世 光 · 世 应 实 验 室
SEESALT ADAPTATION LAB
Where adaptation finance
stops looking
世应 = 适应
Shìyìng (World Resonating) = Shìyìng (Adaptation). One sound, two meanings: adaptation emerges when the world resonates together.
A flood or a dam failure can be read as a climatic extreme or as a governance failure, and the reading is settled after the fact. States anticipate this. They build visible response capacity, which gets credited, and underbuild quiet prevention, which does not. The instruments that decide where adaptation finance goes do not keep the two apart at the point of decision.
What we build
Open-source diagnostics, released as digital public goods
A separating diagnostic
A published protocol that reports a country's disaster-governance capacity as a pair — response side and prevention side — with a confidence interval for each, rather than a single score. Coding rules and judgement criteria are public, so the result can be reproduced and contested.
A multi-hazard data layer
Evidence of prevention rarely appears in what governments report. It appears in land-use change, infrastructure condition, hazard history and exposure. The data layer is the measurement floor the separation requires — not a demonstration.
We do not publish country rankings. A ranking reinstates the aggregation that erases the distinction in the first place.
Where the distinction dies
A narrower claim than we first made
The INFORM Risk Index — scientific lead, the European Commission Joint Research Centre, developed with the IASC Reference Group on Risk, Early Warning and Preparedness, covering 191 countries — measures a "lack of coping capacity" dimension built from two categories: an institutional category covering the existence of disaster-risk-reduction programmes aimed largely at mitigation and preparedness, and an infrastructure category measuring capacity for emergency response and recovery.
Two things follow. The distinction exists in the model and is then removed by aggregation — both categories are combined into one score. And the prevention side is measured by whether programmes exist, not by whether they work.
A correction we owe the reader. We used to say no instrument makes this distinction. That is too strong. The WorldRiskIndex separates lack of coping capacities from lack of adaptive capacities explicitly, and it predates us. What survives is narrower: it cuts on time horizon where we cut on attributability, the separation is still collapsed into a composite at the point of decision, and no instrument we have checked measures the prevention side by effect rather than by existence. The full correction and the prior art are set out in the method.
How to prove us wrong
Falsification conditions
A framework that only works on its home case is a description, not a theory. Three ways this one fails:
- If states facing high attribution ambiguity show no systematic difference in the ratio of prevention to response investment, the claim is wrong.
- If response and prevention capacity cannot be distinguished in obtainable data, the tool cannot be built regardless of whether the claim is right. This is our largest technical risk.
- If practitioners tell us they already make this distinction and simply have not written it into public methodology, then we should be doing something else — or nothing.
The third is the priority for the next six months. If you assess adaptation finance for a living and the gap we describe does not exist, we would rather hear it now.
Where we are
Stated precisely, on purpose
- Our multi-hazard data overlay was named a global champion in Challenge 2 of the UN Tech Over 2025 open innovation challenge, with Tiago Maluta. Code released openly in June 2025, with no measured downstream adoption since. What we built, and what the result did not establish.
- The methodology paper, Attributive Ambiguity: How Anticipated Deniability Deforms State Disaster Capacity, has been accepted for presentation at the AAG Annual Meeting (New York, 8–12 February 2027). Abstract and cases.
- No revenue, no funding, and no paying clients. Costs are absorbed by volunteer time. We have also not yet spoken to a single buyer — no conversations with adaptation funds, MDBs or implementing partners have taken place. That is the next six months' work.
- No completed diagnostic tool, and no partnership or implementing agreement with any UN agency or multilateral institution. Adaptation Lab is a project team within SEESALT, not a separate legal entity.
Who we are looking for
Open roles
Global health security
Someone who knows the JEE, IHR core capacity assessments and the GHS Index literature. The same attribution logic appears in pandemic preparedness, and that line of argument is currently an outline.
Remote sensing & geospatial
Someone who can take the data layer from competition prototype to reusable tool. The hard part is not modelling — it is testing whether prevention leaves an observable proxy at all.
Development finance & procurement
Someone who knows what capacity indicators GCF, the Adaptation Fund and the MDBs actually cite in eligibility assessment. This decides how large the gap is, or whether it exists.
In-country researchers
Researchers based in the case countries. Attribution contests happen in local languages and local politics; remote coding misses the part that matters.
You do not need to be at Yale, or a student anywhere. We recruit from SEESALT's 135-person consultant pool and from outside it on the same terms.
What we can offer: co-authorship, an open-source contribution record, shared design authority over the methodology, and a working environment that does not hide its own limits. What we cannot offer: salary. Better said before a first conversation than after.
If you do none of these but know someone who does, an introduction is the most useful thing you can send us.
Read further
Four pages, in the order they were written
Premises →
Why we treat institutional ambiguity as the medium of governance rather than as measurement error.
Method →
The protocol at v0.1: prior art with sources, candidate indicators, and the condition that would disqualify each one.
UN Tech Over →
The build that started this: inputs, weights, code, and what winning did not establish.
AAG 2027 →
The paper, its three cases, and an open request to session organizers before November.
Get in touch — or tell us we are wrong
Zhijun He — zhijun.he@yale.edu
Research programme: seesaltorg.org/zhijun/research
播 种 光 明 ， 收 获 永 恒