I attended two accessibility related events week beginning 23^rd June. I was virtually at Nottingham University’s Digital Accessibility 2025 Conference on Wednesday 25th and on Friday 27th went to the JISC ‘AI and accessibility skills: building the accessibility professional of the future’ event in London.

JISC Event Highlights

The JISC AI and Accessibility event was interesting; I went ready to have my scepticism about AI challenged. I’ve been at too many AI centered events where someone tells me with great confidence that ‘In the next year, AI will revolutionise [Insert whatever here]’ and all that seems to happen is there are less fingers in the AI generated photos and search engine AI results get things incorrect more confidently.

Ethical, environmental and quality issues with GenAI/LLMs aside, I’m not anti-AI but the way it’s being shoehorned into everything (whether it’s any good or not) is a terrible way to deploy a technology. While the AI you get on your new phone, TV, toaster or Office Suite is generally a substandard experience, I suspect when the current bubble bursts AI will mature into a solid tool in some fields, with coding and accessibility being strong candidates.

So it was marvellous to attend something without the snake oil; what was covered was measured and punctuated with discursive elements. It probably helped being sat on a table with some other Learning Technologists, someone from The Government Digital Service and a rep from W3C – we had a good group with a variety of accessibility experience and knowledge.

The event was led by Dr Sarah Lewthwaite with colleagues from the University of Southampton. We were introduced to their research into ‘Teaching Accessibility in the Digital Skill Set’. Their research covers investigating the teaching of digital accessibility in both academic and workplace settings. Themes such as recognition of disability as a site of expertise, engagement with models of disability, and challenging techno-ablism and bias were introduced.

I was particularly taken with the concept of ‘Questioning what is already known’; when The Digital Education Office ran “lived experience” events in collaboration with AbilityNet in 2019 I had some of my existing preconceptions challenged. Some stuff is just cemented into ‘what is known’ and understood, but disabled folk will often powerfully challenge and take down this received wisdom. It shouldn’t be revolutionary that we take measures to include them in discussion and development from the very start.

We were asked ‘Is this a moment of transformation?’ – a great question with a muddy answer from my point of view. With the Equality Act 2010, the Public Sector Bodies (mobile and websites) Accessibility Regulations 2018 and now the European Accessibility Act 2025 all creating requirements for Universities to ensure that our digital estate is accessible to everyone, it feels like the potential for AI to fix issues and make things actually accessible is massive. AI could mitigate the very real financial, reputational and workload issues we currently face.

We heard from Dimple Khagram, CEO of Purple Beard, about her thoughts on humanising AI so that it works for humans. Dimple had some interesting takes on what it takes to “Be the human AI can’t replace”. It was very much about collaboration rather than competition with AI. I still can’t get past the fact I’ve yet to save time with AI here for the majority of use cases. I can create images, edit audio, write words faster and more authentically than when I’ve attempted to with AI, when you factor in the significant editing and checking of AI generated content. But I will admit I am possibly not the target audience in this use case.

Accessibility consultant Tim Scannell spoke to us via two BSL Interpreters about AI attempts to create BSL translation software and how a good idea suffered from implementation – namely there hadn’t been any engagement with BSL users, so the tech focused on the fingerspelling and missed the additional context given from body language and facial expressions was missed. He questioned where the data used to create the robotic avatars had come from, it was missing some information and things like BSL slang and regional dialects were absent. The phrase ‘Nothing about us without us’ was used several times and I feel that nailed a central theme of the days topics.

The final presentation ‘Accessible by default?’ was from Dr Benjamin M. Gorman from Bournemouth University’s Department of Computing and Informatics; it focused on the skills we can’t afford to lose. The fact that GenAI needs to be explicitly told to make things accessible was highlighted, along with the fact humans need to have a deeper understanding of accessibility to be able to create the right prompts and pick up on things that cannot be codified – the visual accessibility of something for instance needs human level checking and editing. The phrase ‘No ethics. Just output.’ was used, which speaks volumes to me about my issues with having to heavily edit AI output to the point I find it faster and less frustrating to just do it from scratch.

The sessions were punctuated by discursive breakout sessions where we were prompted to answer questions such as ‘Can Accessibility function at the speed of AI?’ and ‘Can accessibility be automated?’

Nottingham Digital Accessibility 2025 – AI Highlights

AI and Accessibility was featured heavily in Nottingham University’s Digital Accessibility 2025 Conference too. I saw some interesting work from Steve Wang at Nottingham around an AI driven tool which could take images of equations and convert them into accessible code for use in their Digital Learning Environment. Accessible equations is a Holy Grail of digital accessibility, so it’s great to see AI helping to make improvements.

The standout session for me at Nottingham’s conference was from Alice Bennet, from University of York’s library team who presented findings from a paper titled ‘AI and Accessibility: assistance, responsibilities and risks’. Alice framed the potential for AI and accessibility but also the risks and dangers of reliance on AI. A strong point was made of the need to unpick Generative AI/LLMs from AI which has been used in assistive tech for decades. She argued that an uncritical reliance on the technology over more traditional accommodations and adjustments could lead to negative outcomes. The loss of agency for disabled students as they are pushed towards AI solutions that might not be quite right really chimed with what Tim Scannell was talking about with the BSL interpretation software at the JISC event.

I’d recommend Alice’s AI and Accessibility: feature, the future or fad post on York’s Digital Accessibility blog as a piece of further reading on the subject.

‘Nothing about us without us’

I think AI has massive potential to lower many barriers, but we still need to keep a skilled human eye on this ‘solutionising’ and actually include people with lived experience of disability; there can’t ever be a one size fits all fix, it’s all too personal, too individual for that. AI could meet this, but steps need to be made to address this now.

I’ve worked with students who are disabled or require adjustments for almost 25 years in both Further and Higher Education and I’ve seen Ed Tech and Assistive Tech used to overcome the barriers they face, to varying degrees of success. ‘Nothing about us without us’ speaks volumes where technology is used to provide solutions. Disabled people need to be front and center of this. The rush to use AI in everything should not be any different.