Related Experiment Videos
Adaptation in a stochastic Prisoner's Dilemma with delayed information
1Faculty of Computer Science and Engineering, University of Aizu, Fukushima, Japan.
Bio Systems
|January 1, 1996
Summary
Learning automata in a Prisoner's Dilemma coevolve strategies. Delays and environmental payoff changes destabilize cooperation, leading to strategy oscillations and failed coordination.
Area of Science:
- Game Theory
- Computational Economics
- Complex Systems
Background:
- The Prisoner's Dilemma models strategic decision-making with conflicting interests.
- Learning automata offer a framework for adaptive strategy adjustment in dynamic environments.
- Real-world systems often feature delayed feedback and imperfect information.
Purpose of the Study:
- To investigate the coevolutionary dynamics of strategies in a delayed-feedback Prisoner's Dilemma.
- To analyze the impact of environmental payoff modifications on strategic stability.
- To determine conditions leading to persistent oscillations in mixed strategies.
Main Methods:
- Modeling players as coadaptive learning automata.
- Employing theoretical analysis and computational simulations.
- Utilizing linear stability analysis for mixed strategies.
Main Results:
- Coevolutionary dynamics of reward-inaction and reward-penalty schemes were elucidated.
- Delayed environmental effects and modified payoffs were shown to induce strategy instabilities.
- Persistent oscillations in mixed strategies were observed under specific conditions.
Conclusions:
- Coadaptive learning in delayed Prisoner's Dilemma settings can lead to unstable strategies.
- Environmental interventions that alter payoffs can exacerbate strategic instability.
- Coordination failures are a likely outcome in such complex, dynamic systems.