The impact of competence and warmth on reactions to human and AI interviewers in a hiring context: the role of interaction quality in human versus AI preferences

Artificial Intelligence (AI) is increasingly used for interactions. While studies show that people prefer human over AI interactions, it is unclear whether this preference holds regardless of the comparative quality of these interactions, where quality is understood in terms of varying levels of warmth and competence. We addressed this using an experimental vignettes study supplemented by open-ended questions. Participants (N = 1067) were subject to a 2 (human or AI interviewer) X 2 (high or low competence) X 2 (high or low warmth) factorial design and asked to rate and explain their perceptions of the interaction. We found that human interactions led to better outcomes on our proxies for interaction quality, compared to AI interactions, and that humans were seen as both warmer and more competent. However, we also found that higher quality AI interactions sometimes led to better interaction quality measures compared to lower quality human interactions. Moreover, we found that while participants were sensitive to variations in both warmth and competence, warmth was the more important factor. These findings have important implications for the ethical deployment of AI in workplaces, given their impact on how satisfied and respected people feel during their interactions with AI.