Opinion piece.
When it comes to artificial intelligence (AI) and its impact on education, it’s fair to say we’re currently in the murky waters of the unknown, trying to make sense of change and any level of certainty is ill-founded. The current context and rapid pace of change doesn’t offer much in terms of evidence to analyse the long term effect on the education sector. That being said, there are interesting ethical and philosophical debates which I find engaging that look at the now and how we set ourselves up to work with the unknown. In particular work on ethical acknowledgement (Phipps and Lanclos, 2023) and how AI may lead to circumventing the academic process in a drive for efficiency (Watermeyer et al., 2024) I find hauntingly pertinent. When it comes to use of generative AI (Gen-AI) in assessment, there are some useful models to explore that balance creativity against integrity, such as UCL and Leeds approach.
In introducing Professor Martin Bean’s keynote at the University of Leeds Digital Summit, DVC Professor Jeff Grabill offered his insights into how we can navigate through the disruption of AI in higher education. Here are three aspects that stood out for me.
Early disruption
Grabill talked of the first few years of AI in education holding the potential to be underwhelming, but in 15 years we’ll be in a position of being overwhelmed. It’s a representation of the hype cycle, where users overestimate capabilities in the early days of emerging tech. Here’s a case in point. When I asked Microsoft Bing running ChatGPT 4 for a summary of my work based on my website, it was all rather generic and missing key points. In some cases it over-emphasised aspects that I had not written about (I’ll share this in a separate post!). This may be down to poor prompt phrasing, or perhaps I don’t articulate very clearly my professional specialisms (am I too generic?), but without that human reflection, it would be easy to say AI isn’t doing what it’s supposed to. The bias in AI responses is also well-documented and uncritical acceptance of Gen-AI outputs is concerning to a range of disciplines. As another example, Gen-AI content-driven explainer videos on YouTube, which sound like Wikipedia-based scripts mashed with more stock b-roll than is healthy for someone to witness, target the right keywords to bring in large volumes of views and advertising revenue. There is backlash against these, as shown in video comments, and a despondency for the extraction of human narrative from the storytelling medium of film. In 15 years time I can well imagine the feeling of overwhelm, either through the significant shift in working practices or the inescapability of Gen-AI. This may bring societal benefits, but certainly will lead to societal and educational change.
Educate, not police
The second point is that we should ‘try to stay out of the business of policing and remain in the world of education’. This, Grabill explains, means not being unduly blunt with the application of policy when students make mistakes, but to be tolerant and enable opportunities to learn.
My understanding is to look for the learning potential, essentially what teaching and learning is about, rather than reach for the rule book when students use AI. Yet academic institutions and educational policy places significant demands of academic integrity on students. Every summative assessment is high stakes, even if their mark weighting is low, as every assessment holds the potential for academic misconduct. There is a risk in some cases that even formative assessment is treated with the same dispassionate appraisal of compliance. The fear of misconduct leaves little room for misinterpretation or, perhaps, creative use of AI. Therefore, it may be argued, there is little opportunity for learning how to learn. Yet, what is often framed as a problem with students is better interpreted, in my view, as a problem with assessment. Not that I have a single answer to what solutions there may be.
There’s much debate on the purpose of assessment in education, particularly higher education where authenticity of task and inclusivity of assessment can sometimes struggle against tendency towards essay writing and closed examination. Gen-AI has already triggered responses that demand handwritten exams in large halls, which cries against the last decade of more inclusive educational practices.
Formative assessment that is designed well has an important role to progress learning, feeds-forward to future learning and enables students to both identify their own learning gaps and celebrate success. Even with the acknowledgement of benefits of assessment for learning and opportunities for synoptic and portfolio assessments, summative assessment still is placed as a gateway activity for qualification that is rarely reflected on and with feedback unlikely to be engaged with constructively. Assessment is often time consuming for both educators and learners, and in some circles there is a view that there is too much assessment. Is it any wonder that students may be using AI for assessment when the dialogue and process of learning through assessment isn’t valued (summative assessment in particular)?
