Research.

I'm completing an MA in AI, Ethics and Society at Birkbeck, University of London, graduating in 2027. My research asks, at base, who AI systems serve and what assumptions are built into them. One essay argues that AI systems are inherently political, in the sense that the philosopher Langdon Winner used the term: artefacts that originate in minds embedded in social contexts, and carry those contexts into the world. The case study is COMPAS, a criminal risk assessment algorithm used in American courts. COMPAS assigns higher risk scores to Black defendants at nearly double the rate of white defendants who did not reoffend. The essay argues this is not a flaw or an accident. It is built into what COMPAS is. Treating it as incidental is what allows the harm to continue.

A second essay looks at how regulation is responding. The dominant approach across the EU, UK, and United States is transparency: require companies to disclose how their systems work, what data they were trained on, and what risks they present. The essay argues that this is not producing accountability. Most disclosures cannot be evaluated by the people they are meant to reach. When companies define the content of their own disclosure, it tends toward managing reputation rather than informing the public. And transparency without enforcement does not change behaviour. The fragility of this approach became clear in January 2025, when the incoming Trump administration rescinded Biden's AI executive order within hours of taking office.

The third project is a dissertation proposal on science fiction imaginaries in AI systems. AI companies reference fiction more explicitly than is usually acknowledged: Asimov's Three Laws appear in robotics governance documents, superintelligence scenarios structure safety research, and cyberpunk aesthetics shape product design. The proposal asks what actually happens when fictional narratives shape real technical infrastructure. Does the fiction become reified as natural fact? Does it materialise into physical systems? And which imaginaries operate without anyone consciously choosing them, shaping what counts as good AI before the question of governance is even asked?

If those assumptions are invisible, they bypass democratic deliberation. The research is trying to make them visible and contestable. I'm open to conversations with researchers or practitioners working on adjacent questions.