Decision Analysis Approach to Competitive Situations with a Pure Infinite Regress*
针对竞争情境中决策者与对手行动相互依赖导致的无限回归问题，提出一种构建决策者偏好排序并同时建立对手信念一致模型的方法。
Abstract If the standard decision analysis approach was applied to a competitive situation, then the decision maker would attempt to assess a subjective probability distribution over all possible actions by the competitors or rivals. However, in assessing this robability distribution, the decision maker would quickly find him/herself in an infinite regress. There is a range of possible dependencies between the decision maker's decision and the rival's action that lead to such a regress. This paper presents an approach to overcoming this infinite regress in performing decision analyses of competitive situations with a “pure” infinite regress. In a pure infinite regress there is a dependency between the decision maker's decision and the rival's choice of action: if the decision maker prefers one alternative to any other, then the rival's actions will reflect a belief by the rival that the preferred alternative is more likely to be the decision maker's decision. The approach builds the decision maker's preferred rank order of alternatives, while simultaneously constructing a consistent model of the rival's beliefs about the decision maker's decision.