Turnitin, a plagiarism detection service used by the University of Edinburgh, now has built-in tools to recognise content generated by ChatGPT and other generative AI software.
In a press release issued by the company last month, they say that their new tools have a 98 per cent success rate at recognising content generated by AI.
Educational institutions that use the company’s services do not have to pay to obtain these new capabilities, as they have been added to existing Turnitin products.
The University of Edinburgh’s policy on generative AI software, issued in March, does not ban its use outright.
Instead, it seeks to allow its use in limited ways, while expecting submitted work to contain a student’s original ideas.
Where assignments do not explicitly allow students to use generative AI software, it encourages students to contact lecturers to ask for permission or clarity.
According to the policy, uncited or unpermitted use of generative AI software counts as a breach of the university’s Academic Misconduct Procedures.
Generating an entire piece of coursework with generative AI software has also been disallowed by the university.
The university says that use of generative AI should be cited as “personal communication” in your chosen referencing style.
Further, use of AI in essay planning – even if AI generated content does not reach the final draft of the assignment – should be cited.
In March, the School of History, Classics and Archaeology (HCA) gave its students more detailed guidelines on how generative AI software can be used in coursework.
HCA did not explicitly ban use of generative AI software, instead telling students to contact course organisers on a case-by-case basis.
The school’s guidance says that all use of AI in essay planning and preparation should be cited, and how the software was used should be explained.
ChatGPT and other tools like it build text by predicting what comes next in a series of words by use of machine learning.
Users of generative AI software provide the program being used with a prompt, which the software attempts to provide a response to.
The output of generative AI programs can range in complexity, from short answers to simple questions to long bodies of text.
Entire essay-length bodies of text can be generated with such software, which is freely available and used by some students for various academic and non-academic tasks.
However, text generated by current AI tools can include factual inaccuracies, invented details, and text copied without permission from other sources.
Image via Joe Sullivan