Our Projects: Hilary 2026
Modelling or visualising the dynamics of viral spread
Deforestation, agricultural intensification, and coastline destruction have well known consequences for ecology and climate. Less discussed is the resulting impacts of this environmental change on the threat of infectious disease spillover. As animal habitats are destroyed and a narrower range of habitats for humans and other mammals becomes available, it is likely that contact between human and animals will increase. My research hypothesizes that this increased contact could facilitate spillover by overcoming previous molecular and environmental barriers. There are two possible avenues of further work in this area:
1. Identify high threat species and interfaces for spillover
Utilizing influenza sequences in areas of recent environmental change, you will model the dynamics of viral spread through space and time and identify viral mutations or ecological interfaces which could be at risk for spillover. This project will feed directly into experiments in the lab and can be used to direct further research.
2. Design an interactive tool to visualize the effects of environmental change on human-animal contact
How does increasing temperature and precipitation increase the rates of human-animal contact? You will use and update previously published geospatial models of human-animal contact to integrate into an interactive user-friendly interface. The interface will integrate the models and be useful to show the effects of environmental change on spillover risk.
Hannah Klim is a Postdoctoral Researcher at Imperial College London in Virology. Hannah completed her DPhil at the University of Oxford in Clinical Medicine in 2024, where she studied immunology, epidemiology, and geostatistics. She was recently awarded an Imperial College Research Fellowship to pursue her own independent research on the causes of viral spillover with a view to improving pandemic prevention.
Analysing AI incident scenarios and emergency responses
This project investigates potential AI incidents and how governance mechanisms and emergency procedures could be used to prevent them. It will explore concrete incidence scenarios and existing safeguards against those. Based on this it should develop proposals for companies and/or governments on emergency procedures that could be implemented to contain these incidents. The focus can be on technical exploration of these possible incidents or on more broader procedural exploration, depending on the interest and expertise of the group.
Sven is a Senior AI Policy Fellow at the Institute for AI Policy and Strategy and works in the European AI Governance team at The Future Society. Before that, he worked as Head of Research Operations at a research institute at Oxford University. He also worked at a non-profit in the sustainability sector and as a management consultant. He has a PhD in mathematics.
Discovering nudges for effective giving
Despite the widespread availability of effective giving information, only a negligible portion of the funds donated worldwide go to expert-recommended effective charities. To date, the main approach implemented at scale to encourage effective giving has been the creation of education and pledging-focused organisations, like Giving What We Can (UK), Mieux Donner (FR), Doneer Effectief (NL) and others. However, for a variety of behavioural and psychological reasons, motivating effective giving through these means has its limits. This project aims to create a synthesis of potential psychological drivers and mediators and (2) actionable design recommendations for effective-giving platforms.
Antilia is completing a DPhil in Economics at Oxford and is a co-founder of Good Wallet, an effective giving project. Her research focuses on the behavioural determinants of effective charitable giving, combining microeconomic theory and experimental methods to study how donors respond to different incentive schemes and choice architectures.
Using semantic entropy to detect unfaithful reasoning
Current frontier AI models often produce natural language “chain-of-thought” reasoning before acting. This provides an affordance to monitor their intermediate computations, and to detect undesired motivation and planning. However, there are many examples where this chain-of-thought can be unfaithful. If we could detect when reasoning steps are post-hoc rationalisations, this could enhance such monitoring. This project would investigate if semantic entropy (or more efficient semantic entropy probes), shown to be effective at detecting hallucination, can also be used to detect unfaithful reasoning steps.
Avi is DPhil student, funded to work on technical AI governance by the AI Governance Initiative at the Oxford Martin School. His previous research has focussed on the effect of stochasticity in training algorithms with the Pivotal fellowship, and the impact of adaptive optimisation on training dynamics in deep neural networks.
Are AI systems ready to provide early warning of critical AI risk?
AI capabilities are assessed more thoroughly after deployment using evaluation methods that didn’t exist while the model was being developed. As AI systems grow more capable, will our evaluation infrastructure keep up?
