14 minute read
ChatGPT – Friend or Foe
What are the implications of large language AI models like ChatGPT for higher education?
In this feature some of our academics give us their opinion on AI tools like ChatGPT and the implications they might have for how we teach and assess our students.
Since its launch in November 2022, ChatGPT has evoked much discussion on its impact on all walks of life from business to education and even the arts. ATU like all other higher education bodies is assessing the rapidly developing field of large language Artificial Intelligence (AI) tools like ChatGPT and their implications for learning, teaching and assessment. We are working on a new Academic Integrity Policy, which will be in place for the next academic year.
Dr Perry Share is Head of Student Success at ATU Sligo. In this article he talks to us about the emergence of AI tools like ChatGPT.
AI has been a scientific goal since the 1950s, with ‘artificial intelligence’ coined in 1955 by American computer scientist John McCarthy. Since then, its fortunes have waxed and waned, as periods of enthusiasm and investment have been interspersed with ‘AI winters’, when the aim of creating machine-based thought has been seen as unattainable.
The current phase of AI development is based on ‘deep learning’ and involves computers ‘learning’ through statistical manipulation of vast data-sets. The outcome of AI research is already visible in applications such as computer spellcheckers and driver assist systems. Uses in education include learning analytics and intelligent tutoring systems.
Current interest in AI, and the realisation that education is on the cusp of a revolution, stems from the release of ChatGPT by OpenAI in November 2022. ChatGPT took an existing technology, GPT3, and presented it to the world as a chatbot that made it easy to use and highly accessible. For the older amongst us, it reflected how the Mosaic web browser transformed the internet in 1993. ChatGPT soon attracted 100 million users (including many students), making it the most rapidly adopted computer application ever.
Since the Oracle of Ancient Greece, humans have imagined a source that could answer any question. At various times the Bible, the encyclopaedia and Google have claimed this role. Generative AI, of which ChatGPT is an example, may have acquired that status. Pose just about any question in everyday language and you will get a well-informed, clearly expressed answer. Probe or prompt further and that answer can be enhanced and developed. If requested, academic or other sources can be identified: often, if not always, accurate.
ChatGPT has launched an AI ‘arms race’, involving major players in Microsoft, Google and, most recently, Amazon. AI is unequivocally here to stay. The challenges to education, and to institutions such as ATU, are fundamental. If AI can answer any question, what does it mean to ‘learn’ something?
New applications such as AutoGPT have the capacity to organise and complete all the steps of a challenge, such as setting up a new company, creating a website or organising a sports event. Where is the role for the business, IT or leisure professional? How will generative AI affect research, where it is already being used to identify research questions and analyse existing literature?
There are significant challenges for learning, teaching and assessment, for research, and for student support. How and what we learn and teach is going to have to change. This is not a matter of ‘tips and tricks’, this requires a rethinking of higher education. It is a challenge that faces the global education community, including ATU.
Dr Valerie McTaggart is Head of Department of Social Sciences at ATU Sligo. Her research focuses on the impact of digital technology to organisations and society. In this article she discusses whether ChatGPT poses a significant challenge to higher education.
ChatGPT as a digital disruptor to education has received significant attention amongst the academic community and the popular press over the last number of months. Like previous digital developments these disruptors have the potential to alter a marketplace at an alarming pace. We have seen several highprofile examples of the impact of the digital revolution on industries who did not capitalise on these digital developments. Remember Blockbuster!
The emerging narrative is that ChatGPT will cause significant challenges for third level education. As for the need for academic lecturing staff, suggestions have abounded that their role will become significantly reduced. Indeed, how can one person compete with all the information available to ChatGPT?
But what is all the hype about?
Being owned in part by Elon Musk, ChatGPT was always going to cause a bit of a stir. Like all great technological advancements its greatest benefit is its convenience. If it were clunky and difficult to navigate, we would not use it. What it does incredibly effectively is to pull information from an array of sources into one coherent unique response. No need to search through several sources to get an answer and then construct your own analysis and discussion. This is what’s currently causing so many educators concern. If there is no effort expended in developing one’s own response to a challenging assignment, then where will a student’s core learning come from? anywhere, and its functionality will only improve. As such we must learn to capitalise on the benefits of these digital developments to support student learning.
It should certainly make us reflect on our approach to education and assessment. However, what we need to be aware of, as education providers, is the potential of this emerging technology. We need to learn how to use the platform and encourage students to use this opportunity to support their learning in different ways. We also need to acknowledge and accept that this technology is not going
As education providers we need to prepare our students for this new world and reimagine what it is to be a student in the 21st Century. Finally, while this platform is new, technology enabled transformation has been around for decades and no doubt ChatGPT will be replaced by the next hot topic soon.
Dr Therese Hume is a lecturer in computing at ATU Sligo. She gives us her opinion on re-thinking teaching and learning assessment in light of the emergence of ChatGPT and other large language AI tools.
The seductive nature of AI tools like ChatGPT and the speed with which they’ve been adopted, despite numerous identified dangers, raises the urgent need to consider the social, ethical, legal and ecological implications of technological developments across the curriculum.
It’s easy to see how tools like ChatGPT that automatically generate essays or sections of essays in polished and persuasive language will continue to tempt stressed-out students and pose a major headache for educators.
