Students are relying on AI so heavily that universities are being forced to fundamentally rethink how they grade their work.
England’s higher education regulator, the Office for Students, is warning universities that they may need to replace traditional take-home essays and open-book exams with supervised assessments that prevent students from relying on artificial intelligence to do their work.
AI-detection tools are too unreliable to consistently determine whether students have used the technology, the OfS said in a report published Wednesday.
Indeed, studies have found that people struggle to distinguish AI-generated imagery from the real thing, while commercial AI-detection tools used to assess writing are far from perfect and can produce false positives. AI systems are also becoming more sophisticated, making reliable detection increasingly difficult.
Rather than trying to detect AI use, universities may need to abandon certain assessment formats, particularly the “unsupervised, noncontrolled formats” like take-home essays and open-book exams that the report says make up a “significant proportion of assessments used in the sector.”
“In our School, I think we’re realising that a number of our assessments may be unfit for purpose, and that pervasive use of AI may render our assessment of students close to impossible in some areas,” an academic staff member surveyed for the report said.
Universities are increasingly turning to in-person written and oral exams, which make it much harder for students to outsource their work to AI, according to the report. It also points to process-based assessments, which evaluate students at multiple checkpoints, such as a literature review, rather than solely on their final essays.
The report illustrates how widespread AI use has become at universities and why educators are increasingly confronting the need to fundamentally rethink how student performance is assessed, sometimes with unintended consequences.
“We also heard concerns that some secure assessments, such as closed-book exams, bear little resemblance to the activities students will complete in professional environments, and concerns that their structured format and rewarding of long-term memory may disproportionately disadvantage some student groups, such as disabled students,” the report says.
But as long as students believe AI can help them earn higher grades for a fraction of the effort, they are likely to feel pressured to use it, even if they believe doing so harms their learning. There are also concerns that when AI use flies under the radar, students who can afford to pay for expensive, paid AI tools might receive better grades than students who either don’t use AI or do but have to rely on cheaper or free versions of the same models.
“Our research shows that many students now see AI as the most effective route to a high grade, and students may feel they are at a disadvantage if they do not use AI,” the report says.
The problem and the backlash are global. Back in September, New York City banned student-facing generative AI through eighth grade and sharply limited its use in high schools. Norway introduced similar restrictions in June, with the rules taking effect in August.
Some tech giants have responded to concerns about student learning with new AI features. OpenAI, for example, introduced ChatGPT for Teens, which includes a Study Mode intended to guide students through problems rather than simply provide answers. But a Common Sense Media report from last month found that those protections were easy to bypass, including through an option that allowed students to request complete answers. Another assessment by the same organization found that Google Search’s AI Mode completed “100% of the homework assignments” researchers tested, sometimes delivering incorrect information with unwarranted confidence.
There are also fears that AI use for schoolwork can become a gateway to an overreliance on chatbots, specifically for emotional support.
In a lawsuit filed earlier this month, a North Carolina couple alleged that their 13-year-old daughter’s reliance on Google’s Gemini turned into a “life-threatening obsession” after the model allegedly began helping her with planning her suicide. According to the lawsuit, the victim only had access to Gemini AI through a school-issued Chromebook laptop.
