Torturing the algorithm: Everyday sadism predicts acceptance of anthropomorphic AI

Abstract

Previous research has identified personality antecedents of attitudes toward AI technologies. However, little attention has been paid to the relationship between dark personalities and acceptance of specific emerging technologies, such as anthropomorphic AI. According to the Computer as Social Actor theory, individuals tend to interact with computers as if they were social actors, applying social norms and emotional responses typically reserved for human interactions. Based on this framework, we theorize that high-sadism individuals will exhibit greater acceptance of AI with human-like features than those with low levels of sadism, as they may derive pleasure from inflicting harm on anthropomorphic AI in a manner similar to their interactions with human partners. In Study 1, we found that self-reported sadism positively predicted acceptance of anthropomorphic AI among Chinese university students. Using a behavioral measure of sadism and a more diverse population, Study 2 provided further evidence for the hypothesized relationship between everyday sadism and acceptance of anthropomorphic AI. Study 3 ruled out alternative explanations, which demonstrates that sadism was specifically associated with acceptance of anthropomorphic AI rather than with AI in general. These findings underscore the importance of considering personality traits when exploring the societal integration of AI.