Psychology Today published an August 1, 2026 article by Rónan Fulton, Alana McCarthy, and Michael Hogan arguing that generative artificial intelligence helps collaborative learning only when teams use it with trust, verification, feedback, and reflection. The article cites three recent empirical studies involving more than 480 learners in higher education and says AI-supported collaboration depends less on raw technical capability than on how people communicate, coordinate, and trust one another while using the tools.

Luo and colleagues (2025) surveyed 308 university students in 66 collaborative groups to study how trust, collective efficacy, and task conflict shaped team creativity when generative AI was introduced. The study measured trust through students' views of whether group members could discuss difficulties openly, approach tasks professionally, and rely on one another to meet responsibilities. It measured task conflict through disagreements about work organization and whether AI-generated suggestions or human judgment should guide decisions.

The Luo et al. findings described AI as double-edged for collaboration. Students who saw AI as intelligent developed stronger collective efficacy, or shared confidence in a group’s ability to succeed, and that confidence promoted team creativity. AI also increased task conflict, and those disagreements reduced creative performance. Groups with stronger interpersonal trust were better able to convert AI-supported collaboration into creative outcomes because trust strengthened the relationship between AI use and collective efficacy.

Lehtinen and colleagues (2026) observed 75 pre-service teachers working with generative AI to design lesson plans, using video recordings and process mining (tracking activity sequences from data). Higher-performing groups drafted lesson-plan sections, searched outside sources to verify AI-generated information, and revised their work after checking evidence. They consulted online resources more than twice as often as lower-performing groups. Lower-performing groups repeatedly re-prompted ChatGPT after weak answers, copied AI-generated text directly into lesson plans, and edited wording afterward instead of independently verifying content.

Gyasi and colleagues (2025) experimentally compared three approaches to human-AI collaboration in online collaborative learning. One condition gave groups AI feedback and feedforward, including personalized summaries after each task, guidance before the next activity, visualizations of how much discussion stayed on topic, and written feedback. A second condition placed an AI chatbot inside group discussions as an additional participant that contributed ideas, asked questions, gave explanations, and helped keep conversations focused. A third condition combined both approaches and produced the strongest results, including higher collaborative knowledge building, cognitive engagement, socially shared regulation, and overall group performance than either intervention alone or no AI support.

The article also cites Xu et al. (2026), which found that shared AI use worked differently from private individual AI use in collaborative learning. When students collectively prompted, evaluated, and revised ChatGPT responses, AI became a shared object for discussion that supported mutual awareness and negotiation. When students used AI privately before group meetings, the article says opportunities for collective evaluation and shared decision-making were reduced.

Psychology Today published an August 1, 2026 article by Rónan Fulton, Alana McCarthy, and Michael Hogan arguing that generative artificial intelligence helps collaborative learning only when teams use it with trust, verification, feedback, and reflection. The article cites three recent empirical studies involving more than 480 learners in higher education and says AI-supported collaboration depends less on raw technical capability than on how people communicate, coordinate, and trust one another while using the tools.

Luo and colleagues (2025) surveyed 308 university students in 66 collaborative groups to study how trust, collective efficacy, and task conflict shaped team creativity when generative AI was introduced. The study measured trust through students' views of whether group members could discuss difficulties openly, approach tasks professionally, and rely on one another to meet responsibilities. It measured task conflict through disagreements about work organization and whether AI-generated suggestions or human judgment should guide decisions.

The Luo et al. findings described AI as double-edged for collaboration. Students who saw AI as intelligent developed stronger collective efficacy, or shared confidence in a group’s ability to succeed, and that confidence promoted team creativity. AI also increased task conflict, and those disagreements reduced creative performance. Groups with stronger interpersonal trust were better able to convert AI-supported collaboration into creative outcomes because trust strengthened the relationship between AI use and collective efficacy.

Lehtinen and colleagues (2026) observed 75 pre-service teachers working with generative AI to design lesson plans, using video recordings and process mining (tracking activity sequences from data). Higher-performing groups drafted lesson-plan sections, searched outside sources to verify AI-generated information, and revised their work after checking evidence. They consulted online resources more than twice as often as lower-performing groups. Lower-performing groups repeatedly re-prompted ChatGPT after weak answers, copied AI-generated text directly into lesson plans, and edited wording afterward instead of independently verifying content.

Gyasi and colleagues (2025) experimentally compared three approaches to human-AI collaboration in online collaborative learning. One condition gave groups AI feedback and feedforward, including personalized summaries after each task, guidance before the next activity, visualizations of how much discussion stayed on topic, and written feedback. A second condition placed an AI chatbot inside group discussions as an additional participant that contributed ideas, asked questions, gave explanations, and helped keep conversations focused. A third condition combined both approaches and produced the strongest results, including higher collaborative knowledge building, cognitive engagement, socially shared regulation, and overall group performance than either intervention alone or no AI support.

The article also cites Xu et al. (2026), which found that shared AI use worked differently from private individual AI use in collaborative learning. When students collectively prompted, evaluated, and revised ChatGPT responses, AI became a shared object for discussion that supported mutual awareness and negotiation. When students used AI privately before group meetings, the article says opportunities for collective evaluation and shared decision-making were reduced.