By Jaspreet Bindra
Last year I completed a Masters in AI and Ethics from Cambridge University. In my final term in late 2023, I wrote a dissertation, and as is customary, submitted it through a plagiarism detection tool, Turnitin.
While this was a practice followed year after year, this particular year there was some apprehension amongst the faculty and some students. ChatGPT had been launched a few months back and had taken the world by storm. Among the first people who woke up to its impact were academia and universities; they worried about how students could use this very powerful tool to write essays and term papers, and yes, dissertations.
Pierag Consulting LLP aims to achieve $50 million revenue target by FY27, says Abhishek Gupta after company rebranding
Apple pushes back against age restriction legislation: What this means for App Store policies
‘Over 1 lakh tonnes cargo capacity’: Noida Airport CEO on multimodal hub and future developments
BrandWagon Exclusive: How AI self service platforms are redefining customer support
Some universities responded by shutting off their Wifi, blocking ChatGPT, or even resorting to pen-and-paper exams. Cambridge University did not do that, but it did worry about whether tools like Turnitin could detect AI-generated content, which was plagiarism of a different kind.
While universities worry about AI being used by students and the lack of original content, therefore, the interesting thing is that Generative AI can also be used to detect plagiarism – one of the biggest issues that universities today face.
There have been innumerable instances where prominent professors and Deans have been tarnished by this, the latest being the President of Harvard University.
So, it’s a bit of a cat-and-mouse game. On the one hand, students are using advanced Gen AI tools to create essays and assignments, raising concerns about authenticity. On the other hand, universities are leveraging similar AI technologies to detect these AI-generated submissions.
It is similar to deepfakes in a way – the same technology used to make them can also be used to detect them. These AI systems are not just searching for copied text. They’re getting smarter, analyzing writing styles and patterns to tease out content that doesn’t quite match a student’s previous work – a sort of literary Sherlock Holmes.
But as AI gets better at detecting its handiwork, it also gets better at evading detection. It’s an endless arms race, much like the virus-antivirus arms race we see in software. Progressive universities are not just leveraging these tools passively; they’re actively engaging with tech firms and partners, and the creators of such tools to constantly update and refine their processes to keep a step ahead of plagiarism.
The aforementioned Turnitin is one example. Advance Publications, the company that owns this tool, has developed a new version of Turnitin which can detect text generated by AI, in response to the increasing use of AI tools by students for completing assignments.
Even OpenAI developed such a tool but with modest success. Prominent universities have adopted this new tool, one of them being Purdue University. They have integrated this into its Learning Management System (LMS) and it scans and evaluates submissions for AI-generated content.
Having said that, it is pertinent to note that Purdue and other universities advise caution in relying solely on these results, as there can be false positives. The first such tool designed by OpenAI, for example, could confidently detect AI-generated content or plagiarism only 26% of the time!
Another interesting example is the University of Adelaide, which conducted tests on Turnitin’s AI detection tool. They realized that while the tool successfully identified AI-generated content in straightforward cases, it struggled when the text was altered to appear more human-like, such as being rewritten in the style of a teenage student.
The University also tested other freely available online AI detection tools, with varying levels of success. Among these, Copyleaks emerged as a relatively reliable tool, showing a high probability of detecting AI-generated content even after human modification.
These are early days, and it is still difficult to use only technology to detect plagiarism. Human intervention and oversight remain significantly important; these tools, at best, can help human evaluators. However, these examples illustrate the strenuous efforts in academia to maintain integrity in the face of rapidly advancing AI technologies.
Besides technology, we will need regulation and policy at the university level, as well as more heightened awareness among faculty and students. The future of academia is not just about avoiding plagiarism; it’s about fostering an environment where original thought and AI coexist, each enhancing the value of the other.
(Author is Managing Director & Founder, The Tech Whisperer)
About The Author
