Passive threats among agents in state oriented domains

Previous work in multiagent systems has used tools from game theory to analyze negotiation among automated agents in cooperative domains. Rosenschein and Zlotkin, using these tools, provided a general mechanism for two-agent negotiation in an isolated State Oriented Domain (SOD) encounter, and also provided a classification that divided these encounters into four basic types. Other multiagent systems work considered the notion of threats during negotiation, but in order to do so introduced additional assumptions on the domain. This paper presents a new model of threats among negotiating agents in State Oriented Domains that requires no additional domainspecific assumptions. We assume that agents may use a "threat of passivity" against other agents - in other words, threatening to remain inactive (and not exploit existing cooperative opportunities), forcing both agents to satisfy their goals on their own (serially). The possibility of this negotiation threat adds interesting complexity to the four basic SOD encounter types, further subdividing them into additional types of encounter. We analyze these new encounter types that arise when there is the possibility of "passive threats", providing a thorough characterization of their properties.