AI in (higher) education: A call for sector-wide collaboration

Image Source: Created using AI tool DALL-E 2 using the prompt ‘educators and students in higher education of mixed race collaborating with AI’.

Artificial Intelligence (AI) has revolutionised every aspect of our society — from the smart devices we use every day to the breakthroughs in medicine and science that are changing the world. The concept of machines facilitating education is not a recent development. This idea dates back to the mid-1920s when Sidney L. Pressey pioneered the use of mechanically driven 'teaching machines' to aid students in learning multiple choice questions. Yet, when it comes to technology assisted education, the application of AI is still in its early stages.

There is much excitement and curiosity around what AI can and cannot do, as well as different visions on how to implement it in education. To the educators, it can serve as an assistant to automate administrative tasks and focus on innovative course design including personalised tools to support their students’ learning. To the students, it can spark initial ideas and content creation to help crack complex concepts. Unlike traditional methods, such as standardised tests, textbooks, and lectures, AI is able to adapt to each educator’s and each student’s needs and preferences. The data on which AI algorithms are trained on can be both reliable and unreliable at the same time. If higher education institutions want to embrace AI, universities have to work collectively to lead the design of pedagogic principles on various do’s and don’ts of AI to provide quality education – the Russell group has provided one such guideline.

There are many unrealised potentials of AI tools which largely remain untested in higher education. A frequently mentioned benefit of AI in an educational setting is that it might offer the opportunity to personalise education and to improve aspects which are frequently critiqued by students, such as the provision of formative feedback. A team of academics and data scientists from Warwick Business  School, led by Isabel Fischer, developed an AI-based formative feedback tool much before ChatGPT took the world by a storm. The software provides students with feedback on their drafts before submission, helping them improve their essays, blogs, reports, and dissertations, and to engage in peer-dialogue. The tool's unique advantage lies in its in-house development, which sets it apart from similar tools and generative AI technologies. Rather than relying on vendor information, we know that from the start we focused on meeting the EU and UK AI ethics requirements, for example, safeguarding data privacy by not requesting any personal data from students and not saving students’ submissions and feedback reports in our databases. Submissions also do not touch the public internet and students’ Turnitin scores are not affected.

A further strength of the tool are the graphs and visualisations which human markers would not be able to produce. Let’s take the example of knowledge graphs. A knowledge graph is similar to a mind map with additional arrows and text showing the logical flow. Imagine having written a dissertation of 12,000 words and then upon submitting your draft assignment the AI algorithms then produce a drawing of all the key concepts you have written about and their relationships. Besides the initial reaction of ‘that’s cool’, students use the graph to revisit their writing to check if they have gone into depth of interconnected concepts or rather remained superficial with many unrelated concepts.

Based on research focussing on an essay assignment last academic year where 512 year 1 students had the opportunity to submit their draft essay to the formative feedback tool, 320 students chose to submit their draft essays, while 192 students chose not to submit their draft essay. Though we did not account for causation versus correlation, i.e. students who submitted their draft essays might have been students who are generally more engaged, students who used the tool had on average a 3.4 points higher mark for their assignment (in percent that is a 6.47% higher mark), this is even more remarkable as we know that a relatively high proportion of international students with English as an additional language participated. Out of the 320 students who submitted their assignment, 115 submitted an additional survey, which revealed that a different demographic of students used the on-demand AI-based formative feedback tool than who attend in-person academic writing tutorials, with international students seeming especially keen to use AI. This additional survey also revealed that the majority appreciated receiving their feedback and that over a third of the students even prefer on-demand AI-generated formative feedback to human feedback.

We have just scratched the surface of the potential use of AI technology in higher education by incorporating a bespoke AI tool in student’s education experience. One of the features of AI is that it is scalable across the sector. Collaborations across the (higher) education sector are needed to enhance the understanding of AI applications and implications. The University of Warwick led one such interdisciplinary and inter-institutional collaboration, where our AI in Education community of practice brought together members of staff and students unlocking the AI debate to disseminate knowledge about AI in education. The endeavour shaped meaningful resources that provide guidance to educators and students on various topics:

• Practical use of AI for assignments

• Assessment designs

• The AI tools for teaching and learning including AI enhanced learning environment

• Academic integrity and AI ethics

To ensure that education remains a fair offering in our an increasingly AI-enabled society, educational stakeholders nationally and internationally must collaborate to keep the momentum and engage the academic community to collectively generate, share, and use knowledge. The full report is accessible using this link. We look forward to hearing your comments.