Multi-Issue Negotiation Processes by Evolutionary Simulation, Validation and Social Extensions

We describe a system for bilateral negotiations in which artificial agents aregenerated by an evolutionary algorithm (EA). The negotiations are governed bya finite-horizon version of the alternating-offers protocol. Several issuesare negotiated simulataneously. We first analyse and validate the outcomes ofthe evolutionary system, using the game-theoretic subgame-perfect equilibriumas a benchmark. We then present two extensions of the negotiation model. Inthe first extension agents take into account the fairness of the obtainedpayoff. We find that when the fairness norm is consistently applied during thenegotiation, agents reach symmetric outcomes which are robust and ratherinsensitive to the actual fairness settings. In the second extension we modela competitive market situation where agents have multiple bargainingopportunities before reaching the final agreement. Symmetric outcomes are nowalso obtained, even when the number of bargaining opportunities is small. Wefurthermore study the influence of search or negotiation costs in this game.