Human-robot cooperation in economic games: People show strong reciprocity but conditional prosociality toward robots

Abstract

<p>Understanding human social interactions with robots is important for designing robots for social tasks. Here, we investigate undergraduate participants’ situational cooperation tendencies towards a robot opponent using prisoner’s dilemma games. With two conditions where incentives for cooperative decisions were manipulated to be high or low, we predicted that people would cooperate more often with the robot in high-incentive conditions. Our results showed incentive structure did not predict human cooperation overall, but did influence cooperation in early rounds, where participants cooperated significantly more in the high-incentive condition. Exploratory analyses revealed other two behavioural tendencies: (1) participants played a tit-for-tat strategy against the robot (whose decisions were random); and (2) participants only behaved prosocially toward the robot when they had achieved high scores themselves. Our findings highlight ways in which social behaviour toward robots might differ from social behaviour toward humans, and inform future work on human–robot interactions in collaborative contexts.</p>

Individuals Expend More Effort to Compete Against Robots Than Humans After Observing Competitive Human–Robot Interactions

“…For examples, Mutlu et al (2006) found that people had more positive attitudes towards the ASIMO robot in a cooperative game context than in a competitive game context [33]. Even when researchers did not intend to frame the robot as competitive, participants can be sensitive to robot's noncooperative decisions and responded to these reciprocally [34]. Therefore, future research on this topic may choose to make more judicious decisions when designing a HRI context and manipulating an agent's intention, to ensure an agent's attribute in third-party encounters is not in contrast with how the agents behave in the actual HRI.…”

Individuals expend more effort to compete against robots than humans after observing competitive human-robot interactions

“…For examples, Mutlu et al (2006) found that people had more positive attitudes towards the ASIMO robot in a cooperative game context than in a competitive game context [33]. Even when researchers did not intend to frame the robot as competitive, participants can be sensitive to robot's noncooperative decisions and responded to these reciprocally [34]. Therefore, future research on this topic may choose to make more judicious decisions when designing a HRI context and manipulating an agent's intention, to ensure an agent's attribute in third-party encounters is not in contrast with how the agents behave in the actual HRI.…”

Tracking Human Interactions with a Commercially-available Robot over Multiple Days: A Tutorial

“…Above and beyond the validation of the animations that are displayed by the Cozmo robot during the interaction with the human user, these data can be used to select and display specific animations. We have successfully used this in studies employing one-off interaction games, by engineering emotions exhibited by Cozmo during various games manipulations (Hsieh et al, 2020a(Hsieh et al, , 2020b.…”

“…It has been successfully used in laboratories, schools, and public events, and with children and adult samples. Researchers have used it in experimental studies to measure perception and behaviour during one-off interactions (Abubshait et al, 2020;Chatterji et al, 2019;Ciardo et al, 2020;Currie & Wiese, 2019;Druga et al, 2017Druga et al, , 2018Gonnot et al, 2019;Hinz et al, 2019;Hsieh et al, 2020aHsieh et al, , 2020bJoshi & Šabanović, 2019;Kumar Singh et al, 2020;Lefkeli et al, 2020;Roselli et al, 2019;Tan et al, 2018) and unrestricted and repeated interactions outside of the laboratory (Cross, Riddoch, et al, 2019;Pelikan et al, 2020). Together, these studies have investigated a variety of human behaviours, from empathy to emotion perception, and from social decision making and learning to the sense of agency in the human user.…”