ChatGPT in Higher Education: My Thoughts & Recommended Actions

Much has been said about chat-based AI following OpenAI's release of ChatGPT late in 2022.

I've been enmeshed in conversations across the sector and with colleagues around this and wanted to take the opportunity to share some of my thoughts and recommendations while things are still fresh in peoples minds.

As academics and digital teams in Higher Education have explored this exciting technology (that has come out of nowhere for so many people), a range of implications have become clear and are gaining increasing airtime in social media and meetings/catchups across the sector.

The most common conversations seem to focus on what it means for assessments, plagiarism and cheating.?This is doubly scary as students across the UK are currently sitting end of semester examinations, many of which are online or digitally driven.?

It is without question that the increasing maturity of technology like ChatGPT has implications for examination.?With increasing understanding of the technology, though, it is clear that implications and responses are quite different depending on specific disciplines.

One might draw an analogy to other technologies that have changed pedagogy and examinations such as calculators (see this excellent article on the history of calculators), the printing press, the internet and internet search, word processors, etc.?To different degrees these have all caused the evolution or transformation of what is being taught, and how it is taught and examined, or the creation of new courses/modules/disciplines entirely.

This all makes it difficult for institutions to respond holistically, except in the most generic of ways.

Just as with other technologies, I suggest that there is a key role for institutions to support the creation of (and to maintain as the technology evolves) discipline specific awareness of the technology and its implications.?There is also a role for educators to consider and evolve what and how they teach and examine.

There are clearly a host of potential benefits to institutions, academics and administrators that become increasingly obvious as one explores the technology.?The ability to rapidly draft personalised responses to politely decline sales outreach, or to rapidly come up to speed with an area in which one is not a domain expert springs to mind, along with the ability to more rapidly do a host of simple little tasks that otherwise one might need to invest significant time in.

?

So what does this mean for Universities?

Like with Office skills, it seems obvious that AI skills will increasingly become expected, or necessary to a range of different vocations, helping people be more productive or effective and giving them an edge in competitive situations.?Universities that are interested in graduate outcomes and employability will need to consider carefully how their students will develop these skills.?

There is much rhetoric in the media and social channels which suggests that Universities (and/or business) which fail to adapt to AI will fail.?Bold statements such as this are great for catching the eye, but there is some truth to the implication that institutions which fail to understand, adapt and adopt these technologies will ultimately struggle to teach well, to graduate students with the skills needed in society, and to be efficient and effective in doing so.

Very rapidly institutions need to (have) respond(ed) during their examination period in one way or another.?The lack of high confidence tools to detect ChatGPT use means that for most institutions responses are likely a restatement/refinement and communication of policy, more so than specific technology implementation, although some institutions have announced plans to move to entirely in person examinations, for example.?It is important to note that while current ai based text generation tools have a relatively low reliability in identifying generated text, it is obvious that this reliability will increase over time, presenting institutions with the opportunity to come back to assessments from this transitionary period and re-test them.?Instituting an approach not dissimilar to athlete drug-testing, in which historic samples are re-tested as tests improve may be an appropriate response.

There is, though, one aspect of the rise of ChatGPT that I see getting little airtime:?the pace and rapidity with which these advances come -- the speed with which ChatGPT reached huge user numbers and awareness dwarfs is remarkable.

Institutions/organisations tend NOT to be able to respond with pace and rapidity.

Many institutions will have commissioned cross-disciplinary working groups to examine and respond to the implications of ChatGPT.?Collegiate, inclusive and engaged discussion on areas such as this form a vital part of the culture of Higher Education and are at the heart of many of the very positive contributions that the sector makes to society.?These groups may, in time, create a set of recommendations for changes that might be made at a local and institution-wide level.

Regularly these changes will need to be endorsed by a body in the institution.?All of this is a sound and appropriate approach towards governance, but struggles to deliver action rapidly.?We are rapidly approaching a time when the pace of release/change of new technology exceeds an institutions cycle time for response and institutions must consider how they increase agility while sacrificing as few as possible of the benefits created by the current models.

There is a large amount to consider here, and as the pace of technology release continues to pick up, we can expect these "it can do what?" moments to occur with increasing regularity!

?

To close out my thoughts -- ?I advocate that institutions use this opportunity to do 4 things

Immediate: Brief educators on AI/Deep learning tools and look to revise curriculum and examination to be AI aware in a nuanced and discipline specific manner.

Immediate: Consider and refine examination and assessment policies, and communicate these.?Consider whether to later re-assess submitted text from this transitionary period for AI use.

Mid-term: Develop a plan to brief staff on, and to invest in AI based technology for personal workflows and institutional processes.?

Mid-term - Long-term: Use the rise of ChatGPT as an example to discuss desired levels of institutional agility.