Model-free and model-based learning processes in the updating of explicit and implicit evaluations

Area of Science:

• Cognitive Psychology
• Neuroscience
• Decision Science

Background:

• Human evaluation involves explicit (self-reported) and implicit (indirectly measured) processes.
• Existing research shows dissociations and associations between explicit and implicit evaluations.
• The underlying mechanisms driving these differences remain unclear, with debates on superficial measurement versus deeper processing distinctions.

Purpose of the Study:

• To investigate the unique and shared aspects of explicit and implicit evaluations.
• To leverage the distinction between model-based and model-free reinforcement learning (RL) to understand evaluation acquisition and representation.
• To provide a novel framework for understanding the context-sensitivity and recalcitrance of implicit evaluations.

Main Methods:

• Utilized a revaluation procedure across three studies with a total of 2,354 participants.
• Employed principles from model-based and model-free reinforcement learning to analyze evaluation updates.
• Varied stimulus exposure and reinforcement schedules (deterministic vs. probabilistic) to test robustness.

Main Results:

• Explicit evaluations are updated by both model-free and model-based RL processes.
• Implicit evaluations primarily depend on model-free RL and are unaffected by model-based RL.
• These findings were robust across different numbers of stimulus exposures and reinforcement types.

Conclusions:

• Implicit evaluations are context-sensitive and exhibit long-term recalcitrance, differing significantly from explicit evaluations.
• The reinforcement learning framework offers a more nuanced understanding beyond traditional dual-process or single-process models.
• Results suggest potential for theoretically guided interventions to modify implicit evaluations.