社会与工程心理学研究室知识库

为了将人类从繁重的驾驶任务中解放，自动驾驶技术受到了社会各界的关注，自动驾驶汽车正式加入道路交通环境指日可待。然而，作为首个由算法而非人类控制的道路交通工具，自动驾驶汽车的推行面临众多困境。其中一个问题是，人们对待自动驾驶汽车的态度和行为可能与对待其他人类驾驶员不同，进而引发安全风险和社会忧患，因此有必要提前了解人们对自动驾驶汽车的态度和行为表现，了解态度和行为差异产生的机制，以达到预测和干预的目的。以往文献表明，人们表现出对待自动驾驶汽车更少的亲社会驾驶行为、更多的攻击性驾驶行为和对道德两难选择更低的接受度。基于“机器是社会行为者”(Computers Are Social Actors, CASA)理论，人们对待自动驾驶汽车和其他人类驾驶员的态度和行为差异可能来源于人们感知到的两类驾驶主体的特质水平不同。因此，本研究尝试描绘并对比现阶段大众对自动驾驶汽车和其他人类驾驶员的特质画像以及态度和行为差异，探究特质画像差异是否对态度和行为差异具有影响，并尝试通过操纵特质感知的方式改变人们对待自动驾驶汽车的态度和行为，改善自动驾驶汽车面临的困境。具体研究如下：

研究一主要探讨人们对自动驾驶汽车群体和人类驾驶员群体的印象、态度和行为差异。首先，研究构建了现阶段大众对自动驾驶汽车和其他人类驾驶员特质印象的工具，检验获得特质感知维度，并在不同维度上对比二者之间的差异。结果表明，对自动驾驶汽车的特质感知可以通过驾驶能力、驾驶风格、社会性（互惠互利）、社会性（公序良俗）和道德风格五个维度进行描述。相比于人类驾驶员，人们感知到的自动驾驶汽车驾驶能力、社会性（互惠互利）水平、社会性（公 序良俗）水平更低。然后，研究通过问卷调研和驾驶模拟实验相结合的方式，探究人们在亲社会驾驶行为、攻击性驾驶行为和道德两难三种不同情境中是否存在 对自动驾驶汽车的态度和行为差异。结果表明，在亲社会驾驶行为情境中，被试对自动驾驶汽车的让行意愿更低；在攻击性驾驶行为情境中，对于被试倾向于向自动驾驶汽车做内部归因的驾驶错误，被试表现出更高程度的攻击意向；在道德两难情境中，当人类驾驶员和自动驾驶汽车做出相同利他选择的时候，被试对自动驾驶汽车选择的接受度更低。

研究二主要探讨特质感知在驾驶主体类型影响态度和行为中的中介作用。在亲社会驾驶行为、攻击性驾驶行为和道德两难三种情境中，均发现了中介效应存在。感知到的驾驶安全性水平中介了驾驶主体类型对让行意愿的影响，感知到的驾驶能力和社会性（公序良俗）水平中介了驾驶主体类型对攻击意向的影响，感知到的社会性（公序良俗）水平中介了驾驶主体类型对道德两难选择接受度的影响。研究初步证实感知到的自动驾驶汽车和人类驾驶员的特质水平同态度和行为差异的关联性。

研究三主要探究操纵特质感知是否影响人们对待自动驾驶汽车的态度和行为。在研究二得到的相关关系基础之上，进一步证明因果关系。通过视频+文字描述的方式，提高人们感知到的自动驾驶汽车驾驶能力和社会性（公序良俗）水平，结果发现人们更信任自动驾驶汽车的亲社会意图，对自动驾驶汽车的攻击性意向更低，对自动驾驶汽车做利己选择的接受度更高。

综上，本论文证明了人们在不同驾驶情境中对待自动驾驶汽车存在态度和行为差异，差异来自于人们感知到的自动驾驶汽车和人类驾驶员的特质水平不同，对特质感知进行操纵可减少对待自动驾驶汽车的态度和行为差异。在理论层面，本研究形成了较为全面的自动驾驶汽车特质画像，并揭示了特质感知影响人们对待自动驾驶汽车态度和行为的机制；在应用层面，本研究指出了自动驾驶汽车在混合车流中可能遇到的实际问题并提供了一种可行的干预方案。

