ZiF Workshop

Explanation and Understanding in the Age of Algorithms

Date: 6 - 9 November 2018
Convenors: Mark Alfano (Delft, NED), Ignaz Rutter (Passau, GER), Dario Martinelli (Kaunas, LTU)

Sophisticated yet opaque algorithms are encroaching on every aspect of contemporary life. They help people find information, make friends, navigate the surface of the Earth, decide what to buy and sell, decide whom to hire and fire, regulate traffic flows, negotiate contracts, predict epidemics, make medical diagnoses, and identify and track criminals and terrorists. Until recently, such activities were the exclusive domain of human decision-making. Our epistemic, ethical, and political capacities enable us to engage in such activities and-in the ideal case-explain our decisions to the people they affect, to the general public, and to ourselves. Human dignity demands that people understand how decisions that significantly affect their own and others' lives are made. This is one reason why the EU recently enshrined the right to explanation in the General Data Protection Regulation. Yet the most effective algorithms currently in use are not fully understood even by the people who design, build, and maintain them. We are at a crossroads: either pursue and use algorithms that make better decisions but that we do not understand and cannot explain, or settle for algorithms that we do understand and can explain but that make less effective decisions. This workshop will bring together philosophers, psychologists, and computer scientists to explore this tradeoff between dignity and utility.

In early November 2018, we convened a workshop entitled 'Explanation and Understanding in the Age of Algorithms' at the Center for Interdisciplinary Studies (ZiF). Participants came from all around the globe, including Europe (Germany, the Netherlands, Lithuania, Scotland, Bulgaria), the United States, and Australia. The topic of our workshop was quite timely, as the public and policy-makers are beginning to worry more and more about fake news, self-driving vehicles, cryptocurrencies, recommender systems, expert systems, and various sorts of Artificial Intelligence. Are these tools trustworthy? What would it mean for them to be trustworthy? Arguably, part of the answer to this question relates to the extent to which they can be understood and their workings explained. However, the most powerful algorithmic systems are also in many cases the hardest to understand and explain. For example, the workings of artificial neural networks with many hidden layers of artificial neurons are often opaque even to their designers—let alone non-experts. To address this issue, we brought together a diverse group of researchers at various career stages from the fields of philosophy (epistemology, ethics, political philosophy), computer science (human-machine interaction, fairness and transparency, network theory), and semiotics (cultural meaning, symbolic interaction). Together, we addressed both descriptive and normative issues. What sorts of algorithms currently exist, and what sorts of algorithms are likely to be developed in the near future? What sorts of auditing, transparency, accountability, and understandability are desirable from these systems? We made some progress in thinking through these questions by narrowing our focus over the course of four days to the problem of trustworthiness and the wise placement of trust. While we are far from having arrived at definitive answers to our questions, we see a promising way forward by applying for a planning grant from the VolkswagenStiftung program on 'Artificial Intelligence and the Society of the Future'. We really enjoyed our time at the ZiF and are extremely grateful for the opportunity to think together across disciplinary boundaries in a comfortable and welcoming environment.

Teilnehmerinnen und Teilnehmer

Kristian Bankov (Sofia, BUL), J. Adam Carter (Glasgow, GBR), Marc Cheong (Clayton, AUS), Jeroen de Ridder (Amsterdam, NED), Giulio Mecacci (Delft, NED), Alan Rubel (Delft, NED), Filippo Santoni de Sio (Delft, NED), Hanno Sauer (Utrecht, NED), Emily Sullivan (Delft, NED), Nava Tintarev (Delft, NED), Suresh Venkatasubramanian (Salt Lake City, USA)

Conference Programme

Please direct questions concerning the organisation of the workshop to Trixi Valentin at the Conference Office. Questions regarding scientific content and contributions should be directed to the organizers.

Tel: +49 521 106-2769
Fax: +49 521 106-152769
E-Mail: trixi.valentin@uni-bielefeld.de