According to new research, people’s moral judgments about sexual assault on AI-powered sex robots are influenced by the robot victims’ mental capacities, interpersonal functions, ontological types, and the nature of their transactional relationships with humans. This work was published in Cognition.
Sexual assault is universally condemned, yet not all incidents are viewed equally. Research has shown that attributes of perpetrators and victims moderate moral judgments in these contexts. Characteristics like race, gender, career success, and even the framing of perpetrators as victims can influence public sympathy and moral evaluation of sexual assault incidents. In this work, Anastasiia D. Grigoreva and colleagues examined whether victim attributes, specifically perceived mental capacities and interpersonal functions, similarly affect moral evaluations of sexual assault.
In Study 1, a total of 790 participants were randomly assigned to one of three conditions, each presenting sex robots with different levels of human-like mental capacities: agency, experience, or mechanism. The participants were tasked with reading scenarios describing interactions between humans and sex robots that demonstrated these capacities. Following the scenarios, participants responded to measures assessing their moral judgments of sexual assault, specifically focusing on the blame and punishment deserved by the perpetrator, the vulnerability of the robot to harm, and the robot’s capacity for sexual consent, presented in random order.
The results indicated that participants attributed more blame and suggested greater punishment to perpetrators who assaulted sex robots described as having human-like mental capacities, either agentic or experiential, compared to robots described in mechanistic terms. This finding was mediated by the robots’ perceived vulnerability to harm and their capacity to consent, suggesting that these perceptions of mental capacities influenced moral judgments.
Study 2 involved 389 participants who were assigned to the same three conditions emphasizing different aspects of the robots. This study aimed to refine the stimuli from Study 1 by presenting a more integrated narrative in the form of a news article about a man and his sex robot to provide clearer distinctions between the conditions. After reading the news article, participants answered questions measuring the same variables as in Study 1, focusing on moral judgments related to the described sexual assault scenario.
Study 2 again found that robots with human-like qualities, either in terms of agency or experience, elicited more moral concern than those without such qualities. The influence of mental capacities on moral judgments was fully mediated by the robots’ perceived vulnerability to harm, highlighting the critical role this perception plays in moral evaluations of sexual assault.
In Study 3, 532 participants were assigned to evaluate either a “social” robot or a “sex” robot, examining whether the perceived function of the robot affects moral judgments of sexual assault. This study used a similar structure of scenarios and measurement as the previous studies but differed in manipulating the robot’s intended interpersonal function (i.e., social vs. sexual) rather than its mental capacities.
The interpersonal function of the robots—whether they were viewed as social or sexual partners—significantly affected moral judgments. Participants attributed less blame and punishment to the perpetrator when the robot was described as a sex robot compared to a social robot. This outcome suggests that the perceived primary function of robots can influence how their mistreatment is morally judged, with sexual function leading to less moral concern compared to social function.
Study 4 involved 393 participants and employed a factorial design to compare responses to sex robots and human sex workers, specifically examining how the context of the relationship—whether sex was purchased or not—impacted moral judgments. Participants read scenarios describing assaults under the four conditions and completed measures assessing their moral judgments, perceptions of the victim’s vulnerability, consent capacity, and overall mind attribution.
The results showed that sexual assaults in non-transactional contexts (i.e., where no sex was purchased) were judged more harshly than in transactional contexts, both for human and robot victims. Additionally, sexual assaults against human sex workers were generally judged more severely than those against sex robots. The transactional nature of the relationship had a more pronounced effect on judgments when the victim was a robot, suggesting that ownership and purchase conditions might reduce moral concern more significantly for robots than for humans.
A limitation noted by the authors is the hypothetical nature of the scenarios involving sex robots. Participants might not have taken these scenarios as seriously as real-life situations, which could affect the applicability of the findings to actual interactions with sex robots.
Overall, these studies suggest that moral judgments about sexual assault are influenced by factors like the victim’s mental capacities and the context of the act, highlighting the complex morality in human-robot interactions.
The research, “When does “no” mean no? Insights from sex robots”, was authored by Anastasiia D. Grigoreva, Joshua Rottman, and Arber Tasimi.