Humans explaining self-explaining machines

Currently, a key question in AI research is how to arrive at comprehensible explanations of underlying machine processes: Should humans be able to explain how machines work, or should machines learn to explain themselves?

'The double-meaning of the name of our conference, "Explaining Machines," expresses these various possibilities: machines explaining themselves, humans explaining machines - or maybe both at the same time,' says Professor Dr. Elena Esposito. The Bielefeld sociologist is heading a subproject at TRR 318 and is organising the conference together with her colleague Professor Dr. Tobias Matzner from Paderborn University.

Dr. Matzer is a media studies researcher who is also heading a Transregio subproject. 'If explanations from machines are to have an impact socially and politically, it's not enough that explanations are comprehensible to computer scientists,' says Matzner. 'Different socially situated people must be included in explanatory processes - from doctors to retirees and schoolchildren.'

The technical and social challenges of AI projects
The organising team emphasises the interdisciplinary focus of the conference. 'It's not enough to develop AI systems solely with the expertise of a single discipline. The computer scientists who design the machines must work together with social scientists who study how humans and AI interact and under what conditions this interaction takes place,' explains Esposito. 'Today more than ever, the challenge of AI projects is both technical and social. With this conference, we hope to encourage the inclusion of perspectives and insights from the social sciences in the debates surrounding explanatory machines.'

Previously, the 'explanability' of artificial intelligence was largely the domain of computer scientists. 'In this research approach, the main view is that explainability and comprehension arise from transparency - that is, having as much information available as possible. An alternative view to this is that of co-construction,' says Professor Dr. Katharina Rohlfing, a linguist at Paderborn University and the spokesperson of Transregio.

'In our research, we do not consider humans to be passive partners who simply receive explanations. Instead, explanations emerge at the interface between the explainer and the explainee. Both actively shape the process of explanation and work towards achieving agreement on common ideas and conceptions. Cross-disciplinary collaboration is therefore essential to the study of explainability.'

Conference talks from media, philosophy, law, and sociology
The conference includes 10 short talks. Among the international, renowned guests in attendance will be Professor Dr. Frank Pasquale, an expert on legal aspects of artificial intelligence, Professor Dr. Mireille Hildebrandt, a jurist and philosopher, and Dr. David Weinberger, a philosopher of the Internet, along with sociologist Dr. Dominique Cardon, legal scholar Dr. Antoinette Rouvroy and media researcher Dr. Bernhard Rieder. Following the keynote talks, participants will have the opportunity to discuss the conference research articles directly with the presenters. These articles will be sent to conference attendees by email after registering for the event.

The conference 'Explaining Macines' is Transregio 318's first major scientific event. Information on the program can be found here. In the next three funding years, further conferences are planned that will focus on the concept of explainability from the perspectives of different scientific disciplines.

Members of the press are welcome to report on the conference: registration is required in advance by sending an email to communication@trr318.uni-paderborn.de. The conference organizers and the Transregio spokesperson will be available during the conference to answer any questions from the press.

Collaborative Research Centre/Transregio (TRR) 318
The strongly interdisciplinary research program entitled 'Constructing Explainability' goes beyond the question of algorithmic decision-making as the basis for the explainability of AI with an approach that requires the active participation of humans with social-technical systems. The goal is to enhance human-machine interaction by focusing on the comprehension of algorithms and examining this as the product of a multimodal explanatory process. The German Research Foundation (DFG) is providing approximately 14 million Euro in funding for this project through July 2025.

Related Links
Bielefeld University
All about the robots on Earth and beyond!

Tweet

Thanks for being there;
We need your help. The SpaceDaily news network continues to grow but revenues have never been harder to maintain.

With the rise of Ad Blockers, and Facebook - our traditional revenue sources via quality network advertising continues to decline. And unlike so many other news sites, we don't have a paywall - with those annoying usernames and passwords.

Our news coverage takes time and effort to publish 365 days a year.

If you find our news sites informative and useful then please consider becoming a regular supporter or for now make a one off contribution.

SpaceDaily Monthly Supporter
$5+ Billed Monthly

paypal only

SpaceDaily Contributor
$5 Billed Once

credit card or paypal

AI Improves Robotic Performance in DARPA's Machine Common Sense Program
Washington DC (SPX) Jun 23, 2022
Researchers with DARPA's Machine Common Sense (MCS) program demonstrated a series of improvements to robotic system performance over the course of multiple experiments. Just as infants must learn from experience, MCS seeks to construct computational models that mimic the core domains of child cognition for objects (intuitive physics), agents (intentional actors), and places (spatial navigation). Using only simulated training, recent MCS experiments demonstrated advancements in systems' abilities - ... read more