Family Tang Hall · Columbia Engineering Innovation Hub
Organizers
- Prof. Lily XuIEOR, Columbia · Co-director, EAAMO
- Dr. David Daogainforest.earth · PL R&D
Motivation
The allocation of scarce resources to shared ends — research funding, humanitarian assistance, maintenance of critical digital infrastructure, stewardship of the global commons — is the central problem of mechanism design for social good. Across these domains, a common pattern has emerged. The populations that rely on these resources are growing, the volume of claims on them is growing faster, and the human expertise required to evaluate those claims is not growing at all. Grant panels are overloaded, scientific peer review is strained to the point of visible failure, humanitarian agencies triage under severe informational constraints, and the maintainers of open source software report being overwhelmed by the cost of reviewing contributions they did not solicit.
The diffusion of AI systems has sharpened each of these pressures simultaneously. Large language models have lowered the marginal cost of producing plausible allocation proposals (e.g. grants, papers, code contributions) while leaving the cost of their evaluation essentially unchanged. This is a structural shock to the incentive and allocation mechanisms on which public goods provision depended in the past.
Workshop aims
This workshop brings mechanism designers, theoretical computer scientists, and empirically-oriented researchers of impact evaluation and funding allocation together with practitioners who operate real allocation systems at scale. The goal is a shared research agenda — not a consensus statement, but a structured map of open problems, promising designs, and the evidence we would need to discriminate between them.
Output: a workshop report summarising the research threads and the state of the evidence.
Research questions
01 — Ex-ante versus ex-post allocation
What are the comparative properties of prospective allocation mechanisms (including quadratic and matching-fund designs) versus retrospective, impact-based reward mechanisms, under realistic assumptions about evaluator capacity, measurement error, and strategic behavior? When do hybrid schemes dominate, and at what operational cost?
02 — Impact evaluation at scale
How should the outcomes of funded work be measured, attributed, and rewarded when the volume of funded projects grows faster than the supply of qualified evaluators? What mechanisms can credibly distinguish impactful work from plausible-looking work, especially when results unfold over years — and which designs transfer across domains?
03 — Evaluation under AI-mediated contribution
How should provenance, attribution, and reviewer effort be incorporated into allocation mechanisms when a growing share of submitted material is machine-generated? What mechanisms can preserve informativeness when the cost of producing plausible submissions approaches zero, and signal scarcity must be engineered rather than assumed?
04 — Network and dependency structure
How should resources flow through networks of dependency and influence — software libraries, scientific citations, organizations delivering services to shared beneficiaries? Shapley-value and cooperative solution concepts, their computational tractability on real networks, and their interaction with existing funders’ decision processes.
Format
Invited talks and a structured working session. Speakers from mechanism design, theoretical computer science, impact-evaluation research, empirical work on scientific and development funding, and digital public infrastructure policy; plus practitioners engaged with the UN Office for Digital and Emerging Technologies and affiliated agencies.
Expected outcomes
A workshop report — published openly — that surveys the research agenda produced by the four threads, identifies tractable near-term empirical work, and names the allocation mechanisms for which better evidence is most urgently needed.
To propose a talk, register interest, or co-sponsor:
daviddao at protocol.ai