Research
My current research is on social intelligence in artificial systems.
I'm interested in how agents understand one another, remember past interactions, form relationships, coordinate in groups, and eventually develop larger social structures. Much of my current work uses multi-agent systems to study these processes computationally.
Alongside this, I've worked on questions in AI safety, particularly around compute governance and the incentives created by different regulatory systems.
Social intelligence in artificial systems
What does an artificial agent need to remember about another agent?
My current work looks at how artificial agents maintain social context over time: how they represent other agents, reason about different perspectives, remember previous interactions, and adapt as relationships change.
I'm particularly interested in social continuity. Relationships depend on history: who knows what, what happened before, what one agent expects from another, and how those expectations change through repeated interaction.
I'm currently developing Relational Belief States (RBS) as one approach to this problem. Most of that work is still private while I develop and test it.
I write more openly about the surrounding questions, literature and experiments on Synthetic Society.
Selected research
SplitComp: Stress-Testing Compute Governance Under Strategic Evasion and Jurisdictional Fragmentation
When does enforcement produce compliance, and when does it push strategic actors towards jurisdictional fragmentation?
We modelled a lab choosing between complying, evading within a single jurisdiction, or splitting activity across several jurisdictions. The project studies how audit probability, detection accuracy, penalty severity and cross-border coordination shape that decision.
One of the main findings is that stronger domestic enforcement does not necessarily produce compliance. When cross-border coordination is weak, it can instead make jurisdictional fragmentation more attractive.
Notes
Some of my ongoing research stays private while it is still being developed. I use Synthetic Society for more exploratory writing: field notes, literature, experiments, unfinished questions and ideas around social intelligence and multi-agent systems.
If you're working on similar problems, please feel free to get in touch. If you're based in Ireland, especially so — there are very few of us here, and I'd love to know more people working in this area.