Governments around the world are using predictive systems to manage engagement with even the most vulnerable. Results are mixed.

More like this: Algorithmically Cutting Benefits w/ Kevin De Liban

Luckily people like Soizic Pénicaud are working to prevent the modern welfare state from becoming a web of punishment of the most marginalised. Soizic has worked on algorithmic transparency both in and outside of a government context, and this week will share her journey from working on incrementally improving these systems (boring, ineffective, hard) — to escaping the slow pace of government and looking at the bigger picture of algorithmic governance, and how it can build better public benefit in France (fun, transformative, and a good challenge).

Soizic is working to shift political debates about opaque decision-making algorithms to focus on what they’re really about: the marginalised communities who’s lives are most effected by these systems.

Further reading & resources:

• The Observatory of Public Algorithms and their Inventory
• The ongoing court case against the French welfare agency's risk-scoring algorithm
• More about Soizic
• More on the Transparency of Public Algorithms roadmap from Etalab — the task force Soizic was part of
• La Quadrature du Net
• France’s Digital Inquisition — co-authored by Soizic in collaboration with Lighthouse Reports, 2023
• AI prototypes for UK welfare system dropped as officials lament ‘false starts’ — The Guardian Jan 2025
• Learning from Cancelled Systems by Data Justice Lab
• The Fall of an Algorithm: Characterizing the Dynamics Toward Abandonment — by Nari Johnson et al, featured in FAccT 2024

**Subscribe to our newsletter to get more stuff than just a podcast — we host live shows and do other work that you will definitely be interested in!**