Just Like a Human? An Experimental Study on University Students’ Perceptions of Warmth and Benevolence Towards Chatbots

Abstract: Human–computer interaction has advanced significantly with the emergence of OpenAI’s ChatGPT. However, it’s unclear whether users perceive these chatbots to have positive human qualities. Using a 2 × 2 experimental design, university students evaluated the perceived warmth and benevolence of a chatbot, whereby we altered the description of the chatbot across experimental groups. Specifically, we altered its description according to Asch’s (1946) concept of warmth, where we varied it to have traits suggesting high or low kindness, and we also varied the language used in the chatbot description (technical vs. anthropomorphic). We hypothesized that describing the chatbot using anthropomorphic (humanlike language) vs. technical language would increase its perceived warmth and benevolence and that descriptions highlighting kindness would enhance the chatbot’s perceived warmth. Results revealed that kindness-related descriptions of the chatbot significantly affected its perceived warmth and benevolence, whereas differences in anthropomorphic or technical language did not. The role of personality traits in shaping AI perceptions is discussed.