Key points are not available for this paper at this time.
This study theorizes and tests language-based mechanisms through which conversations with chatbots can shape users’ loneliness. Lonely college students ( N = 96) engaged in emotionally meaningful conversations with the chatbot Replika daily for one week. Results showed a small but significant decrease in users’ loneliness from baseline to post-test. Linguistic analyses revealed two distinct well-being pathways. When participants used more positive emotion language, so did the chatbot, which was linked to higher positive daily affect and lower post-study loneliness. Conversely, when participants used more negative emotion language, so did the chatbot, which was linked to lower positive daily affect and higher post-study loneliness. Findings provide evidence for an expression effect , where users’ own positive (but not negative) linguistic framing shapes their well-being, and a chatbot-response effect , where chatbots mirror and amplify users’ emotional tone. This study theorizes chatbots as emotional mirrors with dual potential to alleviate or exacerbate users’ loneliness.
Choi et al. (Fri,) studied this question.