About

Hey! I’m Víctor, a 2nd-year PhD student at CitAI Research Centre. Under the supervision of Marc Serramia Amoros and Eduardo Alonso, I'm currently working on applying Mechanistic Interpretability techniques to elicit computational representations of moral values. I've recently work on developing value system aggregation methods from different mathematical formalisations of morality. My next steps involve exploring different areas related to ensuring a safe and ethical AI development oriented by my mathematical and technical background. My interests extend to the following topics:

• Collective decision-making processes like liquid democracy, participatory budgeting, and the role of deliberation; also, collective intelligence.
• Modelling ethical decision-making (machine ethics) via value systems, and how we can obtain such value systems.
• The intersection between game theory, multiagent systems, and cooperative AI.
• The alignment problem (in AI or democratic contexts) and the control problem.

Academic journey

I hold a BSc in Mathematics with Economics from University College London and an MSc in Artificial Intelligence, where I earned distinction (85%) for my dissertation. Currently, I'm pursuing a PhD at City St George’s, University of London funded by the School of Science and Technology Studenship.

During my PhD, I have attended to conferences like IJCAI 2025 and AIES 2025, as well as workshops like FEAR 2025 and DemocrAI 2025. Also, Summer Schools like How to bypass the Turing Trap? and The 2025 Cooperative AI Summer School. Within City, St George's I serve as a Graduate Teaching Assistant and as a marker for Bachelor's and Master's dissertations. I hold the Advance HE Associate Fellowship (AFHEA), a UK-wide recognition of professional standards in university teaching and learning support.

Before my PhD, I collaborated on research projects with the Oxford AI Safety Hub Labs (2023) and the AI Safety Camp (2024) — see more technical projects here. I also took part in the Machine Learning Safety Scholar (MLSS) summer program (2022) and the Topics in Economic Theory & Global Prioritization summer program (2023), led by Phillip Trammell. Additionally, I received a grant from Open Philanthropy (2022) to foster a community around existential risks among London university students.

Values and beyond

Moral values are criteria to decide what's right or wrong. They the abstract motivations that drive our opinions and actions, yet it is difficult to conceptualize them computationally. As AI systems gain autonomy, it becomes increasingly urgent to manipulate faithful representations of morality to guide AI alignment. This raises a fundamental question: what values should AI systems align with? — and, consequently, what do we mean by “moral values” in the first place? My research builds on the value pluralism framework, where individuals hold distinct and sometimes conflicting moral values that shape their judgments of ethical actions. My plan is to keep expanding the literature of AI value alignment, with the goal of investing my time on ensuring safe deployment of technology.

Beyond academia, I enjoy running, hiking, playing frisbee and padel, reading Spanish poets, discussing philosophy and freestyle rap. I have engaged in platforms of participatory democracy , volunteered with my local Scouts community, and I have also signed the 10% pledge.