认知与发展心理学研究室知识库

摘要：户对智能辅助系统的信任是指个体所持有的对智能系统的一种态度，认为在处于不确定和易受损的情境时，智能系统仍然能够实现主体的目标。系统可靠性是影响信任的首要因素，但用户对系统的信任可能偏离系统可靠性，产生过度依赖或弃置不用的不合理交互行为。因此研究者积极探索其他影响信任的因素以达到“校准”信任的目的。研究表明，智能系统的拟人性对人机信任存在影响，而群体身份特征这一新的拟人特性最近开始受到研究者的关注，但之前的研究多关注交互偏好等问题。本研究将探索群体身份对人机信任的影响，认为用户会更加信任具有内群体身份的智能系统。本研究采用2（智能辅助系统的群体身份：内群体身份/外群体身份）×3（智能辅助系统的可靠性水平：95%/70%/45%）的混合实验设计，其中智能辅助系统的群体身份是被试内变量，由最简群体范式进行操控；可靠性水平是被试间变量；因变量是TNO信任任务（TNO Trust Task，T3）中被试对系统推荐的信任和遵从程度。结果发现：智能辅助系统的可靠性水平越高，被试的信任程度越高；群体身份会影响被试对智能辅助系统的信任，在三种可靠性水平下，被试对内群体的智能辅助系统信任都比外群体智能辅助系统的信任高。我们进一步就该发现的理论和应用价值进行探讨。

Abstract: User’s trust in intelligent system can be defined as the attuite that a system will achieve an individual’s goals in a situation characterized by uncertainty and vulnerability. The reliability of intelligent system is the primary factor of trust, but user’s trust in the system may deviate from system reliability, resulting in irrational interactive behaviors that are misuse or disuse. Therefore, researchers actively explore other factors that influence trust in the system to achieve the goal of calibrating trust. Researches showed that anthropomorphism of intelligent system affects human-computer trust, and group membership, the new anthropomorphic characteristic, has attracted researchers’ attention recently, but previous studies have focused on the issues such as interaction preference. This study will explore the influence of group membership on human-computer trust and consider that users have more trust in the in-group intelligent system. A mixed design of 2 (group membership of intelligent system: in-group, out-group) × 3 (reliability level of the system: 95%, 70%, 45%) was used in this study, group membership was a within-subject factor, the minimum group paradigm was adopted to control this factor; and reliability level was a between-subject factor; dependent variable in this study is trust that complies with system recommendation in TNO Trust Task. Results showed that the higher reliability level of the system, the higher trust of the system; group membership will affect participants’ trust in intelligent system, the trust in in-group intelligent systems is higher than that of out-group intelligent systems in all three reliability levels. We will further explore the theoretical and practical value of this discovery.