Students not yet dissuaded by AI exposure in choice of degree subject, report says

Undergraduates continue to choose courses with heavy AI exposure, despite leading to fewer job opportunities in some fields

Prospective students are undeterred by applying for courses with heavy AI exposure

Applications to study degree subjects which are heavily exposed to artificial intelligence have seen an increase among prospective students, even as entry-level opportunities in those fields begin to decline, according to new research.

Following the public release of ChatGPT in November 2022, applications for university courses vulnerable to AI disruption - including computing, mathematical sciences, and engineering - saw an unexpected increase.

A study conducted by Dr Bouke Klein Teeselink, from King's College London, with Aristotle Vossos, Andrei Andronic and Kartik Akileswaran, of the Supervised Program for Alignment Research, examined domestic application trends across 1,090 higher education degrees from 2020 to 2025.

By linking millions of LinkedIn profiles with UK admissions data from UCAS, the researchers tracked how degree choices aligned with graduate career exposure to large language models.

Instead of pivoting away from fields where automation threatens career returns, the study found prospective students were doing the opposite.

After ChatGPT launched, degrees ranking one standard deviation higher in AI exposure experienced a six per cent relative growth in applications compared to less exposed programs.

Yet, while student interest in exposed subjects increased, the proportion of junior and entry-level positions in those target occupations fell by 0.25 percentage points per standard deviation of exposure.

Potential mismatch in skills supply and demand

Researchers warned that if the trend were to continue, it could contribute to a mismatch between the skills graduates acquire and those needed by employers, and countering this could require intervention by policy-makers.

However, the authors added it was not necessarily the case that students were making a mistake in their selections. AI may yet create new tasks and jobs for those graduates, as technologies have done in the past, and entry-level hiring may recover.

The report said: “To counter any mismatch, policymakers could first improve the information on which students base their choices.

“Beliefs about labour market prospects are often outdated and overly optimistic, but experimental evidence shows that accurate labour market information corrects such beliefs.

“Secondly, policymakers could invest in retraining. A degree chosen today reaches the labour market only years later, so even well-informed students will sometimes graduate into subjects that weakened along the way.

“Conversion courses and other forms of retraining can absorb part of this mismatch, and they will become more important as AI continues to shift the demand for skills.”