ChatGPT and Generative AI in Higher Education: Implications for Academic Integrity and Pedagogy
Master's Thesis · ~94 pages · English
Abstract
This thesis investigates the transformative impact of ChatGPT and other large language models on higher education, examining both the threats to academic integrity and the opportunities for pedagogical innovation. Through a mixed-methods study combining content analysis of 200 AI-generated essays across four disciplines, survey data from 350 faculty members, and institutional policy review from 40 universities, the research evaluates the effectiveness of AI detection tools, the adequacy of current academic integrity frameworks, and emerging pedagogical approaches that integrate generative AI as a learning tool. Findings indicate that AI detection tools produce unacceptably high false positive rates, that prohibition-based policies are unsustainable, and that assessment redesign emphasizing process-based evaluation and authentic tasks represents the most promising institutional response.
1. Introduction
The release of ChatGPT in November 2022 triggered an unprecedented crisis in higher education, challenging fundamental assumptions about assessment, authorship, and the purpose of writing assignments. Within months, universities worldwide issued emergency policies ranging from outright bans to cautious integration, often with minimal empirical basis.
This thesis provides systematic evidence to inform institutional responses to generative AI. The central argument is that effective responses require understanding both what AI can and cannot do in academic contexts, and how assessment practices can evolve to remain meaningful in an AI-augmented environment.
2. Detection and Integrity Challenges
Content analysis of AI-generated vs. human-written essays reveals:
Detection Tool Accuracy - Leading AI detection tools (GPTZero, Turnitin AI) achieve 70-85% true positive rates but with 8-15% false positive rates, disproportionately flagging non-native English speakers.
Output Quality - GPT-4 generated essays scored within the B to B+ range across disciplines, with strongest performance in structured analytical tasks and weakest in creative and experiential writing.
Prompt Engineering Effects - Students who used iterative prompting and editing produced work that was substantially harder to detect and higher quality than direct copy-paste submissions.
Faculty Survey Findings - 68% of faculty reported suspecting AI-generated submissions but only 23% felt confident in their ability to identify them without technological assistance.
3. Pedagogical Integration Framework
The thesis proposes a four-tier framework for AI integration in assessment:
Tier 1: AI-Prohibited Tasks - In-class handwritten exams, oral examinations, and supervised practical demonstrations where AI use is neither feasible nor pedagogically appropriate.
Tier 2: AI-Transparent Tasks - Assignments where students document AI use, reflect on AI contributions, and demonstrate critical evaluation of AI-generated content.
Tier 3: AI-Collaborative Tasks - Projects requiring students to use AI tools effectively as part of the learning objective, assessing prompt engineering and output refinement skills.
Tier 4: AI-Evaluated Tasks - Student work assessed partly by AI tools with faculty oversight, enabling personalized feedback at scale.
The framework emphasizes process-based assessment, requiring documented drafts, revision histories, and reflective metacognitive components.
References
- [1]Kasneci, E., Sessler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., ... & Kasneci, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences, 103, 102274.
- [2]Cotton, D. R. E., Cotton, P. A., & Shipway, J. R. (2024). Chatting and cheating: Ensuring academic integrity in the era of ChatGPT. Innovations in Education and Teaching International, 61(2), 228-239.
- [3]Stokel-Walker, C., & Van Noorden, R. (2023). What ChatGPT and generative AI mean for science. Nature, 614(7947), 214-216.
- [4]Mollick, E. R., & Mollick, L. (2023). Using AI to implement effective teaching strategies in classrooms: Five strategies, including prompts. The Wharton School Research Paper.
This is a sample excerpt. Full papers include complete chapters, verified citations, and downloadable formats.
Free to try · No credit card required · Free to start, 3 credits/day