For computing students, ChatGPT is already acting as a useful tutor through generating and explaining code, even if it doesn’t always work. Other uses include helping students develop critical thinking skills and supporting students with learning difficulties.
We might need to assist our students to develop new sets of skills like “prompt engineering” to make the most effective use of language models or other AI systems in areas such as software development, feature-writing, law, accountancy and so on. It would be critical here to address the need for awareness of the limitations of these systems, including the conditions that would need to be met for their safe and ethical use.
While technologies such as ChatGPT will have many implications regarding their use as tools within education, the speed at which they are developing means that it is also important to consider them within a much broader context, and in relation to other societal challenges such as war and climate change.
Dr Eoghan Furey is a lecturer and researcher in Machine Learning and AI at ATU Donegal. He explains why his opinion of ChatGPT is not nearly as positive as the description ChatGPT created itself and he illustrates how we can protect the integrity of ATU by adjusting how we assess our students.
“ChatGBT, an AI language model, has the potential to transform learning and teaching in higher education. It can provide personalized learning experiences, instant feedback, and accessibility to students with disabilities. Additionally, ChatGBT can assist teachers in designing and delivering courses, and researchers in analyzing and synthesizing data. However, the use of ChatGBT also raises concerns regarding dependence on technology, potential errors, ethical considerations, and cost. It is important to consider both the positive and negative implications of ChatGBT in higher education and address these concerns to ensure that the technology is used effectively and responsibly.”
The above paragraph was written by ChatGPT. There is no obvious indication that a human with a decent command of English did not write it (American spelling aside). This is what we are up against! My own opinion of the software is not nearly as positive as the description ChatGPT created itself!
These new large language AI models will force higher education institutes to re-evaluate how degrees at all levels are assessed. The integrity of our awards is at real risk, impacting all the disciplines and courses that we teach. How we approach this software will have far reaching consequences for our institutes, our students, our industry partners and all other higher education stakeholders and affiliates both nationally and internationally.
AI language models like ChatGPT can analyse, evaluate and to a certain extent, they can create. With a skilful user, they can write a degree level dissertation that will pass examination.
There are undoubtedly students who are already using ChatGPT and similar tools. Assignments are being submitted and marked without consideration of these models and we have no definitive way to prove if it is the students own work or not. The landscape has changed, and we have no choice but to change with it.
So how do we deal with it? We must be realistic in our approach. Unless we have witnessed the writing taking place, any work submitted by students could potentially have been created by using AI. Our business is not to grade the work of a computer. Our business is to educate and assess human students. Currently we have no silver bullet to directly address this challenge.
If we don’t adjust our assessments now, then we pay the price of lower quality graduates. This will result in a future workforce with underdeveloped skills. We must protect the integrity of ATU, the qualifications we award and most importantly, we must protect our students. The game has changed, so must we.
Dr Davy Walsh is a lecturer in Applied Social Care and Outdoor Education at ATU Galway-Mayo. He debates whether ChatGPT and other large language tools are a curse or blessing for higher education.
It’s hard to sum up my thoughts on this technology for several reasons but mainly because it’s still in its infancy.
We do, however, have a range of options that have all been tried and tested. Until we have a proper framework for addressing this issue, we can go back to what worked in the past; written exams, in class paper-based tests, and oral examinations. The Viva Voce (lively discussion) examination is designed to overcome any questions on the authenticity of the work and to assess the student’s understanding. We can return to these types of assessment to ensure that our students are deserving of their marks.
It might be best to start with what ChatGPT is enabling us to do. Can it really produce a solid well-argued academic essay? The answer is yes and no. The more detailed your request is, the more you get back.
Students who have a better understanding of a topic are more likely get a better response from ChatGPT. However, it currently comes with limitations which may not be the case in future roll outs of the software. For example, its language pack is a few years out of date, so its ability to discuss contemporary issues is seriously lacking.
That said, if we wish to retain the archaic ‘academic essay’ as the total measure of a student’s worthiness to obtain a degree then, ChatGPT may be a curse rather than a blessing, especially as improvements to the software are rolled out. For anyone really concerned about ChatGPT there is a Chat Zero app that will detect AI generated text, which might become available on Moodle or other education platforms in the not-so-distant future.
“ChatGPT may be a blessing if it forces academics to move into a space where assignments and exams become more individualised, creative, diverse, meaningful, and maybe community orientated.”
A more diverse form of assessment that comes about through collaboration and co-construction could make ChatGPT obsolete for the most part in higher education.
For example, group projects, practical’s, reflective writing, skills assessments, interviews, viva’s, presentations, poetry, and art projects would likely leave ChatGPT where it belongs – to the lower tiers of Blooms Taxonomy of learning. When it comes to research, literature reviews may still fall victim to abuse by ChatGPT, but I think that primary empirical research is out of reach.
I am not overly concerned with ChatGPT as its real impact will likely be outside of academia, perhaps in areas of entrepreneurship, bureaucratic administration, political rhetoric, coding, and software. Anything that challenges the status qua of academic practice is a positive development. Whether we retain our current practices or move forward with new practices, both will be stronger by the currently perceived threat of AI.
The AI generated image, pictured below recently won a photography competition in Australia. In a daring experiment Sydney-based company Absolutely AI entered an AI-created drone ‘photo’ of a beach at sunset into DigiDirect’s free monthly photo competition.
They said they did it to prove the world is at a turning point with artificial intelligence.