Event Details

Session: Trustworthy data as a foundation of trustworthy AI to support policy making

Sep 12, 2023, 01:30 PM
1 hour, 30 minutes

Conveners

Bertrand Loison (Swiss Federal Statistical Office)
Stefan Sperlich (University of Geneva)
Diego Kuonen (University of Geneva)

Speakers

Craig Burgess (World Health Organisation)
Diego Kuonen (University of Geneva)
Stefan Sperlich (University of Geneva)
Yara Abu Awad (Swiss Federal Statistics Office)

Description

What are data quality and quality management? What are the most important moments in the life of data and what do they have to do with trustworthy AI? When can you have trust and why? Why is trust not always a “given”? A general misunderstanding of modern machine learning methods is the belief that more sophisticated, flexible methods have fewer requirements on data quality. This has led to the incautious use of machine learning with “alternative data sources” and “big data” with the confidence that these methods can “deal with” any issues in the data. However, in reality it is the other way around; these potentially powerful tools often have even stronger requirements on data. This is even more true for causal analyses. We tackle this topic by first discussing trustworthy AI in the context of the Swiss federal strategy whose vision is: “human-centric and trustworthy data science and AI for public good and policy”. We will then give some specific insights regarding how we can make AI more trustworthy using certain causal inference methods. The presentations and the following round table discussion will consider these questions, related issues, and more.

Back to Top