Dear University of Southern Queensland,

We, the current students are writing to express our concerns regarding the university's use of Al detection software to accuse students of using Al in their written assignments. We believe this practice is flawed and unfair for the following reasons.

Flawed technological basis:

1.                The AI detection tool, Turnitin, claims a 98% accuracy rate in detecting Al generated content. However, this figure has not been independently verified. In fact, all research we have found demonstrates the true accuracy rate to be significantly lower. 

2.             According to Turnitin's own website, their software alone should not be used to accuse students of academic misconduct. This cautionary statement highlights the known limitations and potential for error in Al detection.

3.                Even OpenAl, the largest Al software company, acknowledges the difficulty in creating an accurate Al detector. If the leading experts in Al development cannot achieve reliable detection, it raises serious concerns over the efficacy of other Al detection software which are not being adequately addressed by the university.

4.                Al software is trained to write like a human, therefore Al detection software is trained to detect writing that, by design, is imitating human writing styles. 

5.             Al detection software disproportionately impacts non-native English speakers and neurodivergent students, leading to these potentially vulnerable students being unfairly targeted.

Opaque functionality and processes:

6.                How Al detection software works is never explained to students, making it impossible for students who are accused of using AI in written assignments to adjust their writing styles to avoid being falsely flagged for using AI. This manner of accusation has the consequence of reversing the onus of proof, requiring students to undertake the often difficult and time-consuming task of proving a negative – that misconduct didn’t take place – while also continuing their normal study workload. Some students have been falsely accused multiple times, and each time they must undertake the same mentally and emotionally taxing process to clear their name.

7.                The reliance on Al detection software undermines the trust between students and educators. Academic integrity should be based on clear expectations and upheld through fair, transparent and timely processes, not by relying solely on potentially flawed technological tools.

Impact on student welfare:

8.                False accusations of academic dishonesty are causing severe mental distress among students. Students are experiencing heightened anxiety whether they have received a mark of 0.02, a letter of accusation of academic misconduct, and even after they have been cleared of wrongdoing. The pressure and stigma associated with these accusations can have long term effects on students’ well-being and academic performance.

9.                As we ought to have learnt from the Robodebt scandal, using automated systems as the sole basis for making accusations of wrongdoing can lead to significant errors and needless human suffering. Technology should not replace human judgement, especially when that technology is so flawed and the consequences of getting it wrong are so profound for so many people.

It is our sincere hope that this letter highlights the serious shortcomings inherent in the existing academic integrity process, and will lead to changes which will benefit current and future students, improve communication and understanding between students and staff, and reaffirm the reputation of the University of Southern Queensland as a fair and compassionate institution.