Abstract:
Generative AI (GenAI) is rapidly transforming study practices; however, Sri Lankan universities
provide only limited, discipline-specific guidance on its appropriate use, creating uncertainty
for both students and lecturers. This study focuses on second-year undergraduates enrolled
in courses in the Department of English Language Teaching (DELT), Faculty of Social
Sciences and Languages, Sabaragamuwa University of Sri Lanka. It examines students’ use
of GenAI in English-language learning, their ethical views on its role in learning and assessment,
and its impact on student–lecturer relationships. However, there is limited qualitative
evidence from the Sri Lankan context on this topic. A qualitative case study was conducted
using semi-structured interviews with 12 gender-balanced students (recruited via purposive and
snowball sampling) and brief naturalistic observations. Data were analysed using reflexive
thematic analysis. UTAUT2 is employed as the theoretical framework because its constructsperformance/
effort expectancies, social influence, facilitating conditions, hedonic motivation,
and habit-align directly with the research questions. Findings indicate students widely use
ChatGPT to understand grammar and meaning, unpack assignment tasks, and brainstorm. Participants
reported peers’ unethical AI use, strong social influence (peer norms), ambiguous
facilitating conditions (unclear rules, uneven detection), and evidence of hedonic use and habit.
Students drew a shifting ethical line between legitimate scaffolding and prohibited substitution,
sometimes crossing it under deadline pressure and perceived norms. Many expressed strong
trust in AI, both as a study assistant and a “friend.” All participants believed lecturers were aware
of AI tools, yet none had encountered formal, course-level rules on acceptable use. About 60%
perceived reduced interaction with lecturers due to AI, while others reported no change. The
study recommends faculty-level, department-level AI policies that are customised disciplinary
level, transparent AI-use statements with submissions, process-oriented assessment, and targeted
AI-literacy workshops. The findings contribute Sri Lanka-specific qualitative evidence
and extend UTAUT2 by foregrounding trust and transparency within facilitating conditions.
