Gaze data reveal distinct choice processes underlying model-based and model-free reinforcement learning.

Area of Science:

• Cognitive Science
• Neuroscience
• Decision Science

Background:

• Organisms utilize distinct learning strategies: model-free (reinforcement-based) and model-based (forward-looking evaluation).
• Previous research suggested both strategies involve value comparison at trial onset, with model-based learning emphasizing future computations.

Purpose of the Study:

• To investigate the temporal dynamics of decision-making in model-free versus model-based learning.
• To challenge the assumption that decisions are always made at a prescribed trial onset.

Main Methods:

• Utilized eye-tracking technology to monitor subject behavior.
• Analyzed decision-making processes in participants exhibiting model-free and model-based learning strategies.

Main Results:

• Model-based learners were observed to make choices *prior* to the official trial onset.
• Model-free learners engaged in a simpler value comparison process during the trial, aligning with sequential-sampling models.
• Findings suggest distinct temporal patterns for different learning strategies.

Conclusions:

• The timing of decision-making differs significantly between model-based and model-free learning.
• Assumptions about synchronized decision-making across learning types in experimental settings may be flawed.
• Eye-tracking provides crucial insights into the cognitive processes underlying learning and decision-making.