The Role of Reciprocity in Developing Trust in Human-Agent Interaction

Abstract

Introduction

This work examines the importance of reciprocity in the human-agent interactions framework (HAI). Within the social sciences, the term reciprocity refers to the practice of collaborating with others for mutual benefit. In this context, it refers to an agent's ability to reciprocate human actions and responses, which can enhance the perception of collaboration in HAI. According to previous studies, reciprocity is meaningful in building trust and promoting engagement [1]. In this study, we want to investigate the role of reciprocity in developing trustworthiness toward the agent in a collaborative drawing task. Collaborative drawing requires interacting with the lines and patterns of others and guessing their intentions to complete a picture together [2]. Therefore, we will conduct an experimental study to analyze the perception of trust toward an agent, testing two opposed attitudes: collaborative and individualistic. We assume that, during the collaborative task, participants rate the collaborative agent, namely the one showing reciprocity, as more trustworthy than the agent with individualistic behavior. The ultimate goal is to investigate the perception of trustworthiness in humans during interaction with agents, contributing to designing a trustable agent.

Methods

We will conduct two experiments. The first is a pilot within-subject study to classify and select object drawing categories pre-selected from the Google Sketch dataset (e.g. truck, face, flower, etc.) based on the users’ reported difficulty and time latency. In the main between-subjects experiment, participants will be assigned to different collaborative drawing tasks, where we will manipulate collaborative vs. individualistic agent behaviors to test the effect on perceived trustworthiness. After collaborative drawing, participants will rate the perceived, likability, enjoyment of the interaction, and legibility of the agent’s intention expression on a 7-point Likert scale. We will also use the Inclusion of Others in the Self (IOS) Scale to measure the perception of trust toward the agent and the Robotic Social Attributes Scale (RoSAS). In conclusion, we expect that participants will be more likely to perceive the collaborative agent as more trustworthy than the individualistic one, highlighting the importance of assessing collaborative systems in promoting trust and effective interactions.

References

[1] J. Zonca, A. Folsø, and A. Sciutti, “The role of reciprocity in human-robot social influence,” iScience, vol. 24, no. 12, p. 103424, Jun. 2021. doi:10.1016/j.isci.2021.103424

[2] T. Backer van Ommeren, H. M. Koot, A. M. Scheeren, and S. Begeer, “Reliability and validity of the interactive drawing test: A measure of reciprocity for children and adolescents with autism spectrum disorder,” Journal of Autism and Developmental Disorders, vol. 45, no. 7, pp. 1967–1977, 2015. doi:10.1007/s10803-014-2353-x