Monte Carlo Planning Method Estimates Planning Horizons during Interactive Social Exchange.

Area of Science:

• Computational neuroscience
• Behavioral economics
• Social psychology

Background:

• Human interactions involve complex reciprocal exchanges, such as repeated trust tasks.
• Modeling these interactions requires accounting for preferences, beliefs, and future planning.
• Existing frameworks like interactive partially observable Markov decision processes (IPOMDPs) are computationally challenging.

Purpose of the Study:

• To develop a computationally precise method for analyzing behavior in multi-round trust tasks.
• To approximate solutions for complex interactive decision-making.
• To enable the inversion of observed behavioral choices.

Main Methods:

• Utilized a variant of the Monte-Carlo tree search algorithm.
• Applied the algorithm to approximate solutions for interactive partially observable Markov decision processes (IPOMDPs).
• Generated behavior to analyze interactive inference.

Main Results:

• Demonstrated the efficiency and effectiveness of the Monte-Carlo tree search approach.
• Showcased the algorithm's capability to invert observed behavioral choices.
• Elucidated the richness and sophistication of interactive inference through generated behavior.

Conclusions:

• The Monte-Carlo tree search offers an efficient and effective approximation for complex interactive decision-making.
• This method advances the computational understanding of human reciprocal exchanges.
• The approach facilitates the analysis of sophisticated interactive inference in behavioral tasks.