Reimagining university assessments for the age of AI

The sun has set over London after 8pm for the last time until April 2025 and when walking the familiar crackle and crunch of leaves can be heard. The pumpkin spice lattes won’t be far behind. Back to school/university/Parliament season is upon us.

Wherever one is returning to this September, AI will play a role. Nowhere is this role more unrecognised than in higher education assessment.

The UK higher education sector is hugely important. It contributes more than ￡100 billion to the economy, supports more than 750,000 jobs and expands the UK’s soft power. Almost 700,000 students come to the UK to study at our universities and three of the top ten global universities are UK-based. Nevertheless, a plethora of challenges face our universities.

Financially, the freezing of domestic tuition fees and the reduction in teaching grants have meant universities face some difficult choices. More than 60 universities have announced budget and job cuts and, by the end of 2023-24, 40 per cent of English publicly funded providers are expected to be in deficit.

Linked to this financial challenge is the struggle to retain staff. Working conditions in higher education are relatively generous compared to other sectors. But long working hours and insecure employment — more than a third of academic staff are estimated to be employed on fixed term contracts — have led to a situation where almost 75 per cent of respondents to a UCU survey reported that they are likely to leave the higher education sector in the next five years.

And a struggle also exists to attract students. Domestic applications via UCAS declined by 3.1 per cent in 2023 and 0.9 per cent in 2024 and there has been a decline in the number of visa applications from prospective international students for study sponsored visas.

These challenges are widely recognised — for example in their General Election manifesto Labour noted that “the current higher education funding settlement does not work” — and need to be addressed. However, addressing the shortage of finance, staff and students will not guarantee the competitiveness of our higher education sector or ensure that it enjoys the confidence of potential students. Universities should be preparing students for their future careers, enthusing them with the confidence and skills required to succeed. To do this, they must adapt assessments for the age of AI.

AI can complete university assessments, on occasions to higher standards than real students, and students are not shy about using AI tools. A recent study found that more than half of undergraduates have used generative AI to help them with assessments and Turnitin’s AI detection tool found that more than 10 per cent of papers it reviewed included some AI writing.

Universities have predominantly responded to this by attempting to detect and punish the use of AI. AI detection software is widespread and since 2022 some universities have handed out more than two hundred academic penalties in response to AI use.

This response is impractical given how hard it is to detect AI-generated content. AI detection tools face the demanding task of avoiding classifying large numbers of false positives (i.e. accidentally assessing human work as AI-generated), recent research has found that AI essays “verged on being undetectable”?and universities in America have told a Senate inquiry that preventing students from using AI is too difficult, if not impossible. This response is also counterproductive. AI is already impacting the world of work. If students are prevented from using AI, universities are failing to prepare them for this world.

Rather than attempting to ban AI — a strategy which is inherently flawed given the growing impossibility of detection — universities should redesign assessments around AI.

Work on this has already begun in other countries, for example in the USA academics are experimenting with group work and oral exams. However progress in the UK has been slow, with less than 10 per cent of undergraduates surveyed by the Higher Education Policy Institute observing a significant change in how their university conducts assessments.

Reimagining assessments would address not only the current misuse of AI but also the “archaic” university assessment culture and rampant grade inflation. Since 2010-11 the proportion of undergraduates receiving top grades has more than doubled.

UK universities are some of the best in the world and have already made progress at adopting AI in other areas of their work, with the Russell Group last year announcing a set of five principles to guide AI’s usage. It is crucial that they and the Government take the bold action needed to reimagine assessments.