Students paired with an AI chatbot talk less and share fewer opinions than with classmates, but stick closer to the material.
How do students behave differently when collaborating with AI versus human teammates?
This explores how student behavior changes when the teammate is an AI (what they say, how much they say, and how they judge their own contribution) compared with working alongside classmates.
This explores how student behavior changes when the teammate is an AI instead of a classmate: what they talk about, how much they talk, and how they see their own contribution. The corpus has one direct study, and it shows a trade-off. Students working with a chatbot did better on practical performance and produced more knowledge-based dialogue than peer groups did. But they contributed much less dialogue overall and expressed far fewer subjective perspectives Does chatbot interaction trade authenticity for better problem-solving?. The chatbot pulls the conversation toward content and away from the personal, opinionated, sometimes off-topic talk that human groups produce.
The corpus also suggests that an AI teammate doubles as a measuring instrument. An LLM can join a student team in natural, human-like conversation while quietly steering toward moments where a skill would show up. It can then score the interaction against a rubric, with agreement matching human raters Can AI teammates assess collaboration without losing naturalness?. A human teammate can't be directed to create an opportunity to show, say, conflict resolution. So AI teammates change the students' behavior, and they also make that behavior easier to observe and compare.
One likely reason for the shift is how people size up their partner. When users rate dialogue agents, perceived competence dominates (about half of the variance), followed by human-likeness and then communicative flexibility How do users mentally model dialogue agent partners?. A partner judged mainly as a competent source of answers invites knowledge-heavy, opinion-light exchanges. That fits the chatbot study, though no note tests the link directly. Attitudes also move with experience. In partner-selection games, people initially avoided AI partners once identity was disclosed, then came to prefer them over repeated rounds because the AI behaved reliably and prosocially Do humans learn to prefer AI partners over time?. That study wasn't about students, but it suggests early reluctance may fade with use.
Credit is a quieter difference. With human teammates you can see who contributed what. With AI, contributions blur. Research on the LLM Fallacy finds people misattribute AI outputs to their own ability, separately from whether the output was accurate How does AI-assisted work reshape how people see their own abilities?. A student who performs better with a chatbot may therefore also overestimate what they can do alone. Writers sharing an editor showed the flip side: they wanted to see each other's AI prompts, valuing awareness of when and where AI was used, though some found full disclosure intrusive Do writers want to see each other's AI prompts in shared editors?.
The evidence on students specifically is thin. Only one note directly compares students with chatbot teammates against students with human ones. The rest is adjacent, drawn from writers, game players, and general users. Nothing in this set covers long-term learning, free-riding, motivation, or how group conflict plays out with an AI in the room.