Can model-free reinforcement learning explain deontological moral judgments?

Area of Science:

• Cognitive Science
• Moral Psychology
• Neuroscience

Background:

• Dual-systems frameworks suggest moral judgments stem from emotional and rational processes.
• Cushman's (2013) dual-system theory links moral judgment to model-free and model-based reinforcement learning.
• Model-free learning associates actions with rewards/punishments, potentially explaining deontological judgments.
• Model-based learning involves causal modeling for planning, aligning with utilitarianism.

Purpose of the Study:

• To critically evaluate the application of model-free reinforcement learning in explaining deontological moral judgment.
• To identify limitations and areas for further investigation in the proposed dual-system framework.

Main Methods:

• Theoretical analysis and critique of the model-free reinforcement learning account of deontological moral judgment.
• Examination of existing evidence and predictions related to reinforcement learning and moral decision-making.

Main Results:

• Model-free learning alone is insufficient to explain deontological moral judgments, as many aversive actions aren't deemed morally wrong.
• Empirical evidence supporting key predictions of the reinforcement account (e.g., differing judgments based on reinforcement history) is lacking.
• Incorporating intention into the model-free framework necessitates unsupported assumptions.

Conclusions:

• Model-free reinforcement learning requires augmentation to adequately account for deontological moral judgments.
• Future research should address the identified challenges concerning aversion, empirical evidence, and the role of intention.
• The model-free/model-based framework offers a valuable starting point but requires refinement for a comprehensive understanding of moral cognition.