中国科学院行为科学重点实验室知识库

基于信息加工过程视角,本研究采用眼动技术检验风险决策整合模型和占优启发式模型。结果表明,自主决策任务条件下决策过程反应时及信息搜索模式均不同于期望价值(EV)迫选任务条件下的决策过程反应时及信息搜索模式;自主决策任务条件下决策过程反应时并未随着选项间整体值(CPT值)差值的变大而变快,且基于特征(attribute-based)的信息搜索多于基于选项(option-based)的信息搜索,不符合整合模型预期。此外,决策者亦未按照占优启发式模型所假定的决策步骤进行决策。基于信息加工过程的检验结果既不利于以累积预期理论为代表的整合模型,亦不利于占优启发式模型。文章建议从决策过程视角检验已有决策模型及建立新的启发式决策过程模型(processmodel)。

Theories intended to describe decision making under risk and uncertainty can be divided into integrative models and heuristic models, according to their theoretical basis. The integrative model implies that individuals integrate outcomes and probabilities in a compensatory way and select the option with the highest weighted sum. The heuristic model, in contrast, assumes that people do not integrate these kinds of information but rely on a repertoire of simple decision strategies, called heuristics, to make inferences, choices, estimations, and other decisions. The present paper tested the integrative model and the priority heuristic model from the point of view of choice process by using an eye-tracking system. The results show that the decision time and information acquisition pattern differ when participants make choices according to their own rules as opposed to making choices according to the imposed EV/PH rule. Specifically, the decision time did not decrease with the increased difference between the CPT values in the self-rule condition, as it did in the imposed EV rule condition. Process measures further indicated that individuals did not rely on deliberate or automatic calculations of weighted sums because attribute-based transitions were observed more frequently than option-based transitions in the self-rule condition. Furthermore, crossed transitions occurred more frequently when the two choices were cross-presented than when they were not. These results conflict with the prediction of the integrative model. Brandstatter et al. （2006） claimed that the priority heuristic could not only provide superior predictions of the output of a choice process but could also account for information acquisition. The information, Brandstatter suggested, should be considered in the order of minimum outcome, probability of minimum outcome, and maximum outcome. Although the measure of decision time supported the PH in the present study, process measures contrasted with the PH. More attention （dwelling time, fixation count and transitions） was paid to the maximum outcome than to the minimum outcome in ＂one-step＂ decision problems, and more attention （dwelling time, fixation count and transitions） was paid to the probability of maximum outcome than to the probability of minimum outcome in ＂three-step＂ decision problems. These results conflict with the prediction of the priority heuristic. In sum, neither the integrative model nor the priority heuristic could account for the data that we observed on choice process. This study encourages the use of process-oriented models and data in decision research rather than simply assessing the predictions of as-if models.