Interview with Kave Noori and Marine Uldry from the European Disability Forum: “Nothing about us without us”, including AI
The image features photos of Kave Noori and Marine Uldry.
The European Disability Forum (EDF) brings together organisations from across Europe, representing over 100 million people with disabilities. In the field of AI, EDF advocates for the rights of people with disabilities within AI policy and its development, while also raising awareness about AI within the disability movement. We spoke with Kave Noori and Marine Uldry about their work on AI and disability justice.
Kave Noori is EDF’s Artificial Intelligence Policy Officer andjoined the organisation in January 2023 as part of the Disability Inclusive Artificial Intelligence project.
Marine Uldry is EDF’s Human Rights Policy Coordinator.
In what ways does AI affect the lives of people with disabilities?
Artificial intelligence can have both positive and negative effects on the lives and rights of people with disabilities. It all depends on how it is developed and regulated. When AI is developed with accessibility, usability and affordability for people with disabilities in mind, AI can make our lives easier and more convenient.
Take for example a voice-controlled personal assistant that lets you control things in your home like lights, TVs or washing machines. In some cases, AI can also be a superpower that enables us to do things we could not do otherwise. For example, when a journalist with limited mobility can write articles thanks to a voice-to-text technology that works. Some AI solutions may be developed specifically for people with disabilities as a target group, while inventions of AI-based products or services may be aimed at all people but may prove particularly useful for people with disabilities.
The problem that can arise in the context of AI and disability is that people with disabilities, in all their diversities, are not considered in the development of AI systems. In such situations, AI can become another barrier. Or in the worst case, it can even seriously hurt us. Canadian professor Jutta Treviranus, who has been working on inclusive design for more than 20 years, was once invited to participate in an experiment to test machine learning models that guide self-driving cars. During the simulation, when the car was confronted with an unusual situation, in this case a friend of the professor who is in a wheelchair and has a habit of pulling her wheelchair backwards across the road at an intersection, the model failed to recognise the professor’s friend as a pedestrian and ran her over. When the model was trained with more data, i.e. more traffic situations, which is a common way to correct AI that does not work accurately, the car ran over this person with even greater certainty (here again, it was a simulation).
What prompted the European Disability Forum to work on AI and why?
AI is now talked about everywhere. With a proposal for a regulation on Artificial Intelligence by the European Commission, EDF started to be more aware of the importance of having the disability movement involved in discussions related to Artificial Intelligence. We also heard examples from other marginalised groups being discriminated against by AI systems. For example, when it comes to social security and access to employment. For us, it was important to do more work on AI to understand the exact barriers, challenges and opportunities for people with disabilities.
It is also important to recall that the moto of the disability movement is “nothing about us, without us”. Because Artificial Intelligence increasingly impacts everyone’s life, EDF decided to work more on AI and to support our member organisations in understanding AI and developing their work on the topic.
How easy has it been to put AI on the agenda and raise awareness among other disability rights stakeholders?
It has been an uphill battle. People with disabilities unfortunately are the group that is often left furthest behind. The EDF has now been working on the AI Act and the rights of people with disabilities for a while, and the Fund’s support has been a gamechanger for our capacity. Besides, there are many of our member organisations that are just starting out in the field of AI and disability justice.
Now we are in the final phase of negotiations on the AI Act. Luckily EDF is in good company as we have joined a coalition of human rights organisations that also work on the AI Act, where among others, we closely cooperate with the European Digital Rights (EDRi), Amnesty Tech, Access Now and AlgorithmWatch. Working together strengthens our efforts because we can give advice to each other and support each other ‘s positions. So the EDF has been able to take action and fight for important things that apply to the whole disability rights movement. But there are also issues that are very important to certain EDF members that would have benefited from being able to do their own lobbying to complement the joint work through the EDF. And it is not that there is a lack of interest. It’s just that in many countries our member organisations have limited resources, and they may be busy struggling to survive by ensuring that their members have access to shelter and social support.
What have been the main milestones of your work and challenges that you’ve faced so far?
Getting this success required many other meetings and discussions in the European Parliament. Through hard work and some miracles, we were able to convince the rapporteurs and the shadow rapporteurs to introduce mandatory accessibility requirements for high-risk AI at the last minute (about 3 days before the vote). The challenge, however, is to ensure that those references are accepted by the Member States during what is called the “trilogue” negotiations.
Another milestone I would like to mention is the work to raise awareness of our members. In March we had a big conference on AI for our members, that took place in Sweden. At the moment we are also organising a series of workshops on AI in relation to other topics such as education, employment, social security, etc. Our challenge there is that AI can be such a complex topic that it takes time to fully raise awareness of our members and persons with disabilities in general.
Have you ever had to counter ableist narratives in the context of technology development? If yes, what stories do these narratives tell and how do you approach them?
Yes, even several times. The most striking example that has stuck in my mind is when people developing AI systems want to make an invention and specifically have in mind that a particular group of persons with disabilities should benefit from that invention, but they don’t ask the group for whom they want to make something useful before they start the development process.
Take the sign language glove as an example: several inventors have won innovation competitions independently of each other. The only problem is that they have put several years into developing the invention, even though it is not useful or helpful to the deaf. The reason is that the glove only takes into account the shape of the hand and the position of the hand and assumes that these are the only components of sign language, and then uses this input to produce text or synthetic speech. However, sign language is much more complex, with about 70% of the message being conveyed through facial expressions and the tilt of the neck and upper body. The only way to distinguish between a question like “Are you hungry?” and the statement “You are hungry” is conveyed through facial expressions like lowering the eyebrows, tilting the head to the side and bending the upper body forward. And then when an inventor wins an innovation award and insists that this invention can be helpful to deaf people without consulting them in its development, this can raise expectations of many. For example, with politicians who might think that we now have an invention that can replace sign language interpreters. The moral of the story is that as an inventor of an AI system, one has to know that you are not (in the majority of the time) the user or the sole user of this system. When it comes to persons with disabilities, AI developers cannot just try to imagine what people with a particular disability or with different disabilities need. People are much more complex than we can imagine. AI developers have to consult people who have experience with different disabilities, and they have to do this already when they have an idea of the new invention.
What books or podcasts on technology/AI have you enjoyed recently and would recommend to others?
I can recommend this resource from the European Center for Not-for-profit Law (ECNL) to anyone who wants to get started on learning about AI and human rights.
I am also very excited to get my hands on the book ‘Unmasking AI: My Mission to Protect What Is Human in a World of Machines’ by Dr. Joy Buolamwini that will be published on 31 October 2023. I’d also recommend to read this super interesting interview with the author.
One podcast that I follow on a regular basis is called ‘Last week in AI’.
Note from the editor: you can also subscribe to the EDF’s AI & Disability Newsletter.