Generative AI has become normalised in British universities, with most students using an AI tool to support studying and only 5% likely using AI to cheat, a first national survey of students and AI since the advent of ChatGPT has revealed. But urgent action is needed to stop a new digital divide emerging, and students want clear AI policies and support.
Among students surveyed, 53% have used generative AI to help with their studies. The most common use is as an ‘AI private tutor’, with 36% using AI to help explain concepts.
“For every student who uses generative AI every day, there is another who has never opened ChatGPT or Google Bard, which gives some students a huge advantage,” said report author Josh Freeman, policy manager at the Higher Education Policy Institute (HEPI), in a release. Male and Asian students are more likely to have used AI than others.
“The divide will only grow larger as generative AI tools become more powerful. Rather than merely adopting a punitive approach, institutions should educate students in the effective use of generative AI – and be prepared to provide AI tools where they can aid learning,” the report stated. There is also a digital divide between institutions, with some embracing and others side-lining AI.
The survey report recommends that universities develop clear policies on AI use, teach students how to use it properly, and provide AI tools to students who cannot afford them. Education authorities should urgently explore how AI will affect assessment.
The study by HEPI, an independent higher education think tank, was published as a policy note on Thursday 1 February. “Provide or Punish? Students’ views on generative AI in higher education” was produced by Freeman, HEPI and digital education company Kortext.
The survey polled 1,250 undergraduate students through UCAS, the Universities and Colleges Admissions Service, in November 2023, exploring student attitudes to a range of generative AI tools that have emerged since ChatGPT launched in November 2022.
In universities, excitement over generative AI’s ability to enhance learning, support students and reduce workloads has been accompanied by concern over its potential use for cheating.
But, said Freeman, universities “have upheld standards of rigour, and they deserve credit. Students trust institutions to identify the use of AI tools and they feel staff understand how AI works. As a result, rather than having AI chatbots write their essays, students are using AI in more limited ways: to help them study but not to do all the work”.
Some other key findings
According to the policy note, 66% of students consider it acceptable to use generative AI for explaining concepts, 54% for suggesting research ideas and 53% for summarising articles. “Only 3% think it is acceptable to use AI text in assessments without editing.”
AI is used to produce text for assessments by 13% of students, who usually edit the content. “Only 5% of students put AI-generated text into assessments without editing it personally – which we expect will be prohibited by most institutions,” said the policy note.
Worryingly, few students see the ‘hallucinations’ of generative AI as a real problem, which may suggest they are not verifying information and may be using inaccurate information.
Regarding student views on how universities have responded to the whirlwind of generative AI, 63% think their institution has a clear policy on AI use, with only 12% thinking it is not clear. Most students (65%) also think their institution can spot work produced by AI.
According to the policy note, 30% of students surveyed agreed or strongly agreed that their university should provide AI tools – only 9% said they currently do. Just 22% of students were satisfied with the support they have received on AI. Almost three quarters of students (73%) expect to use AI after they finish their studies.
The digital divide
The survey uncovered digital divides based on both ethnicity and wealth. Among students from the most privileged backgrounds, 58% use generative AI for assessments against 51% of students from the least privileged backgrounds. Students with Asian ethnic backgrounds are much more likely to have used generative AI than white or black students.
While there is not much difference in the overall proportion of male and female students using AI, there is a gender divide in the way it is used, according to HEPI. Male students use AI text generators more, and are more likely to use AI for data analysis and coding, while female students more often use AI for editing, translating text and transcription of speech.
The ‘digital divide’ may apply to institutions’ use of generative AI as well. “The UK higher education sector is often accused of moving slowly, but that is only partly true of its approach to generative AI,” stated HEPI. “While many institutions have kept their historical approach, others have moved rapidly to integrate AI with their educational provision.
“Both approaches appear to carry risks. But given the potential benefits of embracing AI in improving the student experience and most students expecting to use AI in the future, it seems unlikely that any institution can avoid tackling it head-on for long.”
