The following article is the second article in our series on the European Accessibility Summit. It focuses on the sessions on Artificial Intelligence.
Also:
- Find the summary of the first sessions: Keynote and session on the European Accessibility Act
- Find the summary of the sessions on Public Procurement.
- Find the summary of the sessions on higher education and on video game accessibility.
The Summit included three sessions on Artificial Intelligence, highlighting how these systems are a game-changer in the digital world.
The section started with a panel on AI for Accessibility. The panellists included:
- Ana Carla Pereira, Director ‘Equality and Non-Discrimination’, DG Justice & Consumers, European Commission
- Jenny Lay-Flurrie, Chief Accessibility Officer at Microsoft
It was moderated by Alejandro Moledo, Deputy Director and Head of Policy of the European Disability Forum.
Artificial intelligence: improving accessibility technology
The panellists highlighted how Artificial Intelligence is also helping to build new solutions on assistive technology and accessibility. Ensuring that the data used to build these systems is representative of the population is crucial, otherwise it risks replicating discrimination.
Highlights from the panel also included:
- How the AI Act requires that high-risk systems are accessible and with measures installed to safeguard fundamental rights.
- The need to create new standards that ensure inclusive design and protection of fundamental rights. They need to be made collaboratively between industry, authorities and civil society.
- Involving persons with disabilities is crucial not only to safeguard accessibility and fundamental rights, but also to ensure the systems reflect accurately persons with disabilities.
- Panellists also touched on the risks associated with data collection, including privacy and security concerns.
Ana Carla Pereira focused on the need to protect fundamental rights and ensure accessibility
I think it’s clear that both things are interlinked, because if you don’t have accessibility embedded in the design of AI, then we are failing in developing AI that enables the rights of Persons with Disabilities.
Jenny-Lay Flurrie shared some thoughts on the consequences of not using the right data:
Data is very important. (…) You have to proactively acquire data and put the right guardrails, the right framework and ‘shelving’ around data..
She added:
Tech companies need to be addressing (accessibility in AI). It absolutely goes back to the core basics: If it’s accessible, (and if you) test and listen to your users.
Alejandro Moledo summarised the disability movement approach on AI, when he stated
We know in the disability community, that AI systems can (…) discriminate against persons with disabilities and that concerns not only the safety, the privacy and the security of persons with disabilities, but also the lack of accessibility of certain AI systems. This is why the disability community (…) (does) our best to keep up with the technological developments and to keep operative conversations with the private and the public sector on the ethical use of AI and disability inclusive AI systems.
David Banes, Chair of the Equitable AI Alliance took the floor to explain their initiatives. The Alliance brings together technology companies, persons with disabilities, their representative organisations and researchers to highlight the opportunities that AI offers and the challenges and risks to be addressed.
The future of technology and disability
The afternoon started with a presentation by Dr Cecily Morrison, Senior Principal Researcher Manager, Microsoft Research Cambridge. She
Dr Morrison leads a diverse team to ensure technological solutions are inclusive. She explained some examples of practical benefits and challenges of this practice:
- Understanding the diversity within marginalised communities.
- Working with consumers with disabilities and ensuring their input is included by specialised teams.
- Guaranteeing that users can personalise their experience.
Dr Morrison then showed a practical example of training AI for image-generation and how it creates an issue of inadequate representation of disability. She highlights the importance of using accurate data libraries and ensuring that AI can pull from real life, not fictional examples.
She summarised the points beautifully stating “no AI about us without us“.
The following panels focused on accessibility and public procurement. Read the next part “European Accessibility summit: Accessibility meets Public Procurement“