糖心TV

Patrick O'Callaghan (糖心TV Mathematics Institute)

Context vs Context-Free Preferences: A No-Brainer?

If context is taken to mean anything that affects preferences, can change and is beyond the immediate control of a decision-maker, then it is clear that context plays a major role in decision theory. I will give a brief background on context-free preferences and using a simple flood example, argue that context preferences provide a more intuitive approach to representing behavior. However, contrary to the work of Gilboa and Schmeidler (1995, 2003) I argue that, when the interpretation of preference is deterministic, a representation that is both linear in contexts and separable across alternatives (e.g., "expected utility theory" or "case-based decision theory") imposes three unreasonable conditions. Worse, jointly these conditions are so strong that, even for surprisingly simple decision problems, they exclude all context preferences! The alternatives are clear:
1. settle for non-utility representations similar to those of Loomes and Sugden (1981); or
2. give up on the linearity requirement; or
3. extend preferences so that the decision-maker ranks contexts as well as alternatives.

Poster-February-2-2012