Strategic negotiations for extensive-form games

When studying extensive-form games it is typically assumed that players make their decisions individually and that it is not possible for them to make formally binding agreements about future moves. As a consequence, many non-zero-sum games have been shown to have equilibria that are suboptimal and arguably counter-intuitive. For this reason we explore a new line of research in which game-playing agents are allowed to negotiate binding agreements. We analyze what happens under such assumptions and define a new equilibrium solution concept to capture this (the Negotiation Value). We show that the outcomes predicted by this new solution concept are more efficient than the Subgame Perfect Equilibrium and, therefore, arguably more realistic. Furthermore, we demonstrate experimentally that a bounded rational agent is able to approximate our solution concept in several games and that it strongly outperforms non-negotiating rational players.