What should HE be teaching students about using GenAI for study and employment?
By Jonquil Lowe, Senior Lecturer in Economics and Personal Finance, and Liz Hardie, Director, Scholarship Centre for Innovation in Online Learning, at The Open University
When ChatGPT burst onto the scene in November 2022, the immediate reaction in HE was alarm at the potentially existential threat to assessment from students using Generative AI (GenAI) to do their assignments for them, especially in distance-learning contexts. But GenAI – meaning large language models used to create predictive text, images and other outputs in response to prompts – will also transform many workplaces and invites us to reconceptualise what is the unique element that humans can bring to an activity. That, in turn, prompts HE to reflect on the need to embed GenAI into the curriculum, but what exactly should we be teaching and assessing about GenAI?
In mid-2024, as part of a small team at The Open University (OU), we set out to answer that question through a mixed-methods research project. On the demand-side of the labour market, we wanted to understand what GenAI skills employers are seeking in graduates and how that might change. We therefore surveyed 150 job ads from three major recruitment platforms (Indeed, Glassdoor and LinkedIn) that included GenAI as a core requirement, and conducted a literature review. For insights into the supply side of the labour market, we held focus groups and a survey of OU students and staff to examine their views, expectations and needs around GenAI learning and teaching.
Figure 1 shows the job ads broken down by sector. The four most popular sectors for posts requiring Gen AI skills were technology (35 ads), financial services (22), information technology (20) and consulting (19). The majority of the education jobs were in higher education (13 of 16).
Figure 1: Job ads by sector (number of ads; total 150)
Most of the jobs specified that candidates should have a university qualification. However, 12 out of the 150 ads were willing to consider equivalent experience, while another 52 ads merely listed required experience and skills without specific mention of qualifications. This is consistent with research by LinkedIn (2023) that has seen a shift in recruitment away from degree requirements towards skills-based hiring. This may pose a challenge to HE to demonstrate its ongoing relevance to evolving workplaces.
We also classified the job advertisements by role. A majority of the jobs (31%) were technical, software-related jobs; 20% were research and development positions; 17% were for data scientists and 12% were solutions-architect roles that bridge the gap between understanding client needs and identifying technical answers.
A minority of the jobs (5%) sought experts in linguistics, law and the humanities to work on training GenAI. This is consistent with research in 2024, when US freelance platform, Upwork , noted a fall in earnings for lower skilled translation, writing and customer service work that it attributed to GenAI displacement effects.
A handful of the 150 jobs were for jobs with firms that are embedding GenAI across their business or whose main business is ‘selling software as a service’. Examples included GenAI applications to legal, audit and underwriting work. This points towards a future where working alongside AI may become the norm in professional services sectors.
This dichotomy of jobs either being displaced or augmented may be a driver of the attitudes of student and staff towards GenAI. On balance, student and staff attitudes tended to be positive (ratio of positive to negative sentiments 1.39 and 1.26, respectively), but there were marked differences by discipline. More numerate and scientific subject areas tended to be the most optimistic (for example, sentiment rating of 2.00 for maths), in contrast to areas such as creative writing and education where negative attitudes dominated (sentiment ratios of 0.73 and 0.67, respectively).
There was general agreement that students needed to develop GenAI skills in order to be able to compete effectively in the jobs market.
‘If we don’t educate [students] on how to use GenAI to help them in whatever material or whatever subject they’re studying, they’re going to be behind the rest of their cohort, their peers, when it comes to finding a job.’ STEM-subject tutor
‘AI Is going to be there in your job in some way, so to be…educated without AI is in my view foolish.’ Social-science student
Drilling down into what should be taught about GenAI, three dimensions emerged as shown in Figure 2.
Figure 2: Types of GenAI teaching required
The first dimension is pragmatic guidance on the permissible use of GenAI as an aid to study and when tackling assessments. This is not simply about deterring misuse but also encouraging appropriate GenAI use – this is especially important in supporting equality, diversity and inclusion where GenAI may, for example, help to overcome disability-related disadvantages in academic reading and writing skills. The student focus groups highlighted the need for universities to go beyond high-level statements and to support students with granular guidance to inform specific use cases. This could be tackled by, for example, building up a bank of frequently asked questions, but crucially requires giving student-facing staff the knowledge and confidence to provide guidance.
The second and arguably most important dimension is teaching and developing students’ critical AI literacy. This embraces aspects such as: understanding broadly how GenAI works including its capacities and effective use cases, and also its limitations, such as risk of bias and hallucination; understanding the need to verify and critique GenAI outputs and developing the skills and ability to do this; and being aware of the potential social and environmental impacts of GenAI and its ethical use in the broadest sense.
The third dimension is teaching students to be able to apply GenAI effectively and ethically in cases that are discipline-specific, such as using GenAI either to conduct the first stage of a task before the human takes over or to check and refine a human output.
More broadly, effective use of GenAI means using it to support not replace learning. This in turn means identifying how a task can be broken down into the part where it makes sense to use a tool (which may be GenAI but could be another type of tool, such as Excel or a calculator) and the part that requires human skills. Effective uses of GenAI as a tool that were identified by the focus group participants included, for example: unfreezing learning where a student has become stuck; summarising; navigating and analysing large data sets; creating images for slides; translation; and creating or checking coding. The human element involves both good use of the tool – such as prompting skills and output evaluation – and the application and integration of the tool’s outputs into the completion of the task.
In the survey of staff involved with creating and delivering the curriculum, almost all 154 respondents thought that GenAI needed to be taught to students as an important employability skill. Understanding GenAIs limitations and being able to critically review GenAI outputs were highlighted by 100% and 99% of respondents respectively. Over 80% thought that GenAI needed to be taught at all levels of study with progression as students moved through their qualification. So, teaching to first-year students should embed appropriate use of GenAI in their studies and assessment, while subsequent levels should build up critical AI literacy and the discipline-specific applications. GenAI teaching should be coordinated across modules and qualifications to avoid unnecessary repetition. It was suggested that it would be useful to have a mapping tool (perhaps at qualification level) to do this effectively.
It is clear that GenAI I here to stay and will have a profound effect on how our students engage with work and life more generally. Making sure that HE prepares students for this new era is exciting but will require major rethinking of what learning outcomes we want students to achieve and how we deliver and evidence those achievements.
