Cheating with ChatGPT and generative AI: why isn’t it possible with TestWe?

Using ChatGPT or any other AI tool during an exam is cheating! The rise of generative artificial intelligence poses new challenges for teachers: how do you prevent students from consulting an AI during a remote test? TestWe, the online exam specialist, has integrated anti-cheating functions optimised to block and/or detect any use of artificial intelligence. How does this work? Arina Munteanu, Quality Manager at TestWe, explains.

The TestWe intelligent assistant blocks the use of ChatGPT and AIs

A downloadable application to prevent cheating with ChatGPT and generative AI

The principle is simple: TestWe is a downloadable application, not a solution opened in a web browser. ‘The application is automatically displayed in full screen mode, and blocks both keyboard shortcuts and screenshots. This prevents access to unauthorised external resources, such as artificial intelligence’, explains Arina Munteanu. With such an operating mode, it’s impossible to hope to get out of the exam to browse another site like ChatGPT. ‘Everything is encrypted. Nor can the candidate extract the TestWe subject and copy it somewhere else.

And also: continuous monitoring during the exam

Students might have the idea of bypassing this protection by using the smartphone version of ChatGPT. But that’s not possible either: the webcam surveillance function takes over. ‘The TestWe application goes round the room before the test starts and checks that the candidate is not wearing headphones. Monitoring is then carried out continuously throughout the test, via the webcam and the computer microphone’. Under these conditions, students cannot attempt to ask someone for help, or use any other device connected to an AI. Except at their own risk.

TestWe AI also helps detect cheating attempts with ChatGPT or other AIs

Putting AIs to work for teachers

‘Our solution prevents the use of AI during online exams, but we didn’t want to completely shun this technology,’ stresses Arina Munteanu. Indeed, with TestWe, artificial intelligence is made available to teachers. The TestWe AI function is a virtual assistant based on ChatGPT. It can not only create exams, but also help to mark them and spot cheating attempts.

Testwe AI, a correction assistant

‘The artificial intelligence correction assistant gives an initial mark according to the scale set by the teacher, and offers personalised comments. The tool is also capable of detecting the use of another AI in the candidate’s answers’ (assuming that the candidate has managed to bypass the anti-cheating systems).

A scalable solution to adapt to new digital and AI uses

Since its creation, TestWe has continued to evolve with digital usage and to offer new functions. In the beginning, it was mainly a question of securing face-to-face tests and finding solutions to the digital divide. Since then, the Covid-19 pandemic, the growing use of distance learning and the development of AI have all played their part. ‘This has given even greater importance to the anti-cheating aspect of TestWe’, adds Arina Munteanu.

As a result, new ideas could be deployed in the future to improve TestWe: strengthening the virtual marking assistant and analysing typing speed, for example, always with the aim of guaranteeing the quality and reliability of online exams.