社会与工程心理学研究室知识库

冲动性作为一个心理特质，常常被用于解释和预测不同的人类行为。然而，以往研究提出了多种冲动性的定义和维度结构，相应又开发出对冲动性的不同 测量方法，学界迄今为止仍未就冲动性的定义和测量方法达成一致；这些不一致使得不同研究的结果无法直接比较，严重阻碍了冲动性研究的进展。面对这些问题，甚至有学者开始悲观地呼吁放弃冲动性这一心理构念。针对冲动性研究领域的这一混乱局面，本论文展开了三项研究（N = 3136），以整合冲动性测量方法、探索冲动性产生的动机因素、检测冲动性对行为的预测能力和对问题 行为人群的辨别能力。

研究一旨在整合不同的冲动性测量方法，检验在它们之间是否存在一个公共因子，回答“冲动性是否是一个心理特质”这一争议性问题。在本研究中，我们进行了目前为止冲动性领域最大规模的研究，对每名被试施测了常用的10个冲动性自我报告量表和10个行为测量任务（N = 1676）。结果显示：（1）冲动性符合双因子结构模型，在所有测量变量中可提取出一个一般冲动性因子I和六个特殊因子；（2）因子I可解释 20%的测量变异，且具有很高的重测信度（r = 0.85）；（3）使用机器学习算法，发现因子I 对冲动购物、冲动饮食、短视频 APP 使用和社交媒体使用具有显著的预测作用；（4）依托数据驱动方法开发出的一个新型冲动性量表可测量一般冲动性，其内部一致性系数（α = 0.94）、重测信度（r = 0.85），以及与因子 I 的相关（r = 0.93）均很高，且在行为预测力上与因子I相当。总体而言，在研究一中，我们发现了一个具有时间稳定性、对行为具有预测作用、且可被良好测量的一般冲动性因子。这一结果为冲动性作为一种有效的心理特质提供了全新且全面的证据，驳斥了之前研究者提出的放弃冲动性这一心理构念的主张。

研究二进一步探索了冲动性产生的动机，即是什么导致了冲动性。我们对被试（N = 500）同时测量了一般冲动性和三种动机变量，分别为认知需求、认 知闭合需求和对不确定性的不容忍程度，并对它们之间的关系进行分析。结果 显示：（1）一般冲动性与认知需求存在负相关、与认知闭合需求以及对不确定性的不容忍程度存在正相关；（2）三个动机变量之间存在较高的相关，因素分析结果显示三者间存在一个公共因子，代表对快速解决不确定性的渴望；（3） 该公共因子可解释 56%的测量变异，且对一般冲动性具有一定的预测力；此外， 在控制了该因子的影响后，认知需求对一般冲动性还表现出额外预测力，主要源于认知需求所包含的认知努力倾向因素。因此，对快速解决不确定性的渴望和低认知努力倾向可能是导致冲动性的两个重要动机。

研究三围绕冲动性测量的应用，旨在探索冲动性对问题行为人群的辨别能力。在本研究中，我们对“在押人员”这一特殊群体的冲动性进行了测量（N = 960），包括三个冲动性量表和两个行为测量任务。我们首先对比了不同冲动性测量指标对在押人员与普通被试（来自于研究一）群体的区分能力，然后应用机器学习算法构建了分类模型，对在押人员和普通被试进行类别区分，最后比较了不同犯罪类型的在押人员之间在冲动性水平上的差异。结果显示：（1）冲 动性行为测量指标的区分能力高于自我报告量表，这可能是在押人员有更强的动机低报自己的冲动性水平所导致的；（2）使用撒谎率和冲动性测量指标作为线索构建的一个二阶段分类模型对在押人员和普通被试的区分具有较高的准确率（d' = 1.78，击中率 = 0.85，错报率 = 0.21）；（3）在不同犯罪类型的在押人员中，诈骗犯的冲动性显著低于其他三类在押人员（即暴力犯、贩毒犯和盗窃犯），而另外三类在押人员在冲动性水平上并无显著差异。这些结果为冲动性作为诊 断工具提供了支持。

总体而言，本研究通过三项大规模研究，对冲动性测量进行了全面整合，并进一步探索了冲动性的动机因素和应用价值。本研究的主要贡献和创新点在于：（1）基于大规模研究，首次在冲动性研究领域提出了双因子结构模型，为冲动性作为一种心理特质提供了坚实的证据支持，为更好地理解不同的冲动性定义和维度因素提供了一个整合框架，为未来冲动性研究提供了理论指导；（2）首次探索了冲动性产生的动机因素，找到了两个可能导致冲动性的动机变量；（3）运用机器学习算法及相关数据挖掘技术开发了更优的冲动性测量方法，为未来研究提供了一个强有力的测量工具，并构建了不同的预测和分类模型，为冲动性的应用实践提供新思路与技术路径。