In order to liberate humans from the burdensome task of driving, autonomous driving technology has attracted attention from various sectors of society, and it is only a matter of time before autonomous vehicles officially join the road traffic environment. However, as the first road vehicle controlled by algorithms rather than humans, the promotion of autonomous vehicles faces numerous challenges. One of these challenges is that people's attitudes and behaviors toward autonomous vehicles may be different from which toward other human drivers, leading to safety risks and social concerns. Therefore, it is necessary to understand the phenomenon and mechanism of people's different attitudes and behaviors toward autonomous vehicles in advance, so as to achieve the purpose of prediction and intervention. Previous literature has shown that people exhibited less prosocial driving behavior, more aggressive driving behavior, and less acceptance of moral judgments toward autonomous vehicles. Based on the "Computers Are Social Actors (CASA)" theory, the differences in attitudes and behaviors toward autonomous vehicles and human drivers may be due to different levels of perceived traits in the two types of drivers. Therefore, this thesis attempts to describe and compare the trait, attitude and behavior differences between autonomous vehicles and human drivers, explore whether perceived trait differences have an impact on attitudes and behaviors, and try to change people's attitude and behavior toward autonomous vehicles by manipulating trait perception. Specific studies are as follows:

The first study focuses on the differences in people's impressions, attitudes, and behaviors between autonomous vehicles and human drivers. First, the study constructed a tool for the current public's impression of the traits of autonomous vehicles and human drivers, acquired and examined trait perception dimensions, and compared the differences between the two driver types on different dimensions. The results show that the trait perception of autonomous vehicles could be described through five dimensions: driving ability, driving style, sociality (reciprocity), sociality (morality) and moral style. Compared with human drivers, people perceived that autonomous vehicles had lower driving ability, sociality (reciprocity), and sociality (morality) level. Then, through the combination of questionnaire survey and driving simulation experiment, the research explores whether people had different attitudes and behaviors toward autonomous vehicles in three different situations: prosocial driving behavior, aggressive driving behavior and moral dilemma. The results show that in the prosocial driving behavior situation, the subjects had lower willingness to yield to the autonomous vehicle; In the aggressive driving behavior situation, the subjects showed a higher degree of aggressive intention toward autonomous vehicles if the driving error was considered as the consequence of autonomous vehicles’ internal traits; In the moral dilemma situation, when the human driver and the autonomous vehicle made the same altruistic choice, the participants were less accepting of the autonomous vehicle's choice.

The second study mainly explores the mediating role of trait perception in the influence of driver type on attitude and behavior. The mediating effect was found in prosocial driving behavior, aggressive driving behavior and moral dilemma. The perceived level of driving safety mediated the influence of driver type on driving intention, the perceived level of driving ability and sociality (morality) mediated the influence of driver type on aggressive intention, and the perceived social level (morality) mediated the influence of driver type on acceptance of moral dilemmas. The study preliminarily confirms the correlation between the perceived trait levels of autonomous vehicles and human drivers and differences in attitude and behavior.

The third study mainly explores whether manipulating trait perceptions affect people's attitudes and behaviors toward autonomous vehicles. On the basis of the correlation obtained in the second study, the causal relationship is further proved. Through the way of video + text description, people's perceived driving ability and social (morality) level of autonomous vehicles were improved. And as a result, people had more trust in the prosocial intention of autonomous vehicles, less aggressive intention toward autonomous vehicles, and more acceptance of egoism choices made by autonomous vehicles.

In conclusion, this thesis proves that there are differences in people's attitudes and behaviors toward autonomous vehicles in different driving situations. The differences come from the different perceptions of trait levels of autonomous vehicles and human drivers. Manipulation of trait perception can reduce the differences in attitudes and behaviors. At the theoretical level, this study has formed a relatively comprehensive trait portrait of autonomous vehicles, and revealed the mechanism of trait perception affecting people's attitudes and behaviors toward autonomous vehicles. At the application level, this study points out the practical problems that autonomous vehicles may encounter in mixed traffic flow and provides a feasible intervention scheme.