Hélöise Eloi-Hammer defends thesis on the development and use of legal artificial intelligence

Hélöise Eloi-Hammer defended her PhD thesis at Sciences Po's Médialab on 1 June 2026. Titled "The Algorithm Has the Floor: An Investigation into the Development and Use of Legal AI," the research offers a rare comparative look at how legal artificial intelligence is built and how lawyers actually use it-a intersection that remains underexamined despite its growing influence on professional practice.

The research path to legal tech

Eloi-Hammer's academic trajectory began with a preparatory course in arts and social sciences, where she "fell in love with sociology at first sight." She went on to complete a dual bachelor's degree in economics and sociology at ENS Paris-Saclay, followed by a quantitative Master's 1 programme run in partnership with ENSAE and the University of Versailles Saint-Quentin. She then shifted to a qualitative Master's 2 in "Interdisciplinary Practice in the Social Sciences" at ENS-Ulm.

"I wanted to master both quantitative and qualitative methods, whilst retaining the interdisciplinary dimension," she said. Dominique Cardon supervised her Master's dissertations and later co-supervised the PhD alongside Laurence Dumoulin, a sociologist of law who had just launched the Just-IA research project.

Tracking the life cycle of court rulings

The thesis adopted a Science and Technology Studies approach to trace what happens to court rulings after they are handed down. Through open data initiatives, these rulings are transformed into datasets that feed AI models. Those models are then used by legal professionals in ongoing cases. Eloi-Hammer's goal was to map this full cycle and examine each stage without dismissing tasks that might appear routine.

"My idea was to study what was happening at the intersection of design and use," she said. "This seemed important to me, as it is a field where professionals can be heavily involved in the development of these tools."

Fieldwork methods and access challenges

The research relied on traditional qualitative methods: interviews with developers and lawyers, observational research inside companies and law firms, and analysis of case files produced by practitioners. Eloi-Hammer also observed a tool trial conducted at a law school and supplemented her findings with archival work and reading investigations.

Access proved difficult. The companies building these tools are private, and trade secrets limit what information they will disclose. "It was tricky to gain access to the field and find companies willing to let me carry out my investigation," she said. Even after obtaining authorisation, specific technical details often remained off-limits. Internal power struggles within organisations-where investigations into AI can become weapons in factional disputes-further complicated matters.

Two main contributions for legal professionals

First, the thesis cross-examined both design and use, a practice common in Science and Technology Studies but not systematic in sociology. The findings challenge the idea that users are passive recipients of technology. "On the contrary, my thesis demonstrates that users are very active: they are involved in the design process through testing, and are by no means passive, as they are perfectly capable of repurposing the tools they use," Eloi-Hammer said.

Second, her survey of developers showed that speaking of "legal artificial intelligence" as a single category obscures real diversity. A legal search engine, depending on its programming, can produce vastly different results and different impacts on legal practice. Usage patterns vary too-how a professional deploys a tool directly shapes the legal concepts they rely on.

Advice for future PhD researchers

Eloi-Hammer cautioned against the pressure to do everything at once. "When you start your first year of your PhD and are given all this information, it's a bit overwhelming, daunting and makes it feel like an insurmountable task," she said. Her advice: focus on the research work first, and let teaching, publishing, and project involvement develop gradually.

She is currently seeking a research contract and plans to apply for positions at the CNRS and as a senior lecturer, continuing work on the topics that emerged from the thesis.

Why this matters for legal professionals

The research confirms that legal AI tools are not a monolith. A search engine's underlying programming choices can shape legal outcomes as much as a lawyer's own judgment. For practitioners evaluating or adopting these tools, the takeaway is straightforward: interrogate what is under the hood, because the same label can mask fundamentally different products. And recognise that your own usage patterns-how you integrate the tool into casework-determine its real impact on your practice.