| The trust problem in human-machine hybrid systems intensifies with the deepening of human-machine interaction,mainly manifested as overtrust and distrust.The two types of problems,distrust,and overtrust,can gradually worsen during the interaction process and even have catastrophic consequences.The current research on trust between humans and machines is mainly theoretical,lacking empirical research.At the same time,the theory of trust restoration between humans and machines is still incomplete,and interpretable artificial intelligence technology has achieved a series of development results in recent years.However,there is relatively little research on interpretable artificial intelligence technology in human-computer interaction,especially in the study of the impact of interpretable technology on human-machine trust.To solve the above problems and find the rules of the impact of artificial intelligence technology on human-machine trust,this article conducted the following research work:(1)In response to the problem of information complexity exposed by interpretable artificial intelligence technology to users,this article proposes a three-layer interpretable model based on situational awareness,which classifies and displays interpretable information,aiming to maximize the utilization of artificial intelligence interpretable information while reducing users’ cognitive load.To further explore the impact of model interpretability technology on trust in human-machine hybrid systems,this article developed an experimental platform that integrates artificial intelligence technology and interpretable artificial intelligence technology.The experimental platform implements a three-layer interpretable model and can efficiently collect experimental data from participants.(2)For the issue of trust measurement,this article adopts a combination of subjective and objective trust measurement methods.By collecting dynamic trust and dependency values of users on the system after each round of tasks,we subjectively analyze the trust changes between human-machine collaboration and calculate the difference between user decision values and artificial intelligence prediction values to objectively analyze the trust changes between human-machine collaboration.And a human in loop experiment was organized to conduct experimental research on the impact of model interpretability on trust in human-machine hybrid systems and to obtain the rules that can explain the changes in trust between humans and machines under the influence of artificial intelligence technology.In summary,this article proposes a three-layer interpretable model based on the situational awareness model,which solves the problem of redundant interpretable information and develops an experimental platform integrating interpretable artificial intelligence technology as the research infrastructure.By using a combination of subjective and objective measurement methods,the experimental results have revealed the changes in human-machine trust under the influence of artificial intelligence,providing a theoretical basis and experimental evidence for solving trust problems between humans. |
