Generative AI in Higher Education: A Narrative Review of Global Adoption, Ethical Challenges, and Behavioural Pathways To Critical Thinking, With Implications for African and Ghanaian Contexts
TL;DR
It is found that GenAI adoption is high globally, and is growing rapidly across Africa, but that structural constraints, including limited infrastructure, low AI literacy, and underdeveloped institutional policy, shape a distinctly African pattern of opportunity and risk that global adoption figures obscure.
Abstract
Generative artificial intelligence (GenAI) has moved from novelty to routine presence in higher education within a few years, yet the evidence on its adoption, ethical implications, and cognitive effects remains fragmented across global, regional, and mechanism-focused literatures that are rarely read together. This narrative review synthesises 43 sources, comprising global systematic reviews and meta-analyses, comparative Global North-South studies, African and Tanzanian policy analyses, Ghana-specific empirical studies, and behavioural models linking GenAI use to critical thinking and ethical creativity. The review traces the evolution of this literature across three waves from 2022 to 2026 and finds that GenAI adoption is high globally, at roughly 74.5% among educators, and is growing rapidly across Africa, but that structural constraints, including limited infrastructure, low AI literacy, and underdeveloped institutional policy, shape a distinctly African pattern of opportunity and risk that global adoption figures obscure. In Ghana specifically, a large tertiary-institution survey shows technical infrastructure as a significant constraint even where urban-rural gaps are not, alongside evidence that roughly a third of surveyed undergraduates report low AI literacy despite substantial GenAI use. Critically, the review identifies that the empirically tested behavioural pathways linking GenAI use to critical thinking and ethical creativity, through self-regulated learning, innovative behaviour, and reduced reliance on prior knowledge, have been tested only in Taiwanese and Chinese settings, leaving open whether the same mechanisms hold under African infrastructural and institutional conditions. The review concludes with practical implications for institutions and policymakers and a concrete future-research agenda centred on this gap, arguing that testing these pathways in Ghana is the field's single most valuable next study.