Speaker
Abstract section 4: Outcomes/results
Of the 157 assignments, 28 were flagged for review and 19 were confirmed to contain AI‑generated texts. All confirmed cases were resolved through a structured rewrite opportunity, allowing students to demonstrate authorship while receiving a penalty proportional to the amount of AI-generated text. The results highlight two key outcomes. First, triangulation seems to reduce the risk of false positive errors, or human-written text flagged as AI-generated. Second, the process supported learning by redirecting confirmed cases toward revision rather than punitive measures. The study demonstrates that triangulation can uphold academic integrity while maintaining fairness, due process, and pedagogical value.
Abstract section 2: Contribution/research questions
The purpose of this study is to explore whether a triangulation approach – consisting of 1) cross-referencing of multiple AI-detectors; 2) comparison of submitted written assignments with previously-collected benchmarks, or authentic examples of each student’s writing; and 3) instructor judgment – could improve the detection of AI-generated text in student papers in the academic writing courses of an English-medium graduate school in Japan.
Abstract section 3: Content/method
Seventy-one graduate students completed a series of controlled writing tasks without access to digital tools to serve as benchmarks. These students later submitted a total of 157 course-related writing assignments that were scanned for AI-generated text using multiple detectors. When both the instructors and detectors identified probable AI‑generated text in their work, students were invited to a confirmation meeting where they could explain their writing process, present drafts, or acknowledge AI involvement.
Abstract section 5: References
Copyleaks. (2025). AI in Action: Students have fully normalized AI in the classroom. 2025 AI in Education Trends Report. https://copyleaks.com/wp-content/uploads/2025/09/2025-AI-in-Education-Trends-Report_1.pdf
Digital Education Council. (2024, August 7). What students want: Key results from DEC Global AI Student Survey 2024. https://www.digitaleducationcouncil.com/post/what-students-want-key-results-from-dec-global-ai-student-survey-2024
Elkhatat, A. M., Elsaid, K., & Almeer, S. (2023). Evaluating the efficacy of AI content detection tools in differentiating between human and AI-generated text. International Journal for Educational Integrity, 19, 17. https://doi.org/10.1007/s40979-023-00140-5
Weber-Wulff, D., Anohina-Naumeca, A., Bjelobaba, S., Foltýnek, T., Guerrero-Dib, J., Popoola, O., Sigut, P. & Waddington, L. (2023). Testing of detection tools for AI-generated text. International Journal for Educational Integrity, 19(1), 1-39. https://doi.org/10.1007/s40979-023-00146-z
Abstract section 1: Relevance
The use of generative AI has become ubiquitous in tertiary education, with around ninety percent of university students saying that they have used AI for academic purposes, and more than half reporting that they use AI at least weekly (Copyleaks, 2025; Digital Education Council, 2024). In order to deter inappropriate AI use, many instructors and institutions rely on AI detectors such as Turnitin to scan student texts for AI-generated text. However, research has found that these tools have questionable accuracy and reliability (Elkhatat et al., 2023; Weber-Wulff et al., 2023).
| Title | Can triangulation improve AI-generated text detection? |
|---|---|
| Teaching Context | College and university education |