社会与工程心理学研究室知识库

生活中，我们的决策往往会涉及两类问题:风险((risk)和模糊(ambiguity)。 17世纪起，学界就己经开始对己知各种结果以及结果出现概率的风险决策进行了系统的研究，诞生了期望价值理论((Expected Value Theory)。即使饱受挑战，风险决策的主流理论，即补偿性的期望家族理论仍保留着期望价值理论所提出的“加权求和”的核心假设，不断地修正、发展。在现实生活中更为普遍的是概率未知或不确定的模糊决策。但是直至Ellsberg (1961)提出双色、三色问题对主观期望效用理论(Savage, 1954)造成了巨大挑战，模糊决策的理论模型才逐渐出现、发展，但其主要思路基本是对期望效用理论的修正。

然而，越来越多的研究表明，人们在进行结果和概率都已知的风险决策时并不会遵循期望法则，而是采用简单的启发式策略。在概率未知的模糊决策中，加权求和变得更加困难，而风险决策中“非主流”的启发式理论，或许能够更好的描述和解释模糊决策。但是模糊决策的研究起步较晚，己有的研究基本上都是基于传统的行为实验，更是缺乏过程追踪方面的证据。因此，本研究在决策结果研究的基础上，应用眼动技术对模糊决策的认知过程进行了直接的测量，以回答:1)个体进行模糊决策是否是“加权求和”的过程?2)与风险决策相比，模糊决策是否更可能是采用简单启发式的过程?

研究一参考Su等人(2013)风险决策研究中的比例任务范式，设计了被试在自然状态下作出选择前会有意识地进行加权求和的模糊比例任务，作为基线任务来检验模糊决策是否符合“加权求和”的假设。从行为结果来看，被试在进行两个任务时，其决策符合期望家族理论预期的比例存在边缘显著的差异，决策时间的差异显著。此外，两个任务的眼动过程差异显著，相比于需要进行加权求和的模糊比例任务，被试进行模糊决策时的信息加工复杂度较低，基于选项的加工更少，并且两个任务的扫视轨迹差异显著。该结果说明，模糊决策的眼动过程并不符合“加权求和”的假设。

研究二采用图片形式呈现概率/概率区间，以最大程度减少由实验材料本身的差异造成的影响，通过对模糊决策和风险决策的直接比较，来探究与风险决策相比，模糊决策是否更可能是采用简单启发式的过程。在行为结果层面，被试的选择结果符合期望家族预期的比例存在边缘显著的差异，而决策时间的差异不显著。进一步对眼动过程进行分析，发现:从信息加工复杂度指标上看，并未发现模糊决策与风险决策有显著差异;对信息搜索方向的指标进行分析，结果发现被试进行模糊决策任务时的结果概率眼跳比例显著低于风险决策任务，说明模糊决策更不符合“加权求和”的眼跳模式;对两类决策任务扫视轨迹的分析表明，二者整体、动态的眼动过程并不相同。此外，在进行风险决策任务时，被试在计算难度低的条件下，与计算难度高的条件相比，反应时更短、基于选项的加工更多;而在模糊决策任务中，并未发现计算难度的影响。以上结果说明与风险决策相比，个体进行模糊决策时可能并不存在“加权求和”的计算过程、而更可能是采用简单启发式的过程。

本文从行为结果和眼动过程两个层面进行了一系列比较。其中对眼动过程特征的分析，主要是基于“补偿/非补偿性”以及“基于维度/基于选项”的决策特征，对不同决策类型的加工过程更符合哪种理论决策模型的假设进行了检验。整体而言，研究一和研究二的结果说明，个体进行模糊决策时并非是“加权求和”的过程，并且与风险决策相比，模糊决策更可能是采用简单启发式的过程。

Decision problems often come in two kinds in our daily life: risk and ambiguity. Since the birth of Expected Value Theory in the beginning of the 17th century, systematically study of risk decision-making that has known results with known probabilities has begun. Although it has been continuously challenged, the mainstream theory of risk decision-making, that is, the family of compensatory model, still retains the core assumption of the "weighting and adding" of the Expected Value Theory, and has constantly revised and developed. More common in real life are ambiguity decisions with unknown or uncertain probabilities. Until Ellsberg (1961) proposed two-color and three-color problem that posed a great challenge to the Subjective Expected Utility Theory (Savage, 1954), the theoretical model of ambiguity decision-making gradually emerged and developed, but its main idea is basically the revision of the Subjective Expected Utility Theory.

