Ensuring Artificial Intelligence systems respect disability rights



Ensuring Artificial Intelligence systems respect disability rights

This article provides a summary of the afternoon sessions of the Conference “Disability Rights, Accessibility and Artificial Intelligence”, dedicated to the impact of Artificial Intelligence on Disability Rights. For the morning sessions, focused on the future of disability rights and accessibility, read “Steps for an accessible, disability-inclusive future: our conference in Sweden


Disability-inclusive Artificial Intelligence?

 

Moderator Elias Tebibel and Executive Director Catherine Naughton
Moderator for the first panel

During the first panel, experts explored how digitalisation, Artificial Intelligence and automated decision-making affect the daily lives of persons with disabilities. The panel was moderated by the Chair of EDF’s Youth Committee, Elias Tebibel. Panellists included UN Special Rapporteur Gerard Quinn, Professor Jutta Treviranus and our Artificial Intelligence Policy Officer Kave Noori.

Among the conclusions, the feeling is that we must address the risks before focusing on the positives.

A critical risk, for instance, is that automation and AI systems replicate discrimination and stereotypes and exonerate individuals from responsibility.

Panellists stated that the algorithm is discriminatory by definition, as it is based on a statistical average. This means that persons with disabilities – who don’t fit the average – suffer discrimination in all areas, from recruitment to security, banking services and access to education. A practical example is that algorithms misinterpret facial disfigurement during recruitment processes and associate it with unworthiness.

Reporting discrimination is complicated by the fact that algorithms are a “black box” where people cannot know how decisions were made and, therefore cannot seek redress.

The panel agreed that to make AI developments more inclusive of persons with disabilities:

  1. Persons with disabilities, and their representative organisations, need to be involved in creating AI systems – including by developers and businesses;
  2. Developers need to avoid a “one size fits all” approach and acknowledge the need for an intersectional approach, as persons with disabilities can also be part of other marginalised groups;
  3. Algorithms and patterns of data exploitation need to be changed to consider the diversity of the human experience;
  4. AI solutions need to not oversimplify support with assistance – for example sign language gloves don’t take into account the fact that sign language uses the whole body;.

To learn more about the impact of AI on persons with disabilities, read the UN Special Rapporteur Gerard Quinn’s report “Artificial intelligence and the rights of persons with disabilities

Legislating AI

MEP Brando Benifei on the screen
Member of the European Parliament in charge of the AI Act, Brando Benifei

The second panel focused on the upcoming EU’s Artificial Intelligence Act, which intends to regulate the development and use of AI systems. Experts explored the Act’s possibilities to protect human rights.

The panel included the Member of the European Parliament responsible for the Parliament’s position on the AI Act: MEP Brando Benifei; Francesca Fanucci, Senior Legal Advisor at the European Center for Not-For-Profit Law and Maureen Piggot, our Executive Committee member. It was moderated by Professor Stefan Larsson.

Panellists discussed:

  1. The call to ban unacceptable AI systems and strongly regulate high-risk AI systems, that include social scoring, biometric identification, applications to migration, law enforcement, education, work, justice and essential service;
  2. The need to include accessibility requirements in the Artificial Intelligence Act, in line with the UN Convention on the Rights of Persons with Disabilities;
  3. The need to develop AI systems with active, informed consent and control from participants that call for strong provisions on human rights and mechanisms to seek redress;
  4. The importance of running risk assessments not only before deployment but while AI systems are being used – to correct for possible oversights.

The last panel focused on concrete tools to implement the inclusive design of AI and automated systems. The presenters – Professor Stefan Larsson, Professor Jutta Treviranus, and Anna Felländer, CEO of anch.AI – showcased how AI can be regulated and mechanisms to ensure inclusion.

Jutta Treviranus talked about the dangers of model maturity, which amplify bias in AI. She warned that systems that use current ethics and datasets excluded the ones that most need assistive devices and support.

Panellists pointed three urgent measures to create disability-inclusive AI systems:

  • Have humans involved at every stage;
  • Creating more collaboration between disciplines;
  • Use a diversity of approaches and offer people alternatives to AI.

A Call for a disability-inclusive AI Act

EU Commissioner for Equality Cabinet Member Nora Bednarski , and EDF Vice-President Gunta Anca

At the closure of the event, our Vice-President Gunta Anca presented it’s the conference’s political outcome: the EDF Resolution “EU Artificial intelligence Act for the inclusion of persons with disabilities”.

Through this resolution, the European Disability Forum and its members are calling on the EU to live up to its obligations under the UN Convention on the Rights of Persons with Disabilities and ensure that all persons with disabilities can benefit from accessible, affordable, and available AI technologies that support their socio-economic participation and independent living.

Closing speech

Nora Bednarski, Cabinet member for European Commissioner for Equality, closed the conference and recounted her visit to Skellefteå, Access City Award winner, highlighting the important role of local authorities in delivering accessibility;

She shared concerns and implications of unregulated Artificial Intelligence and discussed the importance of the AI Act in safeguarding fundamental rights and establishing solid guarantees in the system so it prevents discrimination.

She stressed that the development of the AI Act has digitalisation of public services and interoperability across member states at its core to make cross-border data exchange more secure.