The computer says no. But can I believe him?: Thoughts on auditing artificial intelligence

2022-04-05T12:00:00 - 2022-04-05T13:00:00
Tuesday, April 5, 2022
12:00 pm - 1:00 pm PDT

Anni Hellman

Sponsored by the Center for Long-Term Cybersecurity. Co-sponsored by the UC Berkeley ​Institute of European Studies, the Berkeley Institute for Data Science (BIDS), and the UC Berkeley Center for Human-Compatible Artificial Intelligence.

An audit is a tool used to verify compliance with regulations and controls. This requires in-depth analysis to understand the functioning of the object of the audit. If the object of the audit is an artificial intelligence-based system, understanding how the system came up with its results may be a challenge: the AI may be very complex, or its operation may be based on training, which it carries out through an iterative process that we may not be able to observe. How then do we know whether the AI is treating us fairly and that its assessments are sound and objective?

The European Commission has issued a proposal for an Artificial Intelligence Act to the European Parliament and Council. This Proposal comprises several requirements specifically for AI-based software that would be used in legislative areas considered to be “high risk,” as they could “create adverse impact on people’s safety or their fundamental rights.” How can compliance with these requirements be verified, and can that be done ex-post, or is an ex-ante audit necessary?

Register Now

This seminar will be held both online & in person. You are welcome to join us either in South Hall or via Zoom.

For in-person participants

Face masks are strongly recommended.

For online participants

Online participants must have a Zoom account and be logged in. Sign up for your free account here. If this is your first time using Zoom, please allow a few extra minutes to download and install the browser plugin or mobile app

Anni Hellman has been a deputy head of unit at the European Commission and is visiting UC Berkeley for the academic year 2021–2022 as an EU fellow. She is a mathematician, a qualified insurance actuary, and a fellow of the Actuarial Association of Finland, and holds degrees in mathematics, actuarial sciences, and computer sciences.

In the European Commission, her unit (media convergence and social media) has focused on the challenges of social media, with two main strands or work: policy on how to tackle the challenges of disinformation, and how to work towards ensuring the integrity of social media through a multidisciplinary approach and through funding of research projects.

At Berkeley, Anni looks into the challenges of how and whether the trustworthiness and fairness of artificial intelligence and algorithms could be verified: Can AI be audited? Anni aims to combine her backgrounds in mathematics, audit practices and experience in working on and with the European legislative instruments.

Last updated: March 18, 2022