Welcome to the WJETT blog

The buzz is endless – AI is taking Higher Education by storm. Since the launch of ChatGPT, everyone seems to have an opinion, and rightfully so. It’s so new and we have yet to fully understand its potential and the impact it will have. Within academia, the general sentiment mostly surrounds concern for responsible use, as many students have heard their professors and lecturers warning them against its use. However, its rapid growth and widespread adoption demonstrate that it’s not going anywhere soon so instead of avoiding it, it should be given the time of day to understand the risks and challenges but also the opportunities it presents. Furthermore, I think the student voice in these discussions has been underrepresented, but really students can be the key to harnessing this technological advancement as an asset to enhancing learning and education.

The WIHEA group have already explored a number of subjects regarding AI in education from student perspectives that can be found on the group’s Artificial Intelligence in EducationLink opens in a new window webpage. These have included emerging questions AI presents, the risks and ethics of academic integrity, evolving assessment styles to mitigate and integrate AI into assessment and how teaching may evolve. I will now explore some of the opportunities that are being presented with the widening availability and access to AI tools for students to enhance their learning and to generate formative feedback. While summative marking has been and continues to be required to be marked by human markers according to the UK Quality Code in Assessment (UKSCQA, 2018), formative feedback has more flexibility, and we are now presented with an opportunity to test and utilise the capabilities of these AI technologies in providing timely, constructive, and developmental feedback.

Existing feedback systems

This notion will be particularly explored with regards to formative elements of summative assessments. Feedback should allow a student to understand strengths and weaknesses of their work and if engaged with effectively, can be used to improve academic performance, and thus learning. Especially throughout the pandemic, we have seen the role of feedback change massively: as more of education has shifted online, reliance on formative assessments has increased as assessments for learning. This is in contrast to summative assessments which more so represent assessments of learning (Wyatt-Smith, Klenowski and Colbert, 2014). Formative assessments also are an opportunity for autonomous learning by developing one’s own skills and relying on self-motivation. It would also be fair to say that formative feedback can be self-assessment of sorts, as even though the formative feedback is generated externally, it is the engagement with, and learning you apply from it that will ultimately make a difference in each student’s performance.

AI generated feedback

So what could incorporation of AI in these feedback systems change? Well, the use of algorithms in generation of feedback is not an entirely new concept. Algorithms, such as Grammarly and Sketch Engine, have been around for a while and they can generate feedback on academic writing and are for the most part freely available, or students are granted access to them by their institutions. But with more complicated algorithms that use machine learning, we can apply them to provide specific and personalised feedback. To make this even more applicable, by integrating what could be different elements of summative marking criteria or rubrics, they could provide some of the most relevant feedback at a moment’s notice.

This application is indeed being explored right here at the University of Warwick. Isabel Fischer, a WBS professor, is trying to pilot a deep learning formative feedback tool that has the potential to provide more comprehensive feedback that was developed with WBS marking criteria at its core. By simply submitting a pdf or word document, the algorithm instantly produces a document of in depth feedback on the four aspects of WBS marking criteria. This could be just the start of developing similar department-specific feedback tools taking into account department-specific assignments, marking criteria, and writing styles for drafts of academic writing. While there are definitely some considerations to look out for, this is fascinating and shows great promise as a tool to increase student autonomy in adapting how they might approach assignments to still have the opportunity to personally benefit from formative feedback.

Considerations of using generative AI

The considerations I mentioned earlier are worth discussing as students are turning to generative AI technologies like ChatGPT more and more. While these technologies are being developed to simulate human intelligence, there are some things they are simply not capable of. For example, it lacks expressions or expressive language. If using them to generate feedback on your writing, you should be aware that they will not always be able to grasp the nuances or expressive language in that writing. In other words, any feedback you receive from AI should be approached critically. You decide what you implement from feedback you receive, and you are responsible for identifying and understanding what truly can improve your work. This is all part of the responsible use of AI, but really also goes for human generated feedback. Your assignment at the end of the day will still be marked by a human marker with in-depth subject-specific knowledge and skills that they are asking you to learn and demonstrate in your assignment. I think this is the quick, irresponsible and neglectful way people have been quick to exploit resources like ChatGPT, where they do not doubt any response it has generated and implement them into a piece of work, only to find that its references are wrong or entirely don’t exist. Firstly, this should not be the way we utilise it, as this is blatant plagiarism, but secondly, a critical approach should be used to (for example) verify references, and critically understand that the way AI answers can lack certain elements of context. Regardless, the point still stands: responsible applications of AI technologies should not be about using it to do your work, but using them to enhance or improve your outputs.

