Neo-satisficing Control for Cooperative Systems: A Non Heuristic Alternative to Optimization-Based Multiagent Control

Classical optimization-based solution concepts do not easily accommodate sophisticated social relationships such as cooperation, compromise, and negotiation. Satisficing decision theory, which seeks to define solutions that are "good enough," is potentially more amenable to such social behavior. The concept of satisficing introduced by Simon, however, relies on heuristic criteria to define "good enough" solutions that do not systematically extend to the multiple agent case. This paper presents an alternative notion, termed "neo-satisficing," that is explicitly designed to accommodate socially conciliatory behavior. The essential feature of this approach is that it replaces a single utility with a pair of utilities: one as a measure of the effectiveness of the outcomes, and the other as a measure of their efficiency. Each decision maker uses its utilities to define a non-singleton satisficing set of outcomes that are "good enough" according to well-defined and systematic criteria. The agents may then combine these satisficing sets to define a satisficing compromise set of outcomes, thereby leading to principled negotiation protocols.