Negotiation with incomplete information about worth: Strict versus tolerant mechanisms

Research in Distributed Artificial Intelligence (DAI) has for years been concerned with mechanisms of negotiation. In previous work, we considered situations where agents' goals were private information [12, 14, 15]. In order to carry out the negotiation, the agents were to declare, in a-l-phase, their goals. We then analyzed what goal declaration strategies the agents might adopt to increase their utility. In this paper, we consider an inverted situation, where the agents' goals (and therefore stand-alone costs) are common knowledge, but where the worth they attach to their goals is private information. The agents declare, in a-l-phase, their worths, which are then used as a baseline to the utility calculation (and thus affect the negotiation outcome). We are concerned with analyzing what worth declaration strategies the agents might adopt to increase their utility. We introduce two mechanisms, one "strict," the other "tolerant," and analyze their affects on the stability and efficiency of negotiation outcomes. The strict mechanism turns out to be more stable, while the tolerant mechanism is more efficient.