Engagement with AI technologies and feedback

A new level of engagement with AI has been seen since the release of ChatGPT and DALL-E. Perhaps this is rooted in the great advancement that this represented or, more sinisterly, the opportunity to exploit the technology to minimise workload. Regardless, everyone’s interest has been piqued, and the level of engagement has been massive, exceeding what anyone might have expected particularly from students. At the University of Warwick alone, students have made a total of 850,000 total site visits to ChatGPT in the first two months only on the university’s Wi-Fi (SWNS, 2023). I think it’s important to try to understand why this might be in order to be able to channel this traffic for good rather than just fear this alleged ‘cheating epidemic’ that the media has been dubbing it.

In contrast to the older algorithm technologies that have been around, like for example the previously mentioned Grammarly and Sketch, which experienced much more moderate levels of engagement and use. Reasons vary from lack of awareness, to limited breadth of feedback to language, or to lack of confidence in the feedback they provide. AI has surpassed some of these limiting factors in that it is capable of generating a wider breadth of feedback that can include language, style, structure and more. The confidence in the feedback it produces is reassured by the continuous concern from educators. If professors are afraid AI technologies can be used to write entire assessments, then they must be capable of doing so.

Further benefits

As a result, we have seen students be a lot more open to trying to use ChatGPT, and I think we should utilise this eagerness in a way that encourages students to enhance their academic achievements. By introducing resources such as Isabel Fischer’s feedback tool or teaching students how to formulate prompts for ChatGPT to generate similar constructive feedback, we can guide a smooth integration of ChatGPT into Higher Education practices. And there are so many more benefits we have the potential to see. For one, this alleviates a massive workload off staff. If such tools are available to take care of the writing style and structure, staff’s role in formative feedback can remain more focused on content. The speed (or even instantaneity) with which AI can produce feedback also makes feedback more accessible. Furthermore, students can engage with it as many times as they like, inputting multiple drafts, as they are less limited by staff work capacity. Also, different students work on different timescales and with different approaches when faced with an assignment. This further widens accessibility to students that might start assignments later that what might normally be a formative deadline. Communicating these advantages is key in order to achieve these outcomes and to harness this technology towards enhancing educational experience for both staff and students.

Conclusion and personal experience

In my experience thus far with using ChatGPT, I have had mixed feelings. On the one hand, I am very apprehensive of the fact that its use is quite contentious at the moment, with some professors explicitly prohibiting its use or consultation. On the other hand, it is a resource that is available and it feels foolish not to use what is at your disposal. Throughout the research that went into this article and discussion with faculty members about its potential to provide feedback, I have been introduced to a very clearly constructive way to engage with ChatGPT, that seems to make both staff and students happy. While we are still in the early stages of understanding the potential and risks of generative AI technology, at the end of the day this is a tool that will have implications for Higher Education and we are being faced with the possibility of either embracing it, in various ways such as to generate formative feedback, or let it escape our control at the cost of academic integrity, because it is clear that prohibiting its use will not prevent people from exploiting it.

For further queries: marabortnowschi@yahoo.caLink opens in a new window

Reference List

SWNS (2023). 'University of Warwick fears cheating epidemic as data reveals huge number of students visiting AI website during exams.' [online] Kenilworth Nub News. Available at: https://kenilworth.nub.news/news/local-news/university-of-warwick-fears-cheating-epidemic-as-data-reveals-huge-number-of-students-visiting-ai-website-during-exams-176836Link opens in a new window [Accessed 19 Jun. 2023].

UKSCQA (2018). UK Quality Code for Higher Education Advice and Guidance Assessment. [online] Available at: https://www.qaa.ac.uk/docs/qaa/quality-code/advice-and-guidance-assessment.pdf?sfvrsn=ca29c181_4Link opens in a new window [Accessed 16 Jun. 2023].

Wyatt-Smith, C., Klenowski, V. and Colbert, P. (2014). Assessment Understood as Enabling. The Enabling Power of Assessment, pp.1–20. doi:https://doi.org/10.1007/978-94-007-5902-2_1Link opens in a new window.