Skip to main content
Queen Mary Academy
  1. Queen Mary Academy
  2. Educators
  3. Innovation and Scholarship
  4. Innovative Pedagogies
  5. Centre for Excellence in AI in Education
  6. Blog
  7. Items

No Shortcuts to Learning with Generative AI?

The generative-AI revolution through which we are living has enormous potential to increase business productivity, and this alone is sufficient reason to ensure our graduates are familiar with the possible uses of large language models and their offshoots.  But it’s also a reason to be cautious.  How does gen. AI fit into the economy of learning, where time is arguably a principal currency?  

Published:
Two people with swirly lines opposite each other

 How does gen.AI fit into the economy of learning, where time is arguably a principal currency? 

I find this question profound and challenging.  Universities generally award a bachelor’s degree for courses of study worth 360 credits spread over three years.  This period can be extended through part-time arrangements but not compressed, no matter how efficiently a student progresses through classes and assessments. Bringing time into the cost side of achievement in higher education would devalue its nature as personal formation: a process of mastering a discipline by submitting to the tuition of more senior scholars as one’s thinking and skills develop through patterns and habits of reasoning and practice. Bringing in time as a cost factor also risks incentivising malpractices like ghost-writing that undermine learning.  The fact that many undergraduates today need to earn a living while studying makes the cost of time all too real but does not negate the threat this poses to genuine educational transformation. 

If ghost-writing has been a threat to the integrity of education for as long as qualifications have had value, the possibilities of passing assessment points by using gen.AI to produce coursework bring new threats to the integrity of higher education.  Are we not living through a perfect storm?  While the opportunity cost of students’ time grows as opportunities for casual work increase, gen.AI brings astonishing potential for almost mindless generation of passable coursework – all while the drive towards more-authentic assessment sees the phasing out of (AI-proof) exams in many areas. 

Jon Jackson and I explore these issues in a short article; "No Shortcuts to Learning". That title makes our main point.  But we don’t argue that students should simply be prevented or discouraged from using gen.AI tools.  Instead, we consider how the various components of learning or assessment tasks may bring varying amounts of formative benefit – while at the same time having varying susceptibility to time-saving uses of AI.  For example, a data science project requires reading and writing (highly beneficial, and potentially avoided altogether by using gen.AI), computer coding (good exercise, but almost perfected by AI), and formatting (not so formative, and potentially well done by AI).  Some assignments require data gathering, and we’ve had cases of students using gen.AI to avoid the task of surveying their colleagues.  Besides being straightforward academic misconduct, this time-saving tactic eliminates the formative benefit of engaging with real people’s views.  But what can be done to navigate this minefield?

The problem is especially acute for us because the students Jon and I teach are degree apprentices, in “Digital Technology Solutions”, who already use AI tools in the workplace that sponsors their university studies. And whereas most undergraduates have yet to enter a workplace where efficient generation of text and code is a sought-after skill, the employers of our apprentices already seek productivity from day to day.  Indeed, the BSc towards which they are working will be awarded on the basis of a prescribed number of days of study alongside their workplace learning.  The value of time is firmly in view for these learners.  Indeed, questions have been aired about how a full degree can be obtained from two days per week of university study (even if typically continued over three annual semesters for four years).  

What we basically recommend is that any time students save by using gen.AI (e.g. in coding, formatting or sense-checking) should be compensated by time spent on more formative tasks (e.g. reading, writing and experimentation, as well as, for apprentices, active observation and reflection in the workplace).  And we need to develop this proposal further by considering how best to structure and schedule assessment tasks.  We appreciate the Queen Mary Academy’s training and guidance on using gen. AI in teaching and learning. 

Beyond this, my suggestions include: 

  • Publishing mark schemes that clearly indicate how credit will be awarded for investment in attributes like understanding, judgement and originality  
  • Giving credit for demonstration of practical work (the “show, don’t tell” principle), including very specific details of interest (the “information is surprise” principle) 
  • Including an assessment early on in a programme that requires critical use of gen.AI but is orally examined, to help instil reflective skills and habits 
  • Demonstrating at an early stage how writing in the style of a chatbot (even if done without gen.AI) is of low value in today’s world and does not enhance employability 
  • AI-testing assessments (e.g. with MS Copilot as approved by QM) to obtain samples for calibrating to a minimal mark such as 40% 
  • Encouraging creative use of non-verbal communication (e.g. diagrams and photographs) 
  • Celebrating the analogue: hand-drawn diagrams, oral presentations, and spontaneous responses to questions (e.g. in vivas or group interviews). 

Fascinating insights into how and why students are using AI appeared in March in a WONKHE study reporting from a survey of 1,055 students across the UK.  And one thing this reveals is the extent to which individual students have personal policies for when and how to use gen.AI.  One undergraduate in that study expressed an idea exactly in line with what we propose: “I find [an AI search engine] cuts the time I spend looking for resources in half and frees up more time to actually critically evaluate sources”.  Stepping back, a headline finding of the study is that students use AI in more constructively formative ways (or simply use it less) when they understand the requirements and rationale of an assessment and know that they will face an in-person moment of accountability, like a viva or presentation.  There’s much to reflect on here, and our teaching team will be examining what else we can take from the study.   

Learning as we work together with our apprentices, we face the challenge to demonstrate and enhance the value of higher education under the watchful gaze of multinational corporations (the employers) and government (Ofsted liked what they saw last year, but we can’t be complacent!).  The pressure is on providers of degree apprenticeships to forge viable practices for teaching and assessment in a furnace where business and academia come closer than perhaps anywhere else.   

Dr Richard Gunton 

Senior Lecturer in Data Science

https://www.qmul.ac.uk/spcs/staff/academics/profiles/rgunton.html 

 

 

 

Back to top