Academic Integrity

Academic Integrity or academic misconduct – it’s something that’s discussed a lot in Higher Education (and indeed at other levels); we’re no different in Dundee. It’s also something that I have been involved with for a long time. In my early days as a Lecturer at Portsmouth we used, for a small number of assessments, what’s now Turnitin. Then, it was the “JISC Plagiarism detection Service” It was very much advertised as a plagiarism detection. Often, students would submit two copies of work, one, on paper for marking, one digital, in case staff had any concerns. Few checks existed to ensure it was the same work (though checking was easier in days when coursework wasn’t anonymous). I, and a number of others, realised that this new tool had other benefits; marking online, and if we wanted to be very radical, allowing students to see their originality report before the final submission. 

Moving forward perhaps a decade. Many staff and students were still seeing Turnitin as the Plagiarism Police; rather than a potential study support tool. As similarity detection grew, so did students trying to find ways round it. Often, this was due to difficulties really understanding how to paraphrase – so tools like ArticleReWriter were created; others tried uploading to translation tools, and then translating back to English. Both of these tended to produce very garbled English, making it easy for staff to see that something wasn’t right – even if Turnitin failed to detect similarity. 

Meanwhile, over the same decade Contract Cheating (Ghost writing), first described by  Clarke & Lancaster () in relation to code, is now a major concern globally, in all subject areas – as any quick search will show. 

Today, staff concerns still include students having difficulty paraphrasing, and it’s what they’re identifying most frequently. Contract cheating remains a major concern, though staff acknowledge the difficulty identifying it. Most recently, though, software, such as OpenAI has been identified as potential issues. OpenAi and others use linguistic data to generate realistic looking written content. This isn’t new – Stribling, Aguayo and Krohn (MIT) used a similar approach when they generated SciGen over 15 years ago. It’s now used in the workplace, posing significant dilemmas to Higher Education (and education in general). Just to add to the mix, I spotted that Meta (Facebook) have now created a GPT-3 (the engine behind OpenAI) alternative – an alternative that’s OpenSource.

I’ve just spotted that OpenAI’s DALL-E 2 can generate art from descriptions, so I’m off to have a play.