Integrating Large Language Models into College-Based Higher Education: A Practitioner Action Research Study on AI-Mediated Adaptive Teaching

Abstract

Large language models bring both real opportunities and genuine challenges to college-based higher education in the UK, where learner cohorts are remarkably diverse and call for teaching that actually adapts to their needs. Over two cycles from October 2025 to March 2026, three different AI modes were implemented at a further education college in North West England with 12 students on a Level 4 Computing programme. These modes - AI-Tutor (personalised tutoring), AI-Student (peer-based learning), and AI-Simulator (scenario-based practice) - were drawn from the Mollick and Mollick (2023) framework. Data came from a practitioner reflective journal, student feedback questionnaires, and classroom observations. What emerged was clear: students with lower prior attainment gained most from AI-Tutor for understanding core concepts, while higher-attainers went deeper with AI-Simulator for applied problem-solving. Several patterns cut across all three modes: reduced anxiety when asking for help, the critical importance of scaffolding students into AI literacy, and the value of matching modes to learner characteristics. A practical framework for other lecturers working in this context is proposed. This research addresses a real gap - college-based HE remains largely invisible in the AI-in-education literature - and shows how practitioner-led inquiry can produce insights that actually matter for how we teach in these settings. The study has limitations (single practitioner, self-reported data), but the implications extend to policy and practice across college-based HE.

Keywords: Action research; artificial intelligence; college-based higher education; adaptive teaching; large language models; practitioner inquiry

DOI: 10.7176/JEP/17-5-04

Publication date: May 30^th 2026