exploitation in an Information-Selective Symmetric Reversal Bandit Task.

Area of Science:

• Cognitive Science
• Neuroscience
• Behavioral Economics

Background:

• Sequential decision-making challenges arise when reward information is incomplete.
• Previous research highlights the exploration-exploitation dilemma in human choice behavior.

Purpose of the Study:

• To model and analyze human decision-making in information-selective scenarios.
• To investigate the strategies employed in sequential decision-making tasks with detached reward information.

Main Methods:

• Introduction of an information-selective symmetric reversal bandit task.
• Development and validation of probabilistic agent-based behavioral models.
• Analysis of choice data from 24 participants using a maximum likelihood approach.

Main Results:

• Participants demonstrated a belief state-based hybrid explorative-exploitative strategy.
• Quantitative evidence supports the role of subjective uncertainty in decision-making.
• Model validation confirmed robust parameter recovery.

Conclusions:

• Human decision-making in complex environments is a blend of exploration and exploitation.
• Subjective uncertainty is a key driver in resolving exploration-exploitation trade-offs.
• The developed bandit task effectively models real-world sequential decision problems.