Generative artificial intelligence in higher education: A systematic review of student use and learning outcomes

Abstract

Generative artificial intelligence (GenAI) is increasingly used in higher education, yet evidence remains fragmented on how students use it in learning tasks and how these uses relate to learning outcomes. This systematic literature review of 39 peer-reviewed empirical articles outlines seven ways higher education students use GenAI for learning and examines the resulting actual and perceived learning outcomes.  Results indicate that students’ actual learning outcomes were predominantly successful, while perceived outcomes vary. Specifically, using GenAI as a creator resulted in an approximately equal proportion of challenges and successes in learning effectiveness, learning efficiency, interactivity, self-regulation and personalised learning. In contrast, when students used GenAI as a translator, refiner, navigator, evaluator or dialoguer, they perceived higher challenge-to-success ratios. Notably, using GenAI as a self-regulatory supporter resulted in the lowest challenge-to-success ratio, with the few challenges attributed to insufficient integrated self-regulated learning skills and prompt strategy. Results suggest that students need to strengthen their integrated self-regulated learning skills to optimise GenAI for learning. Teachers and institutions must address these challenges by providing ready-to-use prompts or prompt training and bound the use of GenAI as a creator with the use of it as a self-regulatory supporter, academic-integrity guardrails and preservation of author voice.

Implications for practice or policy:

• Educators and instructional designers can improve student learning outcomes by designing curricula and pedagogical interventions that account for different ways of GenAI use and uneven learning outcomes.
• Institutions should develop policies and guidance to address the uneven learning outcomes associated with students’ use of GenAI, as well as data protection for certain uses.
• Policymakers need to consider different ways of GenAI use when allocating funding, shaping regulations and strengthening ethical oversight mechanisms.