Who’s Afraid of ChatGPT?
An audio version of Martin reading this post can be found on his blog.
I have been teaching in one way or another for 30 years this month. Every time I think I’m getting the hang of it, something new comes along heralding an education revolution.
But then I am reminded that change in education is often incredibly slow and, at best, superficial. The shiny things often quickly lose their allure. How we teach, how we assess and, most importantly to my mind, how we relate to our students need not be fixed but effecting positive, social justice-informed changes, such as those represented by Teachers Going Gradeless, is notoriously difficult. Despite frequent frustrations in this domain, I often marvel at how optimistic I am. I mean this only in a professional sense, though. Like many a stereotype of my generation and nationality, I am afflicted with eternal pessimism when I think of my (English) football team’s chances and the—perhaps slightly more important—prospects of the UK’s economy under our current government. But, despite daily exasperation and disappointment elsewhere, I tend to be optimistic and annoyingly ‘can do’ at work. This came in very handy when the implications for education of Covid-19 started to become apparent and likewise recently when contemplating the potentials and threats to traditional methods of assessment (and ‘life as we know it,’ if some commentators are to be believed) of novel text- and image-generation tools, collectively and commonly referred to as ‘AI.’
Sometimes circumstances throw opportunities or necessities our way and we witness sudden changes in practice. Covid-19 of course necessitated huge and swift changes in modality, pedagogy, and assessment practices, but the prevalence of deficit narratives in relation to how education happened in this period—typically (and unhelpfully) referred to as ‘online learning’—are indicative of the power of conservatism and strong structural desires to resist sustained changes and return to the comfort of ‘normal,’ despite many clear benefits such as those identified by Disabled Students UK.
The immediate panic in some quarters at the start of the pandemic have been echoed recently as teachers the world over gasped in horror as Open AI tools like ChatGPT offered credible, grammatical, and passable responses to some of their standard assessment questions (see this really good two page summary of what ChatGPT means to educators). It should be mentioned at this stage that this isn’t as ‘new’ as some might imagine. AI in education has been a source of anxiety for decades: witness the calamity promised by the pocket calculator, the mooted threats to basic language skills created by spelling and grammar checkers, and the end of language learning signaled by instant translation tools. Even the concerns so prevalent as I write were being discussed as imminent way back in the heady days of pre-Covid, in-person conferences when my colleague at UCL, Professor Rose Luckin (who knows a lot more about this than I do!) was arguing that we should be thinking carefully about the consequences of the weight we put on knowledge-based education.
In my role (I work as a faculty developer in a very large English university), I leaned heavily on my ‘compassion first’ approach at the start of the pandemic, broadly aligned to what we at UCL have framed as a Pedagogy of Care (hu)manifesto. My message then, as now, was ‘Don’t panic: let’s talk this through.’ Whereas Covid-19 precipitated necessary changes, the phenomenon itself has been perceived as transitory; a hurdle to overcome on a fixed path.
AI is here to stay, so it’s not just hurdles we need to navigate but decisions about which gates to open as we choose new paths to follow. The capabilities will only improve but, to be clear, I do NOT think we’re on the precipice of a take over by robots and very much appreciate the wisdom of experts able to articulate its limitations such as Janelle Shane on her blog AI Weirdness. If this is a crisis, then this could be another catalyst for systemic and sustained change to the way we do assessment (and feedback). It is also entirely possible that the knee-jerk reactionaries will attempt to ban it and regress further into an imagined idyllic past where assessment for everyone is a handwritten lengthy essay about facts they have memorized, produced in a huge hall policed by people with very squeaky shoes.
The largest part of my career was in what are called ‘Access’ programs in vocational (Further Education) colleges. These programs are designed for those who, like me, had left school at 16 or who failed the exam-heavy, so-called ‘gold standard’ university entry qualifications (‘A’ levels) but were looking for an alternative path into university as mature students. By design, these qualifications were inclusive, minimally graded, dialogic, flexible, and innovative. We designed our own assessments with deliberate contrasts to the A level and, inspired by student ideas, these took interesting forms and often incorporated digital innovations.
As my career transitioned into teacher and lecturer education, my enthusiasm for pedagogic and assessment innovation grew proportionate to my realization of how my experiences with Access programs were very much an exception to the rule. My optimism and belief in the possibilities and affordances of doing things differently led me to explore and experiment with a range of digital assessment and feedback alternatives to the typical and conventional. I trialed a range of multimedia assessments; audio and video feedback; digital portfolios; open badges and all manner of subversions along the continuum of ungrading possibilities. Some of these were spectacularly unsuccessful (open badges) whilst others were incredibly rewarding and transformational. In particular, I found audio feedback (and audio submissions) transformed the ways in which I was able to engage with my students and they, in turn, were able to engage with and use feedback productively (establishing ‘relational literacies’ as Ameena Payne and colleagues put it). However, I soon learnt that the growing body of experiential and research evidence barely chipped away at conventional practices, despite my advocacy and enthusiasm!
So it was (originally towards the end of 2020 and then updated in 2022) that I asked a similar question to the one I am asking in this article: Will Covid-19 finally catalyze the way we exploit digital options in assessment and feedback? I was tentatively optimistic that many of the practices such as greater use of multimedia assessment and feedback I saw colleagues across the sector exploring by necessity would become normalized. While the events of the last three years have undoubtedly advanced the breadth and depth of engagement by teachers and lecturers with these approaches, we are a long way from normalization. That distance means that associated benefits central to ‘going gradeless’ are less likely to be normalised too.
Whilst ‘lack of time’ or ‘I’m not allowed’ are usually the go to reasons for not exploring digital assessment and feedback opportunities (and all the equity potentials implied therein), we are witnessing an attitudinal sea-change as a consequence of teacher experiences of new innovations in AI. It matters not that it is misnamed. It is largely inconsequential that the generation tools have massive flaws. What matters is that there is a perceived and pressing need for substantial change and, whether we like it or not, control or at least influence of the response narrative is imperative. We need to do all we can to respond to fears by offering our colleagues insights and alternatives other than the reactionary. While increased surveillance and monitoring may be the instinctive response for many, we have in front of us an opportunity to do the things many of us have been advocating for years.
I work closely with the Faculty of Life Sciences at UCL and several colleagues have put essay and short answer questions through ChatGPT Open AI. Even from my lay perspective, it becomes quite clear the types of knowledge-based questions that are largely unproblematic and the types that tend to flummox it. Locating outputs away from the purely factual and theoretical to the applied, seeking explicit connections to in-class experiences and, above all, finding ways to de-center the grade and re-energize student interest in the process—not the product—of their education should be driving us all right now.
My optimism is tempered by the concern fear will drown out reason, but is simultaneously emboldened by the knowledge that the groundwork is in place. We have to hand years of sound and compelling scholarship of teaching and learning to draw on as well as a new necessity that is not going away. I believe the machines could and should herald and help facilitate a dawn of teaching where we can realize a more humanized, compassionate, inclusive, process-focused approach. The more of us that unify behind this message, the more likely that it will be realized.
For fun, I asked ChatGPT about these matters and this is what it came up with:
Martin Compton is an Associate Professor at the Arena Centre for Research-Based Education at University College London with a focus in program design, curriculum development, as well as teaching, assessment, and feedback enhancement. You can follow him on Twitter @mart_compton and on his blog.