Student and staff uses of AI
The survey found that a popular application of AI, used by 37% of students, is for enhancing and editing writing, while 30% of students said they had used AI tools like ChatGPT to generate text and 25% have used AI for translation, for instance using Google Translate.
In free-text answers, students described a range of other uses for AI such as creating an essay structure, having AI ask questions from a block of text to test student knowledge, generating AI images, or simply for “inspiration”.
Interestingly, reported the policy brief: “Just over half (52%) agree that academic staff ‘understand well’ how AI is being used by students, with a minority (15%) disagreeing.” Further, 65% of students are ‘quite’ or ‘very’ confident that lecturers can detect if AI is used, while 22% are ‘not very’ or ‘not at all’ confident.
There is evidence again of a digital divide, with wealthier students much less confident of their university’s ability to spot AI use than other students. “This may suggest that more privileged students are better at using AI without detection,” reported the policy brief.
Asked how adaptable institutions have been regarding assessment, 9% of students said assessment had changed significantly while 24% said it had stayed the same, 23% believed it had changed a little and a substantial 23% were unable to say.
Students were divided equally on whether it was their university’s responsibility to provide paid-for AI tools. About 30% agreed and 28% disagreed while 33% were neutral.
Regarding whether their university provided AI tools, 58% of students surveyed said no, only 9% said that it did and a third were unsure. “It is clear many more students want AI tools to be provided by institutions than currently are,” reported the policy brief.
Further, the policy brief said: “While students are quite confident that staff understood their AI use for assessments, they are much less positive about how prepared staff are to help them work with generative AI.” Only 18% of students agreed that staff are well-equipped.
Students articulate their views
Students were also asked to explain answers. Regarding institutional support, there were some very positive and very negative responses. Among the positives: “It was talked about from day 2. I know what’s ok to do and what’s not.” And: “My institution is very clear on how to use AI effectively and the problems with blindly following it.”
Among the negative responses, many students highlighted that AI was seen only as a threat or means to cheat.
One said: “We are told to not use it and that’s it.” And another: “We are neither being informed, advised, instructed, or assisted to learn more about and deal with AI. All that is done is threatening us that if we use it, we face consequences.”
Some students were very positive about the role of AI, including a dyslexic student who reported using AI to ensure that work was correct and understandable. There were many negative comments, including that generative AI repeats itself, “generates bollocks”, does not really apply as yet to the humanities, and that work should be completed by students.
Conclusions and recommendations
The survey made clear that most students have quickly taken up generative AI. Anti-cheating systems seem to have the desired effect so far: “The higher education sector can consider this an early success,” HEPI concluded.
The ‘digital divide’ beginning to emerge has some groups of students adapting more quickly to and benefiting from AI more than others. Still, few students know basic facts about generative AI, such as how often it ‘hallucinates’.
Students want more support with AI and more AI tools to be provided for them. “But many institutions may be understandably wary of encouraging the use of AI tools,” said the policy note, and later: “There is work to do to constructively integrate AI with learning in a way that does not undermine educational provision.”
The policy note recommends, first, that universities “develop clear policies on the acceptable use of generative AI in learning and assessment, and efforts should be made to communicate these to students from the beginning of their course”.
Universities and departments should identify which generative AI tools “can support learning constructively and which may be detrimental”, and these should be regularly reviewed. There is guidance issued by, for example, the Joint Council for Qualifications.
Second, universities should teach all students how to use beneficial AI tools appropriately and effectively, according to the policy note. “This might include instruction on the different kinds of AI that are available, how to write an effective ‘prompt’ (input), and how to evaluate the quality of content generated by AI,” it states.
Third, efforts should be made to equalise access to generative AI tools that help learning. When students need access for paid-for AI tools such as GPT4, universities should consider funding subscriptions – “as many institutions already do for other digital tools”.
Finally, national policymakers, such as the Department for Education in England, should urgently commission a review of how assessment will be affected by AI, and should publish routine guidance on how to identify and respond to AI challenges in assessment.
“The threats to robust assessment are likely to grow more pronounced over time as AI tools become more sophisticated and students more adept at using them,” stated the policy note.
2024 University World News
