The contribution of striatal pseudo-reward prediction errors to value-based decision-making

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Neuroscience

Background:

• Traditional learning studies focus on simple stimulus-response associations.
• Real-world learning involves complex action sequences, potentially managed by hierarchical reinforcement learning (HRL).
• HRL decomposes tasks into subgoals, using pseudo-reward prediction errors (PRPEs) for intermediate reinforcement.

Purpose of the Study:

• To investigate if pseudo-rewards bias choice behavior without inherent value.
• To determine if this bias correlates with striatal representation strength of PRPEs.
• To explore the neural mechanisms of subgoal learning in HRL.

Main Methods:

• Developed a novel decision-making paradigm for two studies (fMRI, behavioral).
• Assessed reward prediction errors (RPEs) and PRPEs.
• Utilized fMRI to analyze striatal activity and its relation to choice behavior.

Main Results:

• Participants showed a preference for pseudo-rewarding options, irrespective of actual monetary gain.
• Individual differences in striatal sensitivity to PRPEs versus RPEs predicted this preference.
• The findings demonstrate that PRPEs influence choice behavior.

Conclusions:

• Pseudo-rewards generate learning signals within the striatum.
• These signals can bias decision-making, even when disconnected from primary rewards.
• This supports the role of HRL in simplifying complex learning through subgoal attainment.