Issues register

EdTech Assessment Toolkit

Generative AI Detector (Lancaster University)

Automated scoring model

Turnitin’s AI Similarity Report emphasises text that may have been generated by AI tools. Turnitin purports to have crafted this tool to aid educators in upholding academic integrity and ensuring equitable treatment for students. The incorporation of Turnitin’s AI writing indicator into the Similarity Report offers insights into AI-produced content within submitted documents. The AI writing indicator displays an overall percentage denoting potential AI-generated text, drawing on Turnitin’s AI model. Language models like ChatGPT are trained using extensive internet text data, rendering them predictable in producing sequences of words based on high probabilities. Conversely, human writing is more varied and distinctive, leading to reduced word prediction probabilities. Turnitin’s classifiers are constructed to discern these variances and excel in pinpointing the word probability patterns of human authors (refer to Turnitin, 2023).

Currently, the precision of AI detectors remains ambiguous, and false positives are notably common (Dalalah & Dalalah, 2023). This is further corroborated by recent controversies, notably the significant number of unfounded allegations against students at Lancaster University (Jordan, 2023). The deployment of these models is also critiqued as a manifestation of technological solutionism, wherein challenges associated with technology are addressed with further technological solutions, carrying significant environmental repercussions (Stokel-Walker, 2023).

Snapshot (July 2023)

System task/function: Detect the use of generative AI in assessments
Model: Automated Scoring Model
Deployment: Generative AI detection
Location of application: Various locations, see e.g. Lancaster University (UK)
Rationale for introduction: Combat academic misconduct facilitated by generative AI
Vendor: Turnitin
Pricing: 3$ US per student annually
Data and computation: Language processing, historical essay data
Inequalities/harms: high number of false positives, especially targeting ESL students
Status: Active
Authority/regulation: University level
Unintended consequences: Previous assignments could be labeled as generated by AI / new AI tools can be developed to counteract a detecting tool.
Sanction/redress: N/A

References/further reading

Dalalah, D., & Dalalah, O. M. A. (2023). The false positives and false negatives of generative AI detection tools in education and academic research: The case of ChatGPT. The International Journal of Management Education, 21(2), 100822. doi:

Turnitin. (2023). AI Writing Detection Capabilities – Frequently Asked Questions. Retrieved from

Stokel-Walker, C. (2023). Turns out there’s another problem with AI – its environmental toll. Retrieved from

Jordan, A. (2023). Lancaster SU calls on uni to stop using Turnitin over fears it falsely detects ChatGPT use Retrieved from