Artificial Intelligence in Education: Transforming Learning Outcomes, Academic Integrity, and Pedagogical Innovation in Higher Education

Abstract

The transformative opportunities and challenges of generative AI tools, especially large language models (LLMs) such as ChatGPT, have emerged rapidly in higher education. The adoption of generative AI in undergraduate education has increased dramatically in 2024–25, whereas institutions' pedagogical approaches, institutional policies, and evidence of learning outcomes are less developed than their use. Previous research has primarily focused on the short-term acceptance of the tools or single academic integrity issues without considering them in relation to each other. Recent systematic reviews and meta-analyses (2025–2026) have begun to measure the impact of GenAI on learning outcomes and trace authorship and integrity concerns in greater detail. However, few studies have systematically investigated the impact of GenAI on learning outcomes, pedagogical design, and academic integrity across a variety of learner populations. This study will explore (1) the observable effects of AI-supported personalised learning on student learning outcomes at the undergraduate level; (2) faculty and student attitudes towards the impact of AI on pedagogical redesign; and (3) institutional policies to ensure academic integrity while supporting and facilitating AI-supported learning. The proposed convergent mixed methods design will involve validation of the survey instrument (n = 420 undergraduate and faculty students across three institutions) and a quasi-experimental pre-post assessment study. Structural equation modelling (SEM) and thematic coding of qualitative data were used in the planned analysis. This study aims to produce a theoretically sound, testable framework that will benefit evidence-based strategies for the adoption of AI in higher education, the AI-Augmented Pedagogy Integration Model (AAPIM). This document outlines the conceptual, theoretical, and methodological underpinning, which has now been further supported by a growing evidence base of 2025–2026 meta-analyses and systematic reviews, with empirical data reported when the data collection is complete