Contesting AI

Accountability is a necessary property of Artificial Intelligence research, development, procurement, deployment  and use.  When a citizen directly interacts with an artificial intelligence agent, and is experiencing difficulty or negative impact by that interaction,   accountability is there to guarantee that the AI agent is examined and possibly adjusted. For accountability to exist there needs to be an adequate socio-technical infrastructure to enforce it. In many regards regulating AI is concerned with building this infrastructure. However, a less explored aspect of the accountability relationship is the right to the individual beneficiary to directly contest the behaviour and specifically the decisions made by the AI agent. 

Existing and in-development regulation, like for example Article 21 of the GDPR  stipulates the right to appeal a decision. However, the operationalisation of rights such as these is also an affordance by the computational design of the AI agent, and it is our task to create that affordance. Contestable AI is a very new field that studies how to make AI systems open and responsive to human intervention throughout their lifecycle, not only after the AI operation is completed, e.g.,  automated decision has been made. 

This workshop seeks to push the scientific progress in contestable AI by providing a forum in which ongoing and new research on this topic can be presented and discussed.  We invite previously unpublished submissions on the topics of AI contestation that include but are not limited to: 

• Formal methods for objection and contestation of AI

• Socio-technical instruments of contestation 

• Human objection to automous machine behaviour

• Explainability and contestation

• Accountability and contestation

• Cognitive and psychological aspects of AI contestation 

We welcome original unpublished contributions. Please use theLaTeX styles and Word templates available at https://www.ijcai.org/authors_kit

Papers must be no longer than 9 pages in total: 7 pages for the body of the paper and 2 pages for references; an optional ethics statement can be placed either in the body of the paper or in the reference pages. For accepted papers, the last two pages can also contain acknowledgements and an optional contribution statement.

We welcome well written papers that describe ongoing or early stage work, but offer opportunity for an interesting discussion on the topics of the workshop.

We will also accept 2 page extended abstract of previously published or ongoing work that relates directly to the workshop and can contribute to an interesting discussion.

Authors commit to reviewing: By submitting to the workshop, each author volunteers to be entered into the pool of potential PC members/reviewers for the workshop.