However, more and more studies have testified that people do not follow the rule of "weighting and adding" when making risky choice with known outcomes and probabilities, but adopt a simple heuristic strategy. In ambiguity decision-making with unknown or uncertain probabilities, the process of "weighting and adding" becomes more difficult, and the "non-mainstream" heuristic theory in risk decision-making may be able to describe and explain ambiguity decision-making better. However, the study of ambiguity decision-making started late and existing research is mainly based on traditional behavioral experiments, and it is also lacking evidence from process tracking. Therefore, on the basis of behavioral results, this study uses eye-movement technology to directly measure the cognitive process of decision-making, in order to answer: 1) Does individual make ambiguity decision based on a "weighting and adding" process? 2) Compare with risk decision-making, is ambiguity decision-making more likely to be a simple heuristic process?

Study 1 refers to the proportion task paradigm in the risk decision study of Su et al. (2013) and designs an ambiguity proportion task in which participants will consciously weight outcome by probability before making choices in their natural state. Thus, study 1 adopted ambiguity proportion task as baseline to test whether the ambiguity decision is consistent with the "weighting and adding" hypothesis. From the behavioral results, the difference of the proportion of choices predicted by EV family is marginally significant, and the difference in reaction time is significant between the two tasks. In addition, the eye movement process of the two tasks is significantly different. Compare to the ambiguity proportion task that require weighting and adding, the information processing complexity of ambiguity decision is lower, the option-based information search is less, and the scanpath of the two tasks is significantly different Results of study 1 revealed that the eye movement process of ambiguity decision-making is inconsistent with the assumption of "weighting and adding".

Study 2 pr0esents probability/probability interval in the form of pictures to minimize the impact caused by the differences of experimental materials. Through the direct comparison between ambiguity decision and risk decision, study 2 explored whether making ambiguous choices is more likely to be a simple heuristic process than risky choices. At the level of behavioral results, marginally significant difference was found in the proportion of choices predicted by EV family between the two tasks, but there was no significant difference in decision time. By further analyzing the eye movement process, we found that: From the information processing complexity index, there was no significant difference between the two decision-making tasks; analysis of information search direction suggested that when subjects were making ambiguity decisions saccades between outcome dimension and probability dimension were significantly less than that of risk decision-making task, indicating that ambiguity decision-making was inconsistent with the eye-movement pattern of "weighting and adding". Analysis of scanpath shown that the overall and dynamic eye-movement process was significantly different between the two tasks. Moreover, in the task of risk decision-making, compare with the high computing difficulty condition, when subjects were making risky choice under low computing difficulty condition, their reaction time was shorter, and they had more option-based processing. But in the ambiguity decision task, no effect on computing difficulty was found. Results of study 2 suggested that compare with risk decision-malting, individuals are more likely to adopt a simple heuristic strategy when making ambiguity decisions rather than to follow the rule of "weighting and adding".

This article made a series of comparisons at the level of behavioral results and eye movement process. The analysis of eye movement process is mainly based on the "compensatory/non-compensatory" and "dimension-based/option-based" decision features to test that the cognition process of different decision types is more in line with which theoretical decision model. Overall, the results of Study 1 and Study 2 suggested that ambiguity decision-making behavior is unlikely to consistent with the process of "weighting and adding", and compare with risk decision-malting, individuals are more likely to adopt simple heuristic strategy when making ambiguity decisions.