To assess the above question, we’ll create a dataset that tracks key metrics from frontier AI company safety reports alongside the broader literature on evaluating AI systems for dangerous capabilities. The dataset will track risk thresholds labs identify as critical, evaluation coverage across risk domains, methodology evaluation, and timing patterns. We’ll analyse the data as we go to answer some of the following questions: (i) what risk thresholds do organisations identify as concerning? (ii) On current trends, how prepared will AI safety reporting systems be to handle emerging risks over 2026-2030?
Paolo is a PhD candidate at Teesside University supervised by Professor The Anh Han. His research studies the dynamics of tech races and the effective design of early warning systems for AI risk. He has built software tools as part of Modeling Cooperation, including an interactive web app to explore tech race scenarios and an app used to facilitate the Intelligence Rising workshops that teaches ideas in AI governance to key decision makers.
Defining Intelligence Recursion
Many experts are particularly concerned about AI models that can substantially automate their own development, allowing for increasingly rapid increases in capabilities. This project examines how to tell when such intelligence recursion has begun. Despite its central importance to governance proposals like MAIM, no consensus definition of this concept exists. Candidates range from fraction of AI-written code and compute thresholds to shares of cognitive R&D work done by AI and observed capability gains. The project will evaluate these definitions to recommend which one(s) governments like the US should adopt. It aims to be the authoritative source for when drastic actions against competitors are justified—not too early, not too late.
Nicholas is an Ellison Scholar at the Ellison Institute of Technology, an Expert Collaborator at the MIT AI Risk Initiative, and a General Committee Member at the Oxford AI Safety Initiative. He works on AI policy and experimental capability evaluations. He has reviewed papers for NeurIPS, collaborated with the IMF to improve World Bank forecasts, and researched a report whose recommendation was adopted by the US and UK governments.
Modelling Land Value Tax in Cape Town
When you tax something, you get less of it. That’s why we tax tobacco. But most taxes fall on things we want more of: income, labour, housing, investment. This shrinks the economy.
Land is the exception. Its supply is fixed. You cannot reduce the amount of land by taxing it. Economists call this “zero deadweight loss.” Milton Friedman called it “the least bad tax.” Joseph Stiglitz - who disagrees with Friedman on almost everything else - demonstrated the “Henry George Theorem”, showing that public investment increases land values by an equivalent amount. This is the economist’s favourite tax that politicians have seldom implemented.
So why don’t we have it? Because landowners vote, and they may not know what they’d gain. We hypothesise that most suburban homeowners would actually pay less under a Land Value Tax (LVT), with the burden falling on speculators and owners of underutilised prime land. But nobody has ever shown voters their personalised “shadow bill.”
This project builds that tool. For the first time, we will show every household in Cape Town exactly how their taxes would change under LVT. This is the missing piece for political viability.
Peter Courtney is a joint PhD candidate in economics at VU Amsterdam (Tinbergen Institute) and Stellenbosch University, on a research stay at the Blavatnik School of Government. He holds the 2025 C. Lowell Harriss Dissertation Fellowship from the Lincoln Institute of Land Policy, an Oppenheimer Memorial Trust Fellowship, and an Emergent Ventures Grant. Publications with Oxford University Press and Brill; policy work for UNU-WIDER and South African National Treasury. Previous positions at J-PAL Africa and UN FAO. More info available here.
Mapping Fever Hotspots to Support Vaccine Trials
This research project will be a descriptive study to map high-risk RVF regions in West Africa (particularly Senegal, the Gambia and Mauritania), collating seasonal livestock movement with environmental drivers and livestock and human distributions. Furthermore, to identify where elevated risk may coincide with surveillance “dead zones.” The work will involvereviewing the literature on migration and trade networks, scraping, collating, and processing publicly available datasets to construct spatial and temporal maps and explore potential hotspots.
This project will directly contribute to our broader RVF forecasting framework and have clear applied impact by supporting local partners in identifying and prioritising potential sites for future RVF vaccine trials.
I am an infectious disease epidemiologist and modeller focused on understanding and mitigating the spread of pathogens in populations. My research involves modelling transmission dynamics and optimising vaccination strategies, including how vaccines are evaluated under dynamic, outbreak-driven conditions.
I am currently a postdoctoral researcher in Professor Fraser’s Pathogen Dynamics Group at the Pandemic Sciences Institute, contributing to modelling vaccine trial designs for WHO pandemic priority pathogens, with a particular focus on Rift Valley fever.
