The Naïve Utility Calculus as a unified, quantitative framework for action understanding

Area of Science:

• Cognitive Science
• Social Psychology
• Computational Neuroscience

Background:

• Human social navigation relies on understanding others' behavior.
• Previous research suggests agents act to maximize subjective utility.
• A formal framework is needed to explain action-understanding.

Purpose of the Study:

• To present a formal theory of Naïve Utility Calculus.
• To model action-understanding using a Bayesian framework.
• To explain how cost and reward tradeoffs influence behavior.

Main Methods:

• Developed a probabilistic generative model for Naïve Utility Calculus.
• Used a Bayesian framework to analyze cost and reward tradeoffs.
• Tested the model's quantitative accuracy in predicting inferences from actions.

Main Results:

• The model accurately predicts how people infer agents' subjective costs and rewards.
• The model extends to complex, multi-episode, multi-agent scenarios.
• Distinguishing desires, goals, and intentions enhances model capabilities.

Conclusions:

• Naïve Utility Calculus provides a formal account of action-understanding.
• The model illuminates basic representations for social navigation.
• This framework advances our understanding of human social cognition.