Research Information System
Applied Soft Computing, vol.197, 2026 (SCI-Expanded, Scopus)
The Iterated Prisoner's Dilemma (IPD), following Axelrod's tournaments, has become one of the most extensively studied models in game theory. Many political, economic, and social interactions can be represented as Prisoner's Dilemma situations, where cooperation is socially optimal but individually irrational. Identifying mechanisms that sustain cooperation in such environments remains a central research challenge. Over time, numerous IPD algorithms have been proposed, differing in cooperation levels, adaptivity, and computational foundations. Recent AI-based strategies have achieved strong performance, often at the cost of interpretability. This paper introduces a Fuzzy Algorithm (FA), an AI-based strategy built on a Mamdani-type fuzzy inference system. The algorithm integrates four opponent characteristics: cooperation rate, adaptivity, forgiveness, and stochasticity, into a compact rule-based decision framework that models graded reasoning under uncertainty. FA is evaluated in classical Axelrod and Stewart-Plotkin tournaments, as well as in large-scale competitions involving over 200 benchmark strategies. Robustness is further tested under 10% implementation noise and varying payoff structures, and evolutionary performance is examined through ecological model and Moran processes. Results indicate that FA consistently ranks among top-performing strategies and demonstrates competitive evolutionary stability, suggesting that interpretable fuzzy inference systems can provide effective decision mechanisms in repeated strategic environments.