Objective: The purpose of this work was to examine the relationship between self-concept and ingroup/outgroup categorization of robots. Background: Social psychological literature can improve Human–Robot Interaction (HRI) through investigations about cultural differences, intergroup dynamics, and more. Parallel to human–human interaction, people categorize robots as ingroup (“my group”) or outgroup (“not my group”) based on a myriad of variables. They favor ingroup robots by viewing them as more positive and human-like versus outgroup robots or humans. Previous work has examined the effect of robot anthropomorphism (i.e., human-likeness) on this categorization process with diverse findings. Method: Examining the self-concept via the two-baskets theory of self-cognitions, we compared the ingroup categorization of Humans, Machine-like robots, Medium Human-like robots, and High Human-like robots using a simple categorization task. Results: Results indicate that robots and human are categorized to the ingroup correlating with the uncanny valley effect, with humans being most likely to the ingroup followed by Medium Human-like, Machine-like, and High Human-like. Conclusions: Self-concept may not be as important for categorization as other factors; however, important categorization differences exist following the trend of the uncanny valley. Application: Those who design and utilize robots should take categorization differences into consideration when designing robots for public interactions. Further, those who purchase robots for use should be careful to consider the implications of visual similarities to human beings to ensure optimal acceptance OSF: ( https://osf.io/x9rqn/?view_only=fd0894404e304422a6c77ccffa013bcd ). Precis: Considering the uncanny valley, we examine the social categorization of humans and robots and the relation of this categorization to the self-concept in the form of self-cognitions. Results demonstrate important differences in the categorization of various types of human and robot images based strongly on the uncanny valley effect.

Involving users in the design of novel technologies, for instance, by adapting them to personal preferences, can improve attitudes and trust towards them. Research on the customization of robot functionalities is still scarce, with existing work predominantly focusing on the impact of perceived, rather than actual user involvement in the design process. Furthermore, it is unclear whether the effects of customization depend on merely choosing options, or on actively implementing these options. Thus, in the present online experiment, we investigated whether actual time invested in choosing options (i.e., words that a robot would learn) and implementing them (i.e., defining the words to a robot in the individual’s own terms) influenced psychological ownership and trust towards the robot. Contrary to our hypotheses, time invested in choosing options slightly decreased perceived control and psychological ownership over the robot. This might be due to task difficulty, which hints at the role of perceived self-efficacy as a determinant of psychological ownership. No effect of time spent in implementing the chosen words by defining them has been found.

As robots increasingly assist more people, tendencies of becoming attached to these robots and treat them well have risen; even to the point of treating robot teammates better than human opponents in laboratory settings. We examined how far this ingroup favoritism extends and how to mitigate it. We did this by making participants play an online game in teams of two humans and two robots against two humans and two robots. After the game, they selected someone to perform an additional unpleasant task (according to the results of our pilot test); we manipulated that task to be equally unpleasant for ingroup and outgroup members in one condition, and more unpleasant for outgroup than for ingroup members in the other condition. We did this to examine if the moral principle of utilitarianism (i.e., social justice and fairness) would outweigh ingroup favoritism. In the results, participants showed typical group dynamics like ingroup favoritism. The opportunity to behave in a utilitarian way failed to reverse the ingroup favoritism effect. Interestingly, participants sacrificed their ingroup robot more than they sacrificed even outgroup players. We speculate about why the study showed these unexpected findings and what it may mean for HRI.

Research on robot groups has often applied psychological principles underlying group processes between humans to interactions with and between robots. However, such research has failed to test empirically whether these principles indeed apply to the robot context. For instance, the notion of a social group may be interpreted differently when facing human versus robot groups. Basic research on this issue is missing. Therefore, the present experiment aimed at integrating social psychological theorizing and research on robot groups by utilizing the principles of group entitativity. We examined the effect of robot number and similarity on the perception of these robots as a (social) group. To do so, participants saw pictures of one to ten robots, appearing low or high in similarity. Results showed that the aspects eliciting the perception of a social “group” in humans seem to differ from the factors evoking robot group perception. According to our findings, at least three robots seem necessary for the perception of a robot “group” to emerge. Social psychological research, however, has proposed that two persons suffice to elicit the notion of a human social group. Basic research is needed to substantiate assumptions drawn from social psychological theorizing before translating it into human-robot context.

The act of revealing personal information, thoughts, and feelings is known as self-disclosure. Self-disclosure represents an important determinant of liking and is central to the development of close relationships among humans. The present study aimed to investigate the role of self-disclosure in human-robot interaction (HRI). 81 participants were randomly assigned to one of four experimental conditions in which they interacted with the humanoid robot NAO. We manipulated whether the robot disclosed personal information or whether the robot asked personal questions to the human interaction partner, so that the participant had to self-disclose. In two control conditions the robot either made factual statements or the robot asked factual questions. Contrary to the hypotheses, the results indicated no immediate statistically significant effects of self-disclosure on the dependent variables robot likability, human-robot interaction quality, future contact intentions, and mind attribution. However, when taking into account participants' tendency to anthropomorphize technology, nature, and the animal world as a covariate, self-disclosure was found to significantly affect participants' tendency to attribute mind to NAO. Furthermore, the results indicate that the form of engagement and dominance in an interaction (i.e., being in the role of a passive listener vs. conversing actively) may affect perceived HRI more than does the content of the verbal exchange. Thus, the paper highlights the importance of considering covariates (i.e., interindividual differences in the tendency to anthropomorphize nonhuman entities) in HRI analyses and points out possible relevant moderators of self-disclosure in HRI.

Although it is widely accepted that robots will be used in everyday contexts in near future, many people feel anxious and hold negative attitudes toward robots. This negative reaction might be stronger when users come into direct physical contact with them, particularly when touch is required between robots and humans, (e.g., when using robots as assistants to help elderly people at home). Intergroup contact research in social psychology has proposed various forms of contact as a means to reduce negative feelings toward outgroup members. The present study examined how Contact Type (Actual vs. Imagined) and Contact Modality (Look vs. Touch) with a NAO robot would impact attitudes toward NAO compared to a no-contact control condition. Results showed that nearly any type of contact effectively reduced negative emotions compared to the control condition. However, for participants with preexisting negative emotions toward robots, contact sometimes produced more negative attitudes. We discuss these findings and the resulting implications for future research.

Negative attitudes towards robots are common and pose an obstacle for successful and pleasant human-robot interaction. To reduce negative attitudes in the context of social robotics, we draw upon social psychological methods of attitude change. Imagined contact represents one such paradigm which to date has mostly been tested and validated in the human-human intergroup context. In the present experiment, we have therefore examined the effectiveness of imagined contact (i.e., the mere mental simulation of contact with a human or nonhuman target) in changing robot-related attitudes, robot anxiety, contact intentions, and psychological anthropomorphism. To do so, participants had to briefly imagine a restaurant scenario as detailed and lively as possible. Crucially, we manipulated the content of the imagined contact scenario as follows: In the control conditions, participants imagined an interaction with a human target or a technical device. In the experimental condition, in contrast, participants imagined an interaction with a robot target. We predicted that imagined contact with a robot (versus a human target versus a technical device) would result in more positive attitudes towards the robot, stronger contact intentions, and higher psychological anthropomorphism whereas robot anxiety should decrease. Contrary to our predictions, however, we found that participants who had imagined contact with a human target reported more positive attitudes and higher contact intentions with a robot prototype than participants who had imagined contact with a technical device. Furthermore, imagined contact with a robot had no effect on the dependent measures. We interpret these findings in light of potential ceiling effects, fluency effects, and the activation of elicited agent knowledge.