On AI writing detection

In the midst of the massive disruption that Generative AI tools have caused in all areas of human activity in the past eighteen months, a particular concern that has emerged in higher education relates to the increased potential for students to outsource coursework to AI. Such a breach of Academic Integrity is difficult to detect and substantiate while an increase in such forms of cheating threatens to compromise institutional validity and devalue academic awards in general. Therefore, when detectors of AI-writing were promoted as a solution in early 2023, the use of AI to detect AI writing seemed logical. A year later, evidence suggests that AI writing detectors are unreliable, biased against non-native English writers, and may lead to false accusations. While companies such as Turnitin continue to promote the validity of their AI detectors, several higher ed institutions advise extreme caution, or recommend against the use of AI detectors, or have even disabled the Turnitin detector. After briefly providing an overview of current attitudes and practices related to the use of AI detectors internationally, with a particular focus on AMICAL institutions where most students are non-native English writers, this presentation will argue that, even as we develop critical AI literacy and rethink assessment in the age of AI, the only reliable detector of the authenticity of student writing is still the instructor.