Experimental Results on the Centipede Game in Normal Form: An Investigation on Learning

Area of Science:

• Behavioral Economics
• Game Theory
• Cognitive Science

Background:

• The centipede game is a classic economic experiment modeling strategic decision-making over time.
• Understanding player behavior in repeated games is crucial for developing accurate predictive models.

Purpose of the Study:

• To analyze player behavior in a repeated centipede game.
• To compare the efficacy of various static and learning models in explaining observed behavior.
• To investigate the impact of sequential decision-making on game outcomes.

Main Methods:

• Experimentation using the centipede game in reduced normal form.
• Subjects played 100 rounds against randomly matched opponents.
• Comparison of static models, quantal response, reinforcement models, and fictitious play.
• Application of learning direction theory to structure period-to-period behavior.

Main Results:

• Significant differences in player behavior were observed based on the timing of their decision relative to the opponent.
• Learning direction theory provided a framework for understanding behavioral changes over time.
• Quantitative learning models showed varying degrees of explanatory power.

Conclusions:

• Player decisions in the centipede game are sensitive to the temporal order of actions.
• Behavioral patterns evolve over repeated interactions, influenced by cognitive learning processes.
• The study highlights the importance of incorporating dynamic learning and cognitive factors into game theory models.