Since it was released to the public in November 2022, the education sector – and organizations in general – have been obsessed with the generative artificial intelligence tool ChatGPT. Much of the talk around this ground-breaking chatbot application is about its ability to create seemingly ‘thoughtful’ text on virtually any subject.
Now, a new AI writing detection tool is being made available to the education sector from Turnitin, which will identify when students have used artificial intelligence software to write assignments. Teachers can then decide if they want to discuss this with the student.
Turnitin is developing such an anti-cheating tool, which will be the first to be introduced in the education market. The company has been producing Internet-based anti-plagiarism software for over two decades, and even the company name is a play on what its software enables – “turn-it-in.”
The Oakland, California-based company had an early jump over other software firms in this AI identification game. Turnitin has had an AI team on board for several years and began work on detecting AI-generated text two years ago. As it followed reports on the progress of ChatGPT at software firm OpenAI, Turnitin put its detection tool product development on the fast track.
Many educators are no doubt already familiar with Turnitin. The company’s software-as-a-service tool is found at approximately 16,000 schools and universities in 140 countries, serving more than 40 million students, according to Turnitin’s Vice President of Artificial Intelligence, Eric Wang.
The Turnitin subscription service enables educators to check submitted documents against its databases to detect plagiarism. The company has a similar product used by newspapers and publishing firms.
The method for detecting AI-generated text in any document might sound simple, but it is anything but simple. Using a proprietary algorithm, the Turnitin tool compares submitted papers to several databases – its own and those of other large academic proprietary databases it has licenses with. This process looks for cases where word choices and placement are blatantly ‘average’ – a sure-fire red flag, Wang explains.
An Internet-based artificial intelligence tool has conceivably all human knowledge at its fingertips. When an AI application is asked to write on a specific topic, it searches out the most common – or average – words to use in each case and the most common word placement order. The result is an absolute most common – or average – example of what a document on the topic would look like.
But that is not how humans think or write, Wang stresses. Nobody is that ‘average’ – at least not in their writing or communication. That is how the Turnitin tool perceives the difference.
“When you and I write, we pick words to convey an overall idea. We’re not picking words because they are the most probable or most likely ‘next’ word,” Wang explains. “We’re thinking in terms of sentences, paragraphs and chapters in the past, and then also about the future. The art of that is how we write. The AI tool doesn’t do that. It is just picking the next most probably used the word, word-by-word.”
The new Turnitin tool should find a very eager audience in the education sector, Wang believes. Teachers and faculty are clamouring for such a service. Many have been asking Turnitin to turn on the features early and even say they are willing to have a beta version. They desperately need something.
The reason is apparent. Educators have long complained about students using online search tools to research homework and class assignments. More recently, they have raised concerns that students would use AI tools to write assignments. In those cases, the text developed by an AI application likely hasn’t appeared before, so an anti-plagiarism tool would probably not detect it. That has helped create this present need.
Executives at Turnitin acknowledge that this beta release of its new AI detection tool may not be perfect. Still, it will give teachers much-needed insight and high confidence that what they are seeing has actually been run through generative AI. Wang says the Turnitin tool will also be constantly updated and improved to keep pace with the rate that the ChatGPT bot is learning.
Wang says he is not surprised at the public release of the ChatGPT tool. But he is astonished at how quickly OpenAI made it available to the public. Turnitin executives were expecting it to be another three or more years before a chatbot tool this capable would hit the market. They recently predicted that OpenAI’s product would revolutionize education.
But the impact of AI chatbots such as ChatGPT could eventually go far beyond education and ultimately impact many industries, Wang believes. For example, Wang foresees a time in the near future when an organization could eliminate the need to do substantial coding for various tasks and processes.
Instead of computer programmers writing or acquiring code to perform a specific task, they could merely tell the ChatGPT application to make it happen – and the app would search out the required code itself, Wang explains.“So instead of saying, ‘I want you to write me code that does this,’ I just say, ‘I want you to do it.’ That’s where we’re headed,” Wang says.
Meanwhile, “November 30, 2022 – when ChatGPT debuted to the public – was a lynchpin moment,” Wang says. “I immediately said, this will take the world by storm. We have to do something about it. This is going to create havoc for teachers.”
“For our two-year’s worth of work toward this, our product team really came together, and our engineering team came together. We are now at the state where we can launch the largest and the most influential AI writing detection capability in edtech,” Wang says.
With the release of its new AI detection capabilities, Turnitin executives don’t expect they have eliminated all worries about using AI chatbots in education. Such tools will never go away, Wang insists.
“When used well, they are incredible tools for learning and improving productivity,” Wang says. “The important thing is how you teach the use of these tools effectively so that the student can still think critically, reason, and argue effectively.”
Toward that end, educators need good, effective and safe detection tools to have visibility into when these AI programs are being used, and they can have the right conversations with students.