Venture Theory: A Model of Decision Weights
提出风险理论模型，解释人们如何评估决策权重，通过锚定概率并心理模拟调整，受收益大小、概率极端性和模糊性影响，并用三个实验验证了关于风险与模糊态度的14个预测。
Several theories suggest that people replace probabilities by decision weights when evaluating risky outcomes. This paper proposes a model, called venture theory, of how people assess decision weights. It is assumed that people first anchor on a stated probability and then adjust this by mentally simulating other possible values. The amount of mental simulation is affected by the absolute size of payoffs, the extent to which the anchor deviates from the extremes of 0 and 1, and the level of perceived ambiguity concerning the relevant probability. The net effect of the adjustment (i.e., up or down vis-à-vis the anchor) reflects the relative weight given in imagination to values above as opposed to below the anchor. This, in turn, is taken to be a function of both individual and situational variables, and in particular, the sign and size of payoffs. Cognitive and motivational factors therefore both play important roles in determining decision weights. Assuming that people evaluate outcomes by a prospect theory value function (Kahneman and Tversky 1979) and are cautious in the face of risk, fourteen predictions are derived concerning attitudes toward risk and ambiguity as functions of different levels of payoffs and probabilities. The results of three experiments are reported. Whereas only a subset of the model's predictions can be tested in Experiment 1, all fourteen are tested in Experiments 2 and 3 using hypothetical and real payoffs, respectively. Several of the model's predictions are not supported in Experiment 2 but almost all are validated in Experiments 1 and 3. The failures relate to the exact nature of probability × payoff interactions in attitudes toward risk and ambiguity for losses. The theory and results are discussed in relation to other experimental evidence, future tests of the theory, alternative models of risky choice, and implications of venture theory for explaining further phenomena.