The Black Box as a Control for Payoff-Based Learning in Economic Games

Area of Science:

• Behavioral Economics
• Game Theory
• Social Psychology

Background:

• The black box method was designed to study payoff-based learning in public goods games by isolating social interactions.
• This method aimed to create a null hypothesis for social behavior, contrasting with rational self-interested agent models (Homo economicus).
• A limitation was the potential for participants to unintentionally influence each other's learning processes.

Purpose of the Study:

• To test a solution for the limitation of the original black box method by introducing simulated players.
• To evaluate the efficacy of a computerised black box in public goods games for studying learning.
• To determine if the computerised black box can be used interchangeably with the original method.

Main Methods:

• Public goods games were employed with participants interacting with either real players (original black box) or simulated players (computerised black box).
• The study involved repeated interactions where participants made decisions about contributing virtual coins with uncertain returns.
• Learning rates were compared between the two conditions.

Main Results:

• No significant differences were found in the rates of learning between the original black box method and the computerised black box method.
• The computerised black box successfully replicated the learning dynamics observed in the original setup.
• Simulated agents can be programmed to exhibit diverse behaviors.

Conclusions:

• The computerised black box method is a viable alternative to the original black box, offering a controlled environment for studying learning in games.
• This method effectively eliminates unintended social influences on learning, providing a more isolated learning environment.
• The computerised black box holds significant potential for future research on individual and group learning in various social dilemma contexts.