Learning, exploitation and bias in games.

Area of Science:

• Behavioral Economics
• Game Theory
• Developmental Psychology

Background:

• Individuals in groups learn and adapt behaviors through competitive interactions.
• Games with negative frequency dependence create scenarios where action popularity impacts payoff.
• Equilibrium configurations occur when no individual benefits from unilaterally changing strategy.

Purpose of the Study:

• To investigate learning dynamics in competitive group development.
• To analyze the consequences of unbiased reinforcement learning in such environments.
• To explore the potential for exploitation and rational avoidance strategies.

Main Methods:

• Modeling a two-action competitive game with negative frequency dependence.
• Defining and analyzing equilibrium configurations based on individual payoffs.
• Simulating unbiased reinforcement learning to observe group behavior evolution.
• Employing evolutionary simulations to test exploitation avoidance mechanisms.

Main Results:

• Groups playing a competitive game with negative frequency dependence reach approximate equilibrium distributions.
• At equilibrium, a preferred action yields higher payoffs than a more prosocial action.
• Unbiased learning creates opportunities for individuals to exploit others by influencing action choices.
• Evolutionary simulations demonstrate that over-valuing preferred rewards can prevent exploitation.

Conclusions:

• Developmental learning in competitive groups can lead to unequal payoffs and exploitation.
• Rational biases, such as over-valuing preferred options, can serve as effective anti-exploitation mechanisms.
• Understanding learning rules is crucial for predicting group behavior and evolutionary outcomes.