关键词: 艾勒悖论, 独立性原则, 齐当别模型

Abstract: Background-Objective: Allais paradox (Allais, 1953) demonstrated behavior in contradiction to the independence axiom of expected utility theory and was then considered as a lever that moved EU. To date numerous revamped theories have been proposed in an attempt to resolve Allais' paradox without discarding the expectation rule, and most of them were based on the assumption that the utilities of outcomes were no longer multiplied by linear probabilities, but by nonlinear decision weights which did not necessarily conform to the rules of mathematical probability. The risky choice behavior was, however, simply seen by the equate-to- differentiate model as a choice between the best possible outcomes or a choice between the worst possible outcomes. The paradox arose because the final choice was not consistently based on a single fixed dimension (either the best possible or the worst possible outcome dimension) in each pair of choices. Method: In searching for evidence of whether decision-makers were indeed guided by the equate-to- differentiate rule in making their choices when faced with the Allais' problems, a so-called "judging" task was designed in the present study to help test the equate-to-differentiate account in further detail. Operationally, the outcomes of prospects on both the best and the worst possible outcome dimensions were paired in the present task. Subjects were then asked to choose the pair with outcomes which were, for them, the most different. If the equate-to-differentiate one-dimensional difference account was correct, then the knowledge of the chosen pair would permit explanation or prediction of option preference. The Allais' two pairs of choice problems, coupled with the judging task, were presented in questionnaire form to business students from the Nanyang Business School of Nanyang Technological University. Results: It was shown in the present experiment that the effect (phi squared) that the judging data were able to account for the variance in choice in relation to the Allais paradox was significant. Conclusion: The results therefore raised doubts about whether people behaved as if they were trying to maximize "something" and added to the evidence that whether the final selection was consistently based on a single fixed dimension was the cause which was most likely to be responsible for a violation of Savage' s sure-thing principle. Though expected utility theory was known to be wrong through the Allais paradox, to rescue it by proposing nonlinear decision weights was very likely to do nothing but hide an old mistake under a new one.

Key words: Allais Paradox, Independence, Equate-to-differentiate approach