Portal de Periódicos da CAPES

Dancing With Myself: The effect of majority group size on perceptions of majority and minority robot group members

2013; Wiley; Volume: 35; Issue: 35 Linguagem: Inglês

1551-6709

Henny Admoni, Bradley Hayes, David Feil-Seifer, Daniel Ullman, Brian Scassellati,

Death Anxiety and Social Exclusion

Dancing With Myself: The effect of majority group size on perceptions of majority and minority robot group members Henny Admoni, Bradley Hayes, David Feil-Seifer, Daniel Ullman, and Brian Scassellati Department of Computer Science, Yale University 51 Prospect Street, New Haven, CT USA Abstract While social psychology has identified characteristics of in- tergroup dynamics, few studies have looked into the percep- tions of robot group dynamics. In this experiment, we sep- arate robots into majority and minority groups based solely on their behavior in a simple dance routine. We attempt to understand how people’s perceptions of robots within those groups change based on group size and features of behavior. Participants viewed the robot dances and rated one robot from each group on a variety of characteristics. We find that be- ing from the minority versus majority group has a significant impact on perceptions of a robot’s creativity, interestingness, anti-sociality, dancing ability, and how much of a team player it is. At the same time, individual behaviors (leading the dance, following the dance, or performing an entirely unique dance) have no statistical effect on participants’ ratings of robot char- acteristics. From these results, we conclude that group size has a larger effect than behavior on subjective evaluations of robots in majority and minority groups. Keywords: Group dynamics; majority group; minority group; intergroup relations; robotics; human-robot interaction Introduction The tendency to categorize people into established groups is an automatic social behavior, and influences much of how we perceive the world. Psychologists have studied the dynam- ics of intergroup relations, particularly when the group sizes are unequal, for many decades (Tajfel, 2010). Among many other findings, the literature reveals that minority groups— those with fewer members—are susceptible to being influ- enced by a unanimous majority group (Asch, 1956), even when the majority’s assertions are incorrect on a point of fact. Furthermore, people are biased toward perceiving their own group (an in-group) with more positive characteristics than a group of “others” (an out-group) (Tajfel, 1974). In this paper, we pursue a systematic, empirical analysis of how simple manipulations of group dynamics can affect the perception of group members’ personal qualities, some of which are not directly related to the behaviors being demon- strated. To accomplish this, we analyze peoples’ judgments of a group of simple robotic agents in terms of characteris- tics such as interestingness, creativity, and sociality. We vary the agents’ behaviors in systematic ways to tease apart which elements of their behavior affect these character judgments. In particular, we compare peoples’ perceptions of a majority group robot—one member of a group of robots all exhibiting nearly identical behaviors—to their perceptions of a minority group robot—a single robot exhibiting a distinct set of behav- iors different from the majority group. In the last century, Michotte investigated the appearance of animacy by systematically varying the behavior of simple moving shapes (Michotte, 1963). The current investigation is inspired by this and, in some sense, continues Michotte’s work by investigating the characteristics of groups of agents by systematically varying the behavior of the group. Through this work, we hope to contribute a greater understanding of social attributions and the dynamics of groups. Another benefit of this work is a greater understanding of the perception of robot groups. As technology improves, groups of humans are being joined by increasingly agentic technologies such as medicine-delivery robots in hospitals and package-retrieval robots in warehouses. Research in the field of human-robot interaction (HRI) has begun to examine intergroup relations in terms of including or excluding robots from human in-groups. Studies have shown that people re- spond more favorably to robots in their in-group than one in an out-group (Eyssel & Kuchenbrandt, 2011; Kim, Kwak, & Kim, 2010; Kuchenbrandt, Eyssel, Bobinger, & Neufeld, 2011; Wang, Rau, Evers, Robinson, & Hinds, 2009). How- ever, much of this research investigates a single robot inter- acting with one or more humans. This paradigm is histor- ically reasonable, because social robots have been typically designed for and deployed in single-robot environments. As robots become less expensive and more socially accepted, however, groups of robots may become more common. In these cases, understanding the dynamics of robot groups will be important for robot designers and users. Few studies have explored intergroup dynamics for groups comprised exclusively of robots. Most importantly, little re- search exists that addresses the perception of groups of robots that are distinguished solely based on the behavior of the group members. We are interested in how the behavioral cate- gorization of robots into majority and minority groups affects perceptions of group members’ characteristics. In this paper, we describe an experiment that attempts to identify the effect of majority group size on people’s charac- terizations of majority group and minority group robots. We use a basic robot behavior—a simple dance—as the distin- guishing feature between groups. In our experimental ma- nipulation, we vary both majority group size (one, three, or seven robots in the majority group versus a single-robot mi- nority group) and the type of dance performed by the minor- ity group robot (same as or different from the majority group dance). We ask participants to rate the robots on a number of characteristics both related to and unrelated to dancing, in an attempt to understand how groups distinguished only by behavior are perceived. Our hypotheses are: H1 The minority group robot will be rated more highly in in-