| Summary: | The rapid integration of artificial intelligence in higher education positively affects student satisfaction, engagement, and learning outcomes. However, students frequently report ethical unease, guilt, and concerns about dependency. The current literature offers a limited explanation for their coexistence, as both have been treated as parallel or independent outcomes. Hence, this review extends and integrates existing theories by reconceptualising cognitive and moral dissonance as a central psychological process that explains how student satisfaction with AI-mediated learning is produced, negotiated, and sustained. Following PRISMA 2020 guidelines, we adopted a two-layer explanatory review design, synthesising 40 Scopus-indexed studies (Layer 1 = 15 studies; Layer 2 = 25 studies) from 2016 to 2025. Layer 1 studies explicitly define dissonance-related explanatory mechanisms that influence satisfaction and continued AI use across contexts such as dissertation writing, programming education, and problem-based learning. Layer 2 encompasses satisfaction-based studies that report ethical or affective concerns in parallel without theorising their interaction. The findings suggest a recurring satisfaction–dissonance paradox, in which students often experience genuine or conditional satisfaction from performance gains while simultaneously managing their psychological discomfort through one or more regulation mechanisms. Further, persistent and escalated dissonance leads to withdrawal or full or partial adaptive behaviour. We propose these dynamics as a testable Dual-Process Satisfaction–Dissonance Framework (DPSDF), which includes five dissonance triggers, five regulation strategies, three feedback loops, and four behavioural outcomes. Further, five domain experts’ suggestions have been taken to provide specific practical implications. This framework extends understanding of AI-mediated learning and provides foundations for future theory and policy development in higher education.
|
|---|
