Explainable Agency

When:  Wellington 1:00 pm-2:00 pm Tuesday 01/11/2022

Where: RH107 Room capacity is 15 or ZOOM Join from PC, Mac, iOS or  Android: https://vuw.zoom.us/j/97733606047

It is now widely accepted that explainability is crucial for supporting an appropriate level of trust in autonomous and intelligent systems, as well as to make these systems understandable, allowing them to be predictable, traceable, and to be challenged. In this talk, I will focus on autonomous agents: software systems that are able to act autonomously. This includes a wide range of physically embodied systems (e.g. robots) and systems that do not have physical embodiment (e.g. smart personal assistants). I will consider two questions. Firstly, how can we get autonomous systems to explain themselves effectively? Secondly, how can we motivate organisations to deploy explanation mechanisms? I consider motivations that can be argued to arise from existing human rights obligations. In other words, the second question can be framed as: in what situations can the right to explanation be positioned as a human right?

Note: this presentation will draw on two papers (and is joint work)

Question 1: Michael Winikoff, Galina Sidorenko, Virginia Dignum, Frank Dignum. Why Bad Coffee? Explaining BDI Agent Behaviour with Valuings. Artificial Intelligence Journal, Volume 300, 2021. doi:10.1016/j.artint.2021.103554

Question 2: Michael Winikoff and Julija Sardelić. Artificial Intelligence and the Right to Explanation as a Human Right. IEEE Internet Computing, 25(2):108-112. doi:10.1109/MIC.2020.

Biography

Michael Winikoff is a Full Professor with the School of Information Management at Te Herenga Waka - Victoria University of Wellington. He is known for his work on engineering aspects of autonomous systems, including the Prometheus methodology. More recently, he has been working on issues relating to trust in autonomous systems, including verification and explanation. He is on the IFAAMAS board of directors, was the Program Co-Chair and General Co-Chair for the conference on Autonomous Agents and Multi-Agent Systems in 2012 and 2017, respectively, and is co-Editor-in-Chief of the Journal of Autonomous Agents and Multi-Agent Systems.

Contact:

Phone:

Email: