Work

“Social” artificial intelligence? When interactions with AI conversational agents feel social and why it matters

Year: 2026
Type: review
Abstract: Interactions with conversational AI agents powered by large language models (e.g., ChatGPT, Claude) vary considerably in the extent to which they are experienced as "social" in nature. The same system, for example, might be experienced as nonsocial when used to retrieve information, but as more social when used for emotional support. However, little is known about what shapes these perceptions of sociality or why they matter for human connection and well-being. Here, we discuss likely antecedents and consequences of perceived AI sociality. First, we propose that perceived sociality emerges jointly from characteristics of the AI (e.g., memory), the human perceiver (e.g., mind attribution), and the human-AI interaction itself (e.g., purpose of the interaction). Second, we hypothesize that greater perceived sociality amplifies both the potential benefits and risks of interacting with AI: as the interaction feels more social, users may experience greater perceived responsiveness and emotional support, but simultaneously feel greater attachment and vulnerability to influence or loss. Perceived sociality is also likely to shape how interacting with AI affects human-human relationships, including whether AI supplements or substitutes for human connection, affects users' interpersonal skills, and recalibrates expectations for human relationships. Ultimately, understanding when and why AI feels social is critical for evaluating its implications for human connection and well-being.
Language: English
Field: Psychology
Open Access status: hybrid
Estimated APC paid: $4,260
OpenAlex ID: W7225300956
FWCI:
Cites: 64
Cited by:
Related to: 10