ISSN 0439-755X
CN 11-1911/B
主办:中国心理学会
   中国科学院心理研究所
出版:科学出版社

心理学报 ›› 2022, Vol. 54 ›› Issue (12): 1517-1531.doi: 10.3724/SP.J.1041.2022.01517

• 研究报告 • 上一篇    下一篇

基于期望值最大化的理论何时失效:风险决策中为自己-为所有人决策差异的眼动研究

刘洪志3, 李兴珊1,2, 李纾1,2,4, 饶俪琳1,2()   

  1. 1中国科学院行为科学重点实验室, 中国科学院心理研究所, 北京 100101
    2中国科学院大学, 北京 100049
    3南开大学周恩来政府管理学院社会心理学系, 天津 300350
    4浙江大学心理与行为科学系, 杭州 310028
  • 收稿日期:2022-01-22 发布日期:2022-09-23 出版日期:2022-12-25
  • 通讯作者: 饶俪琳 E-mail:raoll@psych.ac.cn
  • 基金资助:
    国家自然科学基金项目(71901126);国家社会科学基金重大项目(19ZDA358);教育部人文社会科学研究青年项目(19YJC190013);中央高校基本科研业务费专项资金(63222045)

When expectation-maximization-based theories work or do not work: An eye-tracking study of the discrepancy between everyone and every one

LIU Hong-Zhi3, LI Xingshan1,2, LI Shu1,2,4, RAO Li-Lin1,2()   

  1. 1CAS Key Laboratory of Behavioral Science, Institute of Psychology, Chinese Academy of Sciences, Beijing 100101, China
    2University of Chinese Academy of Sciences, Beijing 100049, China
    3Department of Social Psychology, Zhou Enlai School of Government, Nankai University, Tianjin 300350, China
    4Department of Psychology and Behavioral Sciences, Zhejiang University, Hangzhou 310028, China
  • Received:2022-01-22 Online:2022-09-23 Published:2022-12-25
  • Contact: RAO Li-Lin E-mail:raoll@psych.ac.cn

摘要:

主流的风险决策理论专家发展了一系列基于期望值最大化(expectation-maximization)的理论, 以期捕获所有人的风险决策行为。然而大量证据表明, 这些基于期望值最大化的理论并不能如同描述性理论那样理想地描述单一个体的决策行为。本研究采用眼动追踪技术, 系统考察了个体在为所有人决策与为自己决策时的风险决策行为及信息加工过程的差异。本研究发现, 基于期望值最大化的理论可捕获为所有人决策或为自己多次决策时的情况, 却不能很好捕获个体为自己进行单次决策时的情况。本研究结果有助于理解基于期望值最大化的理论与启发式/非基于期望值最大化的理论的边界, 为风险决策理论的划分和发展提供实证参考。

关键词: 风险决策, 为所有人决策, 期望值最大化, 为自己-为所有人决策差异, 眼动追踪技术

Abstract:

Mainstream theorists in risky decision-making have developed various expectation-maximization-based theories with the ambitious goal of capturing everyone’s choices. However, ample evidence has revealed that these theories could not capture every individual’s (“every one’s”) actual risky choice as descriptive theories. Substantial research has demonstrated that people do not follow the logical process suggested by expectation-maximization-based theories when making risky choices but rather rely on simplifying heuristics. From our perspective, the possible reason why mainstream decision theorists did not abandon the framework of expectation is that these theorists never doubted the validity of the expectation rule as a descriptive rule in describing decision-making under risk. We believe that expectation-maximization-based theories may capture risky choices when individuals make decisions for everyone. However, whether these theories could capture risky choices when individuals make decisions for themselves cannot be taken for granted. We thus used an eye-tracking technique to explore whether a theory for everyone would work well for every one.

A total of 52 college students participated in the experiment. Three risky choice tasks were conducted in the present study: a D-everyone task, a D-multiple task, and a D-single task. In the D-everyone task, participants were asked to choose the more optimal option out of two options under the assumption that their selection would be the final decision for everyone who was facing the same choice—that is, everyone would be subject to the same choice but could receive different outcomes. In the D-multiple task, participants were asked to choose between the two options under the assumption that their selection would be applied a total of 100 times. In the D-single task, participants were asked to choose between the two options under the assumption that their selection would be applied only once to themselves. The participants’ eye movements were recorded while they performed the tasks.

Behavioral results revealed that, compared with the D-single task, participants selected more choices correctly predicted by EV and EU theories, and took a longer time to make a decision in the D-everyone and D-multiple tasks. Furthermore, eye movement measurements revealed the following. (1) The scanpath patterns of the D-everyone task and D-multiple task were similar but different from those of the D-single task. (2) The depth of information acquisition and the level of complexity of information processing in the D-everyone task and D-multiple task was higher than that in the D-single task. (3) The direction of information search in the D-everyone task and D-multiple task was more alternative-based than that in the D-single task. (4) The eye-tracking measures mediated the relationship between the task and the EV-consistent choice. In summary, behavioral and eye movement results supported our hypotheses that participants were likely to follow an expectation strategy in the D-everyone and D-multiple tasks, whereas they were likely to follow a heuristic/non- expectation strategy in the D-single task.

We found that expectation-maximization-based theories could capture the choice of an individual when making decisions for everyone and for self in a multiple-play condition but could not capture the choice of an individual when making decisions for self in a single-play condition. The evidence for the discrepancy between everyone and every one, which was first reported in our study, implied that the possible reason why expectation-maximization-based theories do not work is that a default compatibility between the full set (everyone) and the subset (every one) does not exist. Our findings contribute to an improved understanding of the boundaries of expectation-maximization-based theories and those of heuristic/non-expectation models. Our findings may also shed light on the general issue of the classification of risky decision-making theories.

Key words: risky choice, decision for everyone, expectation-maximization, discrepancy between everyone and every one, eye-tracking

中图分类号: