随着信息化时代的到来，人们在工作、学习和生活中不可避免地需要与智能系统合作完成任务。为了最大限度开发人机协作系统的潜能，需要在交互过程中建立适当的信任水平。以往对人际间交往的研究表明，急性应激会影响基于信任的决策，但对人机信任的影响研究相对较少，主要集中在实验室场景下的任务，且对神经生理相关表现关注不足。本文通过两项研究探索了急性应激对人机信任的影响，并选取大型传染病疫情场景中不同类型的信任决策，从行为、心理和生理等多个层面分析急性应激的影响机制。研究一考察了急性应激和时间压力对人机信任的影响。采用经典信任博弈范式，探索与人工智能系统合作时不同应激源和时间压力对参试者决策的影响。研究二聚焦实际应用，分析疫情场景下不同应激水平对参试者与辅助决策的人工智能系统间信任的影响。两项研究都在实验过程中同步采集了行为、心理、生理、神经等多类数据。本文研究结果如下：第一，急性应激会降低信任博弈中对人工智能系统的信任程度。与正常状态相比，应激状态会导致决策行为上更低的投资金额以及主观上更低的信任评分，但不同急性应激源间无显著差异，且未发现应激状态对决策时间的影响。第二，时间压力也会导致行为和主观上更低的信任程度且对急性应激的影响有调节作用；无时间压力时，社会评估压力的存在会进一步降低信任水平，投资轮次对投资金额也存在显著影响，第一轮显著高于第三轮。第三，疫情背景下决策过程中急性应激水平与人机信任水平变化呈负相关趋势。无论从人工智能系统推荐采纳率还是参试者主观信任评分来看，应激状态都对应更低的信任水平，且随着应激水平的升高，信任水平随之降低。第四，应激处理和信任决策刺激会引起前额叶皮层脑血氧水平的变化，应激状态下背外侧前额叶皮质两侧均有较大活跃区域，且应激状态脑区活跃程度较正常状态更高。本文揭示了急性应激对人机信任影响的变化规律，基于信任决策行为、主观信任意愿、皮肤电水平、心率和大脑血氧水平等多维指标，丰富了人机信任相关的综合研究。同时，本文还验证了与实际应用紧密结合的疫情场景下应激处理任务与信任决策任务范式的有效性，为未来基于信任的人工智能辅助决策相关的研究与应用推广提供了新的思路。

With the coming of information age, people inevitably need to cooperate with intelligent systems to complete tasks in their work, study and life. In order to maximize the potential of human-computer cooperation system, it is necessary to establish an appropriate level of trust in the interaction process. Previous studies on interpersonal communication have shown that acute stress can affect trust based decision-making, but there are relatively few studies on the impact of human-computer trust, mainly focusing on tasks in laboratory scenarios, and paying insufficient attention to the performance related to neurophysiology. This paper explores the impact of acute stress on human-computer trust through two studies, and selects different types of trust decisions in large-scale infectious disease epidemic scenarios to analyze the impact mechanism of acute stress from behavioral, psychological and physiological aspects.The first study examined the impact of acute stress and time pressure on human-computer trust. The classical trust game paradigm was used to explore the influence of different stressors and time pressures on participants‘ decision-making when cooperating with artificial intelligence systems. The second study focused on practical application, and explored the influence of different stress levels on the trust of participants and decision support AI systems in the epidemic situation. Both studies simultaneously collected behavioral, psychological, physiological, neurological and other types of data during the experiment.The results of this paper are as follows. First, acute stress will reduce the trust level of AI system in trust game. Compared with the normal state, stress state will lead to lower investment amount and lower subjective trust score in decision-making behavior, but there is no significant difference between different acute stressors, and no impact of stress state on decision-making time is found. Second, time pressure can also lead to lower trust in behavior and subjectivity, and has a moderating effect on the impact of acute stress; When there is no time pressure, the existence of social evaluation pressure will further reduce the level of trust, and investment rounds also have a significant impact on the amount of investment. The first round is significantly higher than the third round. Third, there is a negative correlation between the level of acute stress and the level of man-machine trust in the decision-making process under the epidemic situation. Whether from the AI system recommendation adoption rate or the subjective trust score of the participants, the stress state corresponds to a lower trust level, and with the increase of the stress level, the trust level decreases. Fourth, stress processing and trust decision-making stimulation will cause changes in cerebral blood oxygen levels in the prefrontal cortex. Under stress, there are large active regions on both sides of the dorsolateral prefrontal cortex, and the brain regions in stress state are more active than those in normal state.This paper reveals the changing law of the impact of acute stress on human-computer trust. Based on such multidimensional indicators as trust decision-making behavior, subjective trust willingness, skin electricity level, heart rate and cerebral blood oxygen level, it enriches the comprehensive research related to human-computer trust. At the same time, this paper also verified the effectiveness of the stress processing task and trust decision-making task paradigms in the epidemic situation closely combined with practical applications, providing a new idea for the future research and application promotion of trust-based AI aided decision-making.