449. Inducing Cooperation in Social Dilemmas
Invited abstract in session MC-10: Optimization, Learning, and Games II, stream Optimization, Learning, and Games.
Monday, 14:00-16:00Room: B100/8011
Authors (first author is the speaker)
| 1. | Stefanos Leonardos
|
| Informatics, King's College London |
Abstract
Social dilemmas illustrate situations where individual interests conflict with collective welfare, often leading to outcomes that harm the group whilst being rational for individuals. Despite this tension, real-life observations suggest that cooperation between individuals not only emerges but is key to the development of human societies. We first analyze social dilemmas through the lens of selfishness level, a game-theoretic metric that quantifies incentives for defection and prescribes the payoff modifications needed to induce prosocial behavior. We then address limitations of the canonical social dilemma model by introducing a novel multi-agent reinforcement learning mechanism that equips agents with partner choice, fostering human-aligned decision-making. Our approach promotes sustained cooperation across diverse social dilemmas and enhances learning robustness, even under unfavorable initial conditions.
Keywords
- Computational game theory
- Optimization for learning and data analysis
- AI based optimization methods
Status: accepted
Back to the list of papers