Economic behavior and emotion reports in trust game interactions with fellow humans and robots.

In Schniter, Shields, and Sznycer (2019) we focus our research on key issues relevant to the topic of trust-based interactions with robots (i.e., agentic forms of artificial intelligence (AI) and automation): (i) that people may not trust robots as they do humans, and (ii) that people may react to robots with different emotions. Using laboratory experiments that model trust-based interactions, we compare trust-based investment and emotions from across three nearly identical economic games: human-human trust games, human-robot trust games, and human-robot trust games where the robot decision also affects another human. In each of these trust games, a human trustor decides how much of a ten dollar endowment to entrust to a trustee (e.g. a fellow human, a robot). The experimenter multiplies the entrusted amount by three - creating potential gains from trust. Then the trustee receives this and decides how much to reciprocate to the trustor. We explain to participants who interact with them that the robots are programmed to mimic humans: they make automated reciprocity decisions based on previously observed behaviors by humans in analogous situations. After conclusions of the trust-game interactions, we provide participants feedback about their interactions and then participants rate how much they feel various positive and negative emotions. In this companion publication, we provide the raw data generated by participants (N = 397) in the role of "Investor" (“Person 1” in the participants’ instructions) and also "Trustee" (either “Person 2” or “the Robot”, depending on condition in the participants’ instructions) in our trust game experiment. This data is organized into two parts (1) the participants’ economic behavior in the trust games and their emotion reports from the post-game questionnaire, (2) a codebook explaining participants’ data.

在Schniter、Shields与Sznycer（2019）的研究中，本研究聚焦于与机器人（即具备自主代理能力的人工智能（AI）与自动化系统）的信任互动相关的核心议题：其一，人们或许不会像信任人类那样信任机器人；其二，人们对机器人的情绪反应可能存在差异。
本研究采用模拟信任互动的实验室实验，对三类近乎一致的经济博弈展开对比分析，分别为人际信任博弈、人-机信任博弈，以及机器人决策同时影响另一人类的人-机信任博弈。在每一类信任博弈中，人类信任者需决定将10美元初始禀赋中的多大比例委托给受托方（例如其他人类或机器人）。实验人员会将委托金额放大三倍，从而创造出信任带来的潜在收益。随后受托方接收该金额，并决定向信任者返还多少比例的资金。我们会向与之互动的参与者说明，机器人已被编程以模仿人类行为：它们会基于人类在类似场景中的既往观察行为，做出自动化的返还决策。
在信任博弈互动结束后，我们会向参与者提供本次互动的反馈，随后让参与者对自身感受到的各类正向与负向情绪的程度进行评分。在本配套研究出版物中，我们提供了参与者在本信任博弈实验中分别扮演"投资者"（参与者指导手册中记为"Person 1"）与"受托方"（根据实验条件，记为"Person 2"或"the Robot"）时所产生的原始数据，样本量N=397。该数据集分为两部分：（1）参与者在信任博弈中的经济行为表现，以及赛后问卷中的情绪报告；（2）用于解释参与者数据的编码手册。