Philosophy Personal Statementfor Cambridge

When I read about the AI Safety Summit at Bletchley Park in November 2023, what stayed with me was not the language of risk but the thought behind it: once a system becomes sufficiently intelligent, we start speaking as if its decisions deserve moral weight. I wanted to know what would justify that change. Intelligence alone seemed too thin a criterion, but consciousness, intention and responsiveness to reasons were even harder to define. That difficulty drew me to philosophy, because it turned a news story about technology into a deeper question about mind and personhood. I now want to study philosophy because it gives me a way to examine those assumptions rather than hide them inside technical language. At university I want to look more closely at ethics and philosophy of mind, especially where claims about artificial systems rely on contested ideas about consciousness, agency and responsibility.

My academic study has helped me test that interest against arguments rather than intuitions. Our work on utilitarianism first attracted me because it promised a clear procedure for hard choices, but that appeal weakened when I started looking at sacrificial cases. Reading Philippa Foot's 1967 essay on double effect, then Judith Jarvis Thomson's reformulation of the trolley problem, made me realise that moral judgement is not only about outcomes. Who acts, what they intend and what kind of duty is being violated all matter. Setting this beside Kant's claim that persons must be treated as ends rather than means made the disagreement sharper, and Bernard Williams' critique of utilitarianism made it harder for me to ignore what gets lost when morality is reduced to an impersonal calculation. I had begun by wanting a theory that would settle difficult cases quickly; these readings made me suspicious of that wish. I explored the issue more systematically in my EPQ on whether autonomous vehicles can make moral decisions or only carry out moral instructions written by others. The project improved once I organised it around one distinction: a machine can follow a decision procedure without being a moral agent. I used that distinction to compare utilitarian models, deontological constraints and a virtue-ethical criticism that collision cases strip away too much context. My first draft tried to cover every theory and became descriptive. The argument only became clearer when I narrowed it to one claim: responsibility does not move from human beings to machines simply because a response has been automated, since the variables, thresholds and priorities have already been chosen by people. That made me treat trolley-style examples as useful tests of principle, but poor substitutes for the uncertainty of real roads.