Impulsivity is often used as a personality trait to explain and predict various human behaviors. However, previous studies have proposed a variety of definitions and dimensional structures of impulsivity, and correspondingly developed different measures of impulsivity. Till now, there is still no consensus on the definition and measurement of impulsivity. These inconsistencies made it impossible to directly compare the results of different studies, which seriously hindered the progress of impulsivity research. In light of these problems in impulsivity research, some researchers even began to pessimistically call for the rejection of impulsivity as a psychological construct. To address these issues, we conducted three studies (N = 3136) to integrate impulsivity measures, explore the motivational factors underlying impulsivity, and test the ability of impulsivity to predict behavior and discriminate among populations with problematic behavior.

In Study 1, we aimed to integrate different measures of impulsivity, to examine whether there is a common factor among them, and to answer the controversial question of whether impulsivity is a psychological trait. Specifically, we conducted the largestscale study of impulsivity to date, administering 10 commonly used self-report scales of impulsivity and 10 behavioral tasks to each participant (N = 1676). Results showed that (1) impulsivity fit a bi-factor model, with one general impulsivity factor I and six specific factors extractable from all measured variables; (2) factor I explained 20% of the variance and had high test-retest reliability (r = 0.85); (3) using machine learning algorithms, factor I was significantly predictive of impulsive buying, impulsive eating, short-video app use, and social media use behaviors; (4) a novel impulsivity scale developed using a data-driven approach to measure general impulsivity had high internal consistency (α = 0.94), test-retest reliability (r = 0.85), and correlation with factor I (r = 0.93), and was comparable to factor I in terms of behavioral predictive power. Overall, in Study 1, we found a general impulsivity factor that is temporally stable, predictive of behavior, and can be well measured. This finding provides new and comprehensive evidence for impulsivity as a valid psychological trait, refuting previous claims that impulsivity should be rejected as a psychological construct.

In Study 2, we further explored the motivation for impulsivity, i.e., what leads to impulsivity. We measured participants’ (N = 500) general impulsivity and three motivational variables (i.e., cognitive need, cognitive closure need, and uncertainty intolerance) and analyzed the relationships between them. The results showed that (1) general impulsivity was negatively correlated with need for cognition, positively correlated with need for cognitive closure, and intolerance of uncertainty; (2) there was a high correlation among the three motivational variables, and the results of the factor analysis showed that there was a common factor among them, representing the desire for a quick resolution of uncertainty; and (3) the common factor explained 56% of the variance and had predictive power for general impulsivity; furthermore, after controlling for the effect of this factor, the need for cognition showed additional predictive power for general impulsivity, mainly representing the cognitive effort tendency. Thus, a desire for quick resolution of uncertainty and a low cognitive effort tendency may be two important motivators of impulsivity.

In Study 3, we focused on the application of impulsivity measures, and examined the ability of impulsivity to discriminate between populations with problematic behavior. We measured impulsivity in a specific group of “prisoners” (N = 960), including three impulsivity scales and two behavioral measurement tasks. We first compared the ability of different impulsivity measures to discriminate between prisoners and non-prisoners (from Study 1), then applied machine learning algorithms to construct a classification model to discriminate between prisoners and non-prisoners, and finally compared the differences of impulsivity between prisoners with different types of convictions. The results showed that (1) the discriminative power of behavioral impulsivity measures was higher than that of self-report scales, which may be caused by the fact that prisoners had a stronger motivation to underreport their impulsivity level; (2) a two-stage classification model constructed using the lie rate and impulsivity measures as cues had a high accuracy in discriminating between prisoners and nonprisoners (d' = 1.78, hit rate = 0.85, and false alarm rate = 0.21); and (3) among different types of prisoners, fraudsters were significantly less impulsive than the other three types of prisoners (i.e., violent, drug dealing, and theft), while the other three types of prisoners did not differ significantly in impulsivity. These findings provide support for impulsivity as a diagnostic tool.

Overall, the whole study provides a comprehensive integration of impulsivity measurement through three large-scale studies, and further explores the motivational factors and application value of impulsivity. The main contributions and innovations of this study are that (1) based on the large-scale studies, a bifactor model was proposed for the first time in the field of impulsivity research, which provides solid evidence support for impulsivity as a psychological trait, provides an integrated framework for better understanding different definitions and dimensional factors of impulsivity, and provides theoretical guidance for future impulsivity research; (2) for the first time, we explored the motivational factors of impulsivity and found two motivational variables that may lead to impulsivity; (3) we developed a better measure of impulsivity using machine learning algorithms and related data mining techniques, which provides a powerful measurement tool for future impulsivity research, and we constructed prediction and classification models to provide new ideas and technical ways for the application practice of impulsivity.