There are positive routes for the use of AI too. In Martin Bean’s keynote (Bean, 2024), he highlighted use cases as diverse as career development, campus security and personalised course materials. For example, devising question sets, producing robust variations of questions based on course-specific training data, would save significant time for educators and allow for basic knowledge and application checks by students. Using AI to assess work may even bring advantages in standardising marking (noted that biases in training data will need compensating for), or recommending study routes to students based on performance. However in some cases, uses of AI lead us into a path of reducing human-human interaction in learning. For example, during my participation at the RIDE Research in Distance Education Conference at the University of London earlier this year, the dystopian vision of AI marking generative AI work emerged from our table discussions and really challenged the whole concept of submitted assessments. If that point is reached, everyone loses: the role of the educator is diminished and the role of the student in learning is undermined.
I’m reminded of the film, War Games (1983), where at the climax a cold war computer simulation battles against itself perpetually leading to mutually assured destruction. The computer ends up playing ‘tic-tac-toe’ (or as is otherwise known, ‘noughts-and-crosses’), developing an understanding of no win scenarios, leading to the concluding line: “the only winning move is not to play”. I am probably not the first person to connect this film to our current concerns of AI in education. It offers a poignant thought though, if the only ‘winning move’ by educators or institutions is not to play the assessment game. Assessment needs reframing as ‘for learning’.
Education is about experiences, not content
The third point from Grabill’s introduction aligns very strongly with my own view as an educationalist: “we’re not in the content delivery business, we’re in the experience business.” Higher education is about learning, experiences and learning experiences, and this can be interpreted at many levels. From the holistic experience of being a student at an institution, through to the day-to-day micro-experiences of interaction with other students and academics. Activity and experience design is based on human interaction.
Bean drew attention to the language of a ‘global skills emergency’ which stems from careers shifting, education needing to prepare students for jobs not yet defined and the advancement of technological solutions. The slow pace of universities to respond to knowledge and skills demands from industry has left a gap, which large multinational tech companies are filling with professional certificates. Offering microcredentials as a route for higher education to support the skills agenda, Bean referenced the World Economic Forum (2023) Framework for Action and the need for “skills-first culture, policies and mindset.” Technical skills, for example, are supported through lifelong learning as technologies change, but the skills to approach challenges, problem-solve, collaborate and communicate are able to be developed through programmes of study. Bean cited the Deloitte (2024) reports on the skills agenda, how human contribution will change with AI and other technologies. There is a continuation of the idea going back to Hagel, Seely Brown and Wooll (2019) of “enduring human capabilities” that enable individuals to adapt and bring the human element to future careers in tandem with AI. It seems to me these capabilities are those which are developed through interaction. They cannot be taught or learnt through content, but through experiences.
AI may offer ways to free up time for both students and educators to interact more. As Watermeyer et al. (2024) caution, freeing up time gives rise to filling that time with other activities viewed as more productive at an institutional level, which are likely to be tangible outputs. However, I would argue some of the most valuable ‘outputs’, which come through educational interaction, are also the most intangible and often the impact is personal and felt much later. Therefore at an institutional level, the benefits of AI have to be framed as making the educational experience more personal, and the language of efficiency challenged with the language of human interaction. If AI really does hold the potential for redirecting educational time towards experiences and interactions, we may find a way to navigate and coexist with AI after all.
References
- Bean, M. (2024). Navigating the Professional Learning Landscape: The Road Ahead. Digital Summit: AI and me. 24-25 June 2024. University of Leeds.
- Deloitte (2024). Global Human Capital Trends 2024. Deloitte Insights.
- Hagel, J., Seely Brown, J., Wooll, M., (2019). Skills change, but capabilities endure. Deloitte Insights.
- Phipps, L. and Lanclos, D. (2023). ‘An offering’, Digital is People.
- Watermeyer, R., Phipps, L., Lanclos, D. and Knight, C. (2024). Generative AI and the Automating of Academia, Postdigital Science and Education, 6. 446-466.
- World Economic Forum (2023). Putting Skills First: A Framework for Action. An Insight Report published by the World Economic Forum’s Centre for the New Economy and Society, in collaboration with PwC.
Leave a Reply