Turnitin Can Now Detect AI Writing

A few Aussie universities are surprisingly *not* on board.

If you’re one of those people thinking you could get away with using AI to write your next assignment – think again.

Every student’s fave submission software ‘Turnitin’ has jumped on the AI hype train to combat potential cheating through its AI writing detection capabilities.

Turnitin’s writing indicator boasts 98% confidence based on data they have run through their own "AI Innovation Lab", which can report ‘the overall percentage of prose sentences contained in a long-form writing format within the submitted document that Turnitin’s model determines was generated by AI’.

The percentage that is generated from the AI writing detection model is separate from the similarity score that we’re all familiar with, however students will not be able to see the AI detection indicator.

Back in January, we discussed how unis were going to combat AI involvement in assignments, with the general consensus that going back to paper and pen might help this issue.

Since then, it seems that the response from unis are mixed.

The Royal Melbourne Institute of Technology told The Daily Aus that it has "opted against using the new [Turnitin] feature", and wants to embrace AI innovation rather than shying away from it.

The Sydney Morning Herald was told by Monash University that "their focus is on helping students understand how generative AI can be used in an ethical, responsible, and effective way in their learning and assesment", as well as their "current or future careers".

Don’t get too excited though – other Aussie unis like the University of Melbourne and the University of NSW are confirmed to be using it.

However, UNSW explained to The Sydney Morning Herald that "staff won’t be relying on it" and believe that the solution to dealing with ChatGPT is through working more closely with students, and improving "assessments in the senior years when there are smaller classes", and "lecturers are better acquainted with their students".

Is this new Turnitin feature actually any good though?

The Washington Post reported that in a test they ran with five high school students, one innocent student was flagged by Turnitin’s AI detector as a cheater. Oops.

Turnitin’s Chief Product Officer Annie Chechitelli even explains in a blog post that given their false positive rate is not zero, instructors need to use their "professional judgement", "knowledge of their students" and "specific context regarding the assignment".

It’s fascinating to see just how fast the discussion and use of AI has become, and how much it has already evolved in a matter of months.

Header Image: Turnitin.