Automated systems in public administration affect real decisions. My work focuses on what accountability actually requires in practice: the impact assessment tool I developed with AlgorithmWatch helps public authorities structure this — from triage to transparency reporting — for use across public services and companies.
Fairness in ML sounds precise but conceals contested normative choices. My research takes these seriously — especially in insurance and health settings where predictions directly shape access to care and coverage. The work exposes the tradeoffs hidden inside algorithmic decisions about healthcare and insurance policies.
Large platforms don’t just carry content — they shape what is visible, amplified, or suppressed. My work with AlgorithmWatch studied the systemic risks these platforms pose to democratic discourse and developed methodologies for identifying them — going beyond individual violations to structural effects on society.
With AlgorithmWatch and Canton Zurich, I developed practical guidelines for evaluating automated decision systems in public administration — what questions authorities need to ask, what documentation is required, and how to build in meaningful oversight. The goal: transparency and accountability that holds up in practice, not just on paper.
I help organizations figure out what changes when the work is done with AI — and what should stay the same.
Most companies adopting AI focus on what it can do. Few are asking who is accountable for what it produces, how decisions get documented, or what happens when something goes wrong. These are not compliance questions. They are operational design questions.
I bring 10 years of academic research on trust, transparency, and bias in AI — and practical experience building assessment tools, fairness methodologies, and governance frameworks with institutions like AlgorithmWatch, Politecnico di Milano, and the University of Zurich. Now I work with organizations directly, designing governance layers that let professionals use AI without losing control over their output.
If you’re thinking about how to structure human oversight in your AI workflows — not in theory, but in practice — I’d like to hear about it.
A governance layer for professionals who use AI — and need to remain responsible for the results.
The complete collection of not-yet-peer-reviewed ideas
The complete collection of my publications by categories: journals, conference proceedings, books, and book chapters
The complete collection of my publications by categories: journals, conference proceedings, books, and book chapters
