Adapting academia: how is St Andrews responding to the AI revolution?

Ros Walker, Digital Accessibility Adviser, Student Services reflects on her experience of using AI in her work to support students. With thanks to Ros for featuring in Chronicle 2024.

AI and disabled students

Working in Disability Services, we can see huge potential for AI to support learners.

Assistive technologies have been using large repositories of data and language for many years. Speech-to-text (read-aloud) software and Text-to-speech (dictation) both rely on large corpora of language to operate effectively. What is now emerging is a further step up from that, in that it can generate language and share ideas.

We have all been familiar with ‘predictive text’ since the advent of the mobile phone, but Large Language Models or Generative AI (GenAI) can produce much more complex language and ideas in response to a stimulus.

• Grammarly is now able to detect spelling and grammar errors more effectively and can also make suggestions for better phrasing. In the Premium version, you can highlight a sentence or paragraph and ask for it to be made ‘more academic’.
• In ChatGPT, it is possible to add an assignment title and ask for it to be broken down into smaller steps. It will supply a full writing framework with word counts for each section.
• Software that will summarise academic papers (Scholarcy, Audemic) can simplify the reading process.
• Flash-card generation is offered by many pieces of software, based on your lecture notes or a transcript from a lecture.

However, students themselves are concerned about how far they can go with using AI. Recently a student asked me: ‘Is using Grammarly cheating?’ This dyslexic student had lost marks in assignments for spelling and grammar errors. The answer would have to be ‘No, it's not cheating, it’s giving you the tools to perform to the best of your ability.’

Another concern is that much of the software is built on OpenAI its use is now being challenged in the courts for breaches of copyright. (See Is ChatGPT more foul than fair?, Harvard Gazette) This means that software that is based on these Large Language Models could be withdrawn in the future. Our students could become familiar with something that supports them – and then it disappears. And that is also a worry.

This article was written by Ros Walker (Digital Accessibility Adviser) and made use of Grammarly to improve the quality of the language.