A systematic review of AI-powered collaborative learning in higher education: Trends and outcomes from the last decade

Synthesis: A systematic review of AI-powered collaborative learning in higher education: Trends and outcomes from the last decade

Key Findings

• A PRISMA 2020 systematic review covering 27 studies of AI-powered collaborative learning in higher education, drawn from a Scopus search (run July 16, 2024, query: TITLE-ABS-KEY "artificial intelligence" AND collaborative AND learning AND higher AND education) that identified 163 records → 145 screened after excluding 18 conference reviews → 59 retrieved → 27 included after eligibility assessment.
• The review organizes AI tools into four functional categories: predictive analytics/early warning systems, language-based systems (chatbots, large language models, natural language processing), recommender algorithms, and intelligent tutoring and monitoring applications — each contributing differently to collaborative learning dynamics.
• Evidence indicates AI-powered predictive analytics and multimodal approaches (including emotional and physiological monitoring) enhance student engagement and motivation, enabling early identification of at-risk students and timely, data-driven interventions (e.g., combining AI with learning analytics for continuous feedback improved collaborative learning performance and satisfaction).
• Personalized learning systems and recommender algorithms were found to underpin effective collaborative environments, while good task design (balancing challenge and Accessibility, embedding collaboration requirements) and emotional engagement and social presence emerged as critical success factors.
• The annual distribution of included articles shows a marked increase in publications over the decade, reflecting growing research interest in AI-enhanced collaborative learning in higher education.
• Key challenges identified include algorithmic bias, over-reliance on technology, and teacher training needs, alongside ethical concerns about transparency, data protection, and balancing full automation with human touch; the review notes AI's role in collaborative learning remains understudied relative to individual learning contexts.

About the Review

The review synthesizes a decade (2014–2024) of research on AI-enhanced collaborative learning in higher education. Inclusion criteria: articles published 2014–2024, written in English, explicitly focused on AI-based collaborative learning in higher education, with eligible designs limited to empirical studies, case studies, and meta-analyses, and populations of higher education students or instructors. Six research questions structured the synthesis, covering the impact of AI-powered multimodal approaches on engagement and motivation; the influence of task design on AI-powered collaborative environments; the role of affective factors and social presence; the effectiveness of remote learning and virtual laboratories; innovative teaching methods enabled by AI; and future directions. Studies were coded and tabulated by focus area, methodology, application, main findings, and AI methods, and grouped by intervention type, outcome measures, and population characteristics.

Key Results

• Academic performance: predictive analytics and machine learning support early identification of at-risk students and continuous feedback loops that improve collaborative performance and satisfaction; AI-driven feedback systems appear across the reviewed studies as a core intervention type.
• Engagement and motivation: multimodal AI approaches that track emotional and physiological responses (e.g., heart rate, facial expressions, electrodermal activity) give instructors real-time insight into engagement, supporting timely interventions to maintain motivation.
• Task design and assessment: effective tasks balance challenge and accessibility; AI can dynamically adjust task difficulty based on group performance, monitor participation, and enforce equitable contribution so dominant members cannot take over and reserved students are not left behind.
• Social presence and emotion: AI tools that increase social presence and emotional engagement make learners feel more connected to the learning process, improving collaborative outcomes.
• Remote and virtual learning: AI-powered virtual laboratories simulate physical lab environments, allowing distance students to collaborate on complex projects with engagement comparable to on-campus settings.
• Innovations and future directions: integration with blended/flipped classrooms, learning analytics, and emerging immersive technologies (metaverse, augmented and virtual reality) are highlighted, alongside calls to investigate underrepresented approaches such as symbolic AI and hybrid systems that merge reasoning-based methods.

What this means for practice

• Instructors. Invest in task design before tooling: the review names challenge calibrated to group ability and collaboration requirements embedded in the task structure as success factors, along with emotional engagement and social presence rather than cognitive outcomes alone.
• Instructors. Configure AI to protect equal participation — monitoring contribution and adjusting task difficulty as group performance changes — so that dominant members cannot take over and reserved students are not left behind.
• Administrators. Fund predictive and multimodal analytics for early disengagement detection (the reviewed systems use learning analytics and emotional or physiological signals for timely intervention), and fund the governance and teacher upskilling the same studies identify as unmet needs.
• Administrators. Treat bias, over-reliance, transparency, and data protection as deployment requirements rather than research caveats: they appear as recurring challenges across the decade of studies and as conditions for using AI in Collaborative Learning at all.
• Researchers. Prioritize what the review finds thin: AI's role in collaborative rather than individual learning is understudied, and group cohesion, equal participation, and effective collective collaboration are the open ground it names.

Limitations

• The review's scope is constrained by its stated eligibility criteria: a single database (Scopus), English-language articles only, and a 2014–2024 publication window, which may exclude relevant non-English or pre-2014 work and non-indexed venues.
• The synthesis draws on heterogeneous study designs and outcome measures, which limits direct comparability of effect sizes across interventions.
• The review itself notes the field's open research gaps — AI's role in collaborative (versus individual) learning is understudied, and future work should address group cohesion, equal participation, and effective collective collaboration.

Citation

Kovari, A. (2025). A systematic review of AI-powered collaborative learning in higher education: Trends and outcomes from the last decade.