Most students report AI errors but only half often check content

The vast majority of students have identified issues with work generated by artificial intelligence (AI) but less than half say they regularly check or verify its output, according to a major new survey on public attitudes towards new technologies. 

The research by the King’s Institute for Artificial Intelligence and the Policy Institute at King’s College London found that those attending universities often had conflicting views on the impact of AI and its usefulness.

More than half (56 per cent) of students polled say they are using AI a few times a week – compared with 33 per cent of the general population. 

Nearly nine in ten students (85 per cent) who use AI have found there were problems with the work or content it produced. The most common problems were factual errors or inaccuracies (37 per cent) and hallucinations such as made-up sources or statistics (31 per cent).

Despite this, 19 per cent of students admitted to rarely or never verifying the final output. Only 15 per cent say they always check, and 28 per cent say they usually do.

There was no clear consensus on whether universities encourage AI use, with a third saying they do, another third saying they are discouraged and the rest saying they have had no clear guidance either way. 

While 60 per cent of students say their university can prepare them well for an AI-shaped job market, only 36 per cent say they currently are being well prepared.

Looking at general attitudes to the potential impact of AI, just 24 per cent of the public as a whole think the technology is positive for humanity – compared with 43 per cent of university students.

And a clear majority (56 per cent) of students are excited about new job opportunities opening up as a result of AI. In contrast, only 35 per cent of workers and 28 per cent of the overall public feel the same.

Within universities, there is a clear gender divide. Male students are significantly more likely to say AI will improve their life (57 per cent) compared with female students (40 per cent).

Male university students are also the most confident that AI is improving their ability to think for themselves (41 per cent) – and female university students are most likely to think it’s having the opposite effect (46 per cent).

But despite their optimism for the use of AI, students tend to be more worried than others about its implications for the economy.

About six in 10 expect AI will have made the job market much tougher for them by the time they graduate. They are nearly twice as likely as the public to blame AI for the graduate jobs slump.

And 56 per cent think widespread AI-driven job losses would be worse than a normal recession – higher than any other group surveyed.

A third of students believe AI will eliminate jobs fast enough to trigger civil unrest – significantly higher than the view of the general public (22 per cent).

Despite acknowledging concern for the economy, people are noticeably less worried about AI’s impact on their own job. Graduate workers (45 per cent) are more likely than school-leavers (34 per cent) to express concern about their own role.

Although 78 per cent of students would still go to university, 30 per cent would choose a different undergraduate degree given the growth of AI. Twelve per cent say they would choose not to go to university at all.

Bobby Duffy, director of the Policy Institute at King’s, said people mostly look to the government, schools and universities to help young people adapt, but there is clearly much more that can be done.

“The public, workers, young people and university students are watching the rapid development of AI with more fear than excitement, with real concern for what it will do to jobs, particularly at entry levels and, therefore, the prospects for our young people and the economy in general,” said Duffy.

The study, which surveyed about 4,000 people overall, including 1,000 university students, found that large sections of society are undecided on how beneficial they think AI will be.

“It is still early days, and our baseline study shows that many don’t yet have firm views or much direct experience of AI’s impact, but that’s likely to change quickly, and we’ll need to outline clear plans on how we will adapt and support people in the transition,” said Duffy.

patrick.jack@timeshighereducation.com