Generative AI and Student Collaboration: A Scoping Review of Group Work Processes, Outcomes, and Risks

Synthesis: Wei and Perkins conducted a PRISMA-guided scoping review of 18 English-language studies (January 2023–March 2025) on how generative AI reshapes student group work in higher education, analyzing them using reflexive thematic analysis. Reported benefits include group knowledge development, idea generation, reflective thinking support, communication efficiency, task coordination, and feedback, while reported risks include reduced peer interaction and engagement under over-reliance, alongside privacy, transparency, and accuracy concerns. Outcomes appear to depend on task design, the degree of AI integration into group workflows, and student motivation and subject expertise.

Key Findings

1. 18 studies mapped across eight themes. The included studies spanned knowledge development, higher-order thinking, communication, team interaction, administrative efficiency, personalized feedback, data privacy/transparency, and bias/fairness/accuracy concerns.
2. GenAI reconfigures knowledge development. Learning shifts from an individual internal process toward a distributed, interactive one in which AI acts as facilitator, collaborator, and cognitive scaffold (e.g., within the zone of proximal development).
3. Higher-order thinking is both supported and constrained. AI can expand ideas, support reflection, and act as a "devil's advocate," but randomized evidence found AI produced more innovative suggestions without significantly boosting participants' overall innovativeness, and could homogenize ideas and invite cognitive offloading.
4. Communication becomes more efficient but can displace dialogue. AI reduced misunderstandings and coordination load, yet studies found group work became more efficient with reduced demand for communication, negotiation, and collective sensemaking (Lin et al., 2024).
5. Bias, fairness, and accuracy concerns are widespread. AI outputs can distort or over-generalize user ideas, hallucinate inaccurate information, and reinforce training-data biases that risk excluding marginalized voices.
6. The evidence base is uneven. Benefits (knowledge, communication efficiency, interaction) are more prominently and empirically supported; risks around privacy, transparency, bias, and accuracy are frequently acknowledged but largely discussed conceptually with limited empirical investigation.
7. Trade-offs are interconnected. Efficiency gains can reduce interaction, and AI-mediated communication can both enhance clarity and compress nuance—positioning GenAI as reshaping multiple dimensions of collaboration simultaneously rather than in isolation.
8. Key gaps remain. Much research is short-term or conceptual, with little longitudinal work on group dynamics, cognitive development, or the social-relational dimensions of group work (trust, community formation, peer support).

What this means for practice

• Instructors. Design group tasks that require human negotiation, since AI-mediated group work became more efficient with reduced demand for communication, negotiation, and collective sensemaking.
• Instructors. Assess process and meta-skills — collaborative problem solving, critical evaluation of AI output, AI literacy — rather than only the hybrid product, because conventional assessment poorly captures AI-enabled collaborative learning.
• Learners. Treat GenAI as a fallible collaborator: its suggestions can homogenize ideas, hallucinate inaccurate information, and reinforce training-data biases, so evaluate outputs before they enter the group's work.
• Faculty developers. Support the shift from content deliverer to facilitator, learning designer, and ethical steward, including the AI literacy and Learning Analytics interpretation that role now requires.
• Researchers. Close the empirical gap on risks, since privacy, transparency, bias, and accuracy concerns were frequently acknowledged in the 18 included studies but largely discussed conceptually.

Limitations

• Scoping review of 18 English-language studies published between January 2023 and March 2025, with a deliberately broad question and no quality appraisal; the authors state the lack of empirical studies makes specific, nuanced recommendations difficult.
• The included higher education contexts and disciplines vary (for example China and the US), so the authors caution that the implications may not apply elsewhere.
• One included study (He et al. 2024) was conducted with professionals rather than higher education students, and differences in participant characteristics and organizational settings may limit transferability to higher education.
• The evidence base is uneven: benefits such as knowledge development and communication efficiency are empirically supported, while risks around privacy, bias, and accuracy are largely discussed conceptually with limited empirical investigation.

Citation

Wei, Y., & Perkins, M. (2026). Generative AI and student collaboration: A scoping review of group work processes, outcomes, and risks. International Journal for Educational Integrity, 22(8).