Dissociable Effects of Curiosity and Hedonic Valence on Reinforcement Learning

Area of Science:

• Neuroscience
• Behavioral Economics
• Cognitive Science

Background:

• Curiosity and exploration are crucial for learning and decision-making under uncertainty.
• The impact of outcome valence (positive vs. negative) on exploration strategies across species is not fully understood.
• Previous human studies often rely on explicit instructions, limiting insight into experience-driven exploration.

Purpose of the Study:

• To investigate how hedonic valence influences novelty seeking, exploration, and reinforcement learning in rhesus macaques.
• To determine if valence-dependent exploration arises from curiosity or prior beliefs.
• To examine the roles of approach and avoidance motivation in learning and decision-making.

Main Methods:

• Rhesus macaques were studied using visual tokens as secondary reinforcers.
• Exploration of novel, uncertain options was compared between conditions predicting gains and losses.
• Reinforcement learning, choice balks, and directed vs. random exploration were analyzed.

Main Results:

• Monkeys showed increased novelty seeking when outcomes involved potential losses compared to gains.
• Heightened novelty seeking was attributed to optimistic prior beliefs about novelty, not a direct valence effect on curiosity.
• Monkeys learned faster from losses than gains, indicating loss aversion.
• Choice balks served as strategic responses to approach-avoidance conflicts and uncertainty, modulating exploration.

Conclusions:

• Hedonic valence does not categorically alter curiosity but influences exploration through prior beliefs and motivational states.
• Approach and avoidance motivations differentially impact reinforcement learning and decision-making strategies.
• The findings reveal dissociable effects of curiosity and valence on explore-exploit tradeoffs and adaptive behavior.