Approach
Four connected specialisms, one habit of mind: respect the system, respect the data, and let the question choose the method.
When system behaviour emerges from many interacting individuals, model the individuals. ABMs make assumptions explicit and testable, support policy experiments you could never run in the real world, and give stakeholders something they can see and question.
Reach for it when
Prediction and pattern-finding done carefully: leakage-proof validation, uncertainty quantification, and models chosen for the problem rather than the fashion. Deep nets when they earn their keep; simpler models when they don't.
Reach for it when
For sequential decisions in systems with feedback — allocation, scheduling, adaptive strategy. Often paired with classical numerical optimisation: the optimiser handles hard constraints, the learned policy handles judgement.
Reach for it when
Networks, nonlinear dynamics, emergence and tipping points — the mathematical backbone underneath everything else here. This is the lens that connects a fishery, an epidemic and an organisation.
Reach for it when
The raw material
Ecological and medical data share a family resemblance: irregular sampling, observation bias, missingness that carries meaning, and ethical stakes. Rusty Data treats the data pipeline as part of the science — provenance tracked, assumptions documented, uncertainty propagated rather than deleted.
Comfortable with
How an engagement runs
Before any code: what generates this data, who acts in this system, and what decision will the model actually inform?
Reproducible data work with uncertainty carried through — the unglamorous half of every successful project.
The right method for the question, validated in ways that would survive a hostile reviewer.
Code, documentation and a model your team can interrogate and extend — not a black box with an invoice.
One practice, many strands
Methods, domains and projects feed one another — each strand of the practice draws on the rest.