社会与工程心理学研究室知识库

风险决策与跨期决策是日常生活中常见的，与人类生存发展密切相关的两类决策。这两类决策在理论发展、行为效应和认知神经基础等多个层面具有相似性，其核心之处在于，这两类决策在决策过程上可能具有相似性。主流的折扣模型认为，两类决策的过程相似，均符合补偿性的、基于选项加工的规则，而非折扣模型则认为，两类决策过程的相似性在于二者均符合非补偿性的、基于维度加工的规则。但是，以往多数研究多基于行为结果的证据探讨二者的相似性，目前罕见研究探索二者是否具有共同的决策过程。
因此，为检验风险决策和跨期决策是否具有共同的过程机制，本文从基于行为效应的“特殊”条件到基于等量转换的“一般”条件，从“简单”的单决策结果情境到“复杂”的双决策结果情境的研究逻辑，采用眼动追踪方法，运用贝叶斯估计、分层贝叶斯模型拟合、眼动轨迹分析等一系列新分析方法，从行为特征、局部过程特征、整体过程特征三个层面，对风险决策与跨期决策进行了综合的过程比较。具体来说，研究关注决策过程的两个基本特征：“补偿/非补偿性”规则和“基于维度/基于选项”加工规则。基于局部和整体过程特征，本文选取反映“补偿/非补偿性”、“基于维度/基于选项”加工决策特征的一系列决策属性和过程指标，基于这些指标对这两种决策特征进行了检验。通过比较了两类决策的加工过程是否一样，进一步检验了两类决策更符合哪种理论决策模型的假设。  
首先，考虑到风险和跨期决策都存在相似的行为效应，研究一对特殊行为效应的过程机制进行了比较。研究一包含两个子研究，选取了风险和跨期决策中的两对相似行为效应，分别比较了单决策结果情境下风险决策的确定效应与跨期决策的即刻效应，以及双决策结果情境下风险和跨期决策的“隐藏零”效应。结果发现，两类决策任务在部分过程特征，如加工复杂程度、注意分配和整体动态的眼动过程等方面存在差异，其他的过程指标上均都表现出相似性。除局部的加工方向外，其他属性均表明两类决策都更符合非折扣模型的假设，具有非补偿的、基于维度加工的决策特征。
其次，为克服参数特异性和个体差异对行为偏好和过程的可能影响，研究二首次使用概率和时间等量转换的研究范式，通过前测生成了每个被试特有的心理等量的风险和跨期决策任务参数，从定量的、去除行为效应的角度对风险和跨期决策的一般过程机制进行了比较。研究二包含两个子研究，分别比较了在单决策结果和双决策结果情境下风险和跨期决策的决策过程。结果发现，当风险决策和跨期决策在决策参数上等量时，被试在风险和跨期决策中表现出不同的行为特征，但在加工过程上，除了在加工复杂性和整体动态的眼动过程上不一致以外，两类决策在其他过程特征上都表现出相似性。除局部的加工方向外，其他属性均表明两类决策都更符合非折扣模型的假设，具有非补偿的、基于维度加工的决策特征。
总之，本文所有研究一致地发现，虽然风险和跨期决策在少数加工过程的细节特征上有差别，如加工复杂程度、整体动态的眼动过程。但是综合来看，在核心的决策基本规则上，风险和跨期决策具有共同的认知机制：和主流的折扣模型理论假设不同，风险和跨期决策都不遵循折扣模型所假设的补偿性的和基于选项的加工规则，即包含“加权求和”和“折扣计算”加工过程；风险和跨期决策都更可能依据简捷的，非补偿性模型所预期的启发式规则进行决策。本研究结果为建立风险和跨期决策的共同解释框架做了有益的尝试，也为不同决策比较研究在研究范式和分析方法上提供了有益探索和新的方向。

Risky choice (RC) and intertemporal choice (IC) are common decisions we make in daily life, and these decisions are vital to the development of human beings. These two types of decisions are similar in terms of theoretical development, behavioral effects, and cognitive neural basis. The core similarity of the two is that  they might involve similar process mechanisms. The mainstream discounting model holds that RC and IC are similar in that they follow a compensatory, alternative-based rule. However, other models suggest that RC and IC are similar in that their processing is  non-compensatory  and attribute-based.  However, the previous research was based on evidence from outcome data and research that explore whether the two have a shared decision process are rare.  
In this paper, in order to examine whether RC and IC  involve similar decision process,  by using the logic from “special” (based on effect)  to  “general” (based on equivalence conversion), and  from “simple” (single outcome) to “complex” (double outcome), we utilized the eye tracking method to compare the underlying process in RC and IC. We adopted a series of new analytical methods, such as hierarchical Bayesian model fitting, Bayesian estimation, and scanpath analysis to compare behavioral characteristics,  local process characteristics and holistic process characteristics of the two types of decisions.  In detail, we focused on two decision factors: compensatory/non-compensatory rule, alternative based/attribute based rule. Based on the local and holistic process characteristics, we select a series of decision attributes that reflect the  compensatory/non-compensatory  rule,  alternative based/attribute based  rule, and use the eye-tracking indexes to test these  decision attributes. By  comparing the decision process the two, we futher  determine which models are more suitable for these two decisions.
In study 1, considering the similar behavioral effects of RC and IC, we compared the process mechanisms of the special behavioral effects of RC and IC from a qualitive perspective. In this study, we compared the certainty effect of RC and the immediacy effect of IC (single-outcome block) and the "hidden zero" effect of RC and IC (double-outcome block).  We found that  although  RC and IC differed in some process characteristics, such as complexity and holistic dynamic eye movement pattern, they are similar in other characteristics. In addition, except for  local processing direction, other characteristics we examined suggested that the two types of decision are consistent with the assumptions of a non-discounting models, and their information processing were more likely to follow the non-compensatory, attribute-based rule.
In study 2, to overcome the disadvantages of the previous studies, which failed to control the confounding for results from parameter specificity and individual differences, we established a paradigm of probability and delay equivalence conversion to set “equivalent RC and IC options pairs,” and we compared the process mechanisms of the two in “general condition” from a quantitative perspective. Study 2 also consists of two sub-studies. We compared the underlying process in single-outcome (sub-study 3) and double-outcome (sub-study 4) blocks. We found that based on equivalence conversion, RC and IC showed different behavioral charactoristics. However, they were similar in terms of information processing, expectation for complexity, and holistic dynamic eye movement pattern. In addition, except for local processing direction, the other characteristics we examined suggested that the two types of decisions are consistent with the assumptions of a non-discounting model, and their information processing were more likely to follow a non-compensatory, attribute-based rule.
All of the studies presented in this paper showed that although differences exist between  RC and IC  in the details of few decision processes, such as  in terms of compexity and holistic and dynamic eye movement pattern. However, at the core of the decision rule, RC and IC share cognitive mechanisms: In these two types of decisions, contrary to classical discounting models, people do not follow a  compensatory, dimension-based rules and undergo a “weighting and summing” or “delay discounting” process to make a decision. However, they are likely to use simple heuristic rules that are expected by non-discounting models. This study provides a theoretical basis for the establishment of a common interpretation framework for RC and IC, as well as opens a new direction for the research methods and analytical methods for comparative research on different decision tasks.