Odyssean Institute
Inaugural Research Fellowship
Get notified of more jobs like this · No spam, ever
How your CV stacks up
Upload your CV to see how well it fits this job role
?%
Inaugural Research Fellowship
Modelling Expert Mental Models of AI Safety
The Odyssean Institute is opening applications for three inaugural Research Fellows to work on a novel methodological project at the intersection of AI safety, complex systems, and LLM-assisted research. This is a three-month engagement with a stipend of €5,000 per fellow and an expected commitment of 20 hours per week or more.
About the Odyssean Institute
The Odyssean Institute is a UK-based research charity. We help people and institutions navigate complex problems in times of crisis and pre-empt future crises, by combining expert knowledge, computational foresight, and participatory decision-making. We call this approach the Odyssean Process, and we use it to address existential and global catastrophic risks, including risks from advanced AI.
Our work draws on complexity science, exploratory modelling, and deliberative mechanisms such as the IDEA Protocol and Citizen Assemblies. Recent projects include a horizon scan of global catastrophic risks, work on AI safety regulation informed by public deliberation, extreme resilience foresight (i.e. nuclear impacts), and research on nodes of persisting complexity for global resilience.
Our team is global and transdisciplinary, with backgrounds in history, political theory, philosophy, earth science, economics, complex systems modelling, sociology, and the arts. We have academic and industry experience from the University of Oxford, the University of Cambridge, the Australian National University, the National University of Singapore, Goldsmiths University of London and beyond.
About the Fellowship
This is the inaugural cohort of the Odyssean Institute Research Fellowship (AI Safety Stream). You will help shape a new programme from the ground up and work on a genuinely novel research problem.
We have collected a substantial body of expert insight on the future of AI safety through our horizon-scanning and elicitation work. The next step is to surface the causal mental models that experts hold about how AI safety unfolds. That means identifying the variables experts see as consequential and the relationships they imagine between those variables, then representing those models in a form that can be reasoned about, deliberated on, and stress-tested.
Fellows will help us do this in four ways.
- Automating thematic coding of expert testimony to identify causal variables and the relationships between them, using LLMs as a core part of the tooling.
- Building causal loop diagrams, and where warranted stock-and-flow models, that render the community's mental models of AI safety visible and comparable.
- Contributing to an open-source project that makes these methods reusable by other researchers working on complex-risk questions.
- Supporting expert elicitation and workshops where relevant, including helping to design and run sessions and integrating outputs back into the modelling.
Reasons to use Rodeo
Start with a chat, not a search bar
Grad scheme, placement, apprenticeship? Not sure what you want yet — that's fine. Your agent talks it through with you and turns "I have no idea" into a shortlist.
It searches the market for you
Every day your agent scans the market matching roles against what actually matters to you, not just keywords on a CV.
Only hits
No noise. No "maybe this fits." Just roles with a clear explanation of why they're right — and where to focus when applying.
This is exploratory work with no fixed recipe. We expect false starts, iteration, and the kind of judgement calls that come with building something for the first time. We want fellows who are excited by that.
What You'll Do
- Develop and refine LLM-assisted pipelines for thematic coding of qualitative expert data.
- Extract causal variables and hypothesised relationships from coded material.
- Build, iterate on, and document causal loop diagrams and, where appropriate, stock-and-flow models representing expert mental models.
- Contribute code, documentation, and methodology write-ups to an open-source repository.
- Participate in team research discussions and, where relevant, help design and support expert elicitation sessions and workshops.
- Submit a weekly timesheet tracking hours and progress.
Who We're Looking For
We're looking for three fellows whose combined skills cover the range below. You do not need every one of these. Tell us what you bring.
Required experience
- Mixed-methods research, with hands-on familiarity with thematic coding of qualitative data.
- Scenario development and analysis.
- Complex systems analysis methods, particularly causal loop diagrams and stock-and-flow modelling.
- Experience using LLMs to build open-source projects, including prompting, orchestration, and evaluation of outputs in a working pipeline.
- Comfort with methodological experimentation and ambiguity. You are willing to try approaches, discard what does not work, and iterate.
Nice to have
- Quantitative risk estimation methods.
- Expert elicitation methods such as the IDEA Protocol or structured expert judgement.
- AI safety, existential risk, or global catastrophic risk research.
- Prior contributions to open-source communities.
You'll thrive here if you are
- Excited by work that sits between qualitative interpretation and formal modelling.
- Happy to work independently with light-touch coordination while contributing to a small team.
- Motivated by using better methods to help societies make better decisions under uncertainty.
Get help with your application
Your very own career expert that helps elevate your application to the next level.
Details
- Three positions available.
- Stipend of €5,000 per fellow for the three-month engagement.
- Time commitment of 20 hours per week or more. This may include additional support for expert elicitation and workshops.
- Weekly timesheets tracking hours and progress.
- Location and working arrangements are remote with option for hybrid collaboration if in the USA or UK.
- Application deadline: 11th September
- Start date: 14th September
How to Apply
Please attempt all of them, but honest, partial answers are welcome. If a method is new to you, tell us how you would approach it rather than leaving it blank. We are reading for judgment and how you think, not for polished or complete work. Please limit AI use to review and grammatical edits.
Please send us the following as a single PDF to operations@odysseaninstitute.org, and contact@odysseaninstitue.org
-
A sketch of an experiment. Pick one of the four tasks described in the fellowship (for example, using an LLM to code expert testimony for causal variables and relationships). In roughly one page, outline how you would run a first small experiment to test whether the approach works. We are not looking for a detailed or finished plan. We want to see how you would set it up, what you would measure to know if it is working, and the one design decision you are least sure about.
-
Evidence of mixed-methods work. Attach or link to something you have produced that involved thematic or qualitative coding of data. A paper, a preprint, a thesis chapter, a report, or a public write-up is all fine. In two or three sentences, tell us what your specific role was in the coding and analysis.
-
Working under deep uncertainty. Imagine you have a causal loop diagram built from expert testimony about how AI safety might unfold. It is a static picture of variables and their links. In a few short paragraphs, how would you begin to explore that model under uncertainty, given that experts disagree and the future is genuinely open? What would you vary, and what would it mean for one of these models to be robust or trustworthy? You do not need familiarity with any particular framework or toolset to answer this well.
“It took my CV and asked me questions relevant to understanding what kind of jobs to suggest for me. Suggestions were almost perfect. Jobs were exactly what I’ve been looking for.”
Jessica, London
Skills
Location