Contestable Artificial Intelligence

With AI systems playing a bigger and bigger role in society, how can we ensure our rights to autonomy and dignity are respected? One way is to make sure AI systems allow human intervention during development and after deployment.

This project aims to uncover new knowledge regarding the design of mechanisms that enables people to:

  • Meaningfully contest individual automated decisions made by AI systems; and,
  • Collectively contest the design and development of AI systems, particularly when it pertains to datasets and models.

Transparency, fairness, and accountability in AI systems are not problems that can be solved purely by technical means. Effective solutions require careful consideration of technical and social factors in tandem, and should take local contexts into account. These are challenges that a design approach is uniquely equipped to meet.

Our project takes a practice-based, action-oriented approach. Our main research activity is aimed at prototyping mechanisms for contestability in new and existing AI systems in and outside of labs, with a special focus on AI used by local governments to make automated decisions pertaining to urban public infrastructure.

Through this project, we expect to be able to present a portfolio containing examples of contestable AI in context, along with generative, intermediate-level design knowledge that will aid others in the research and design of AI systems that respect human rights.


  

Partners:

  • Amsterdam Institute for Advanced Metropolitan Solutions (AMS Institute)
/* */