Scientific Visibility Is a Coordination Problem — An Experimental Approach

Important scientific and technical work can remain obscure because it is difficult to discover, difficult to evaluate, badly indexed, poorly presented, or disconnected from the people who could use it. Science DAO treats this as a coordination problem rather than claiming to have a unique or proven solution.

Our working hypothesis

We are testing whether a combination of AI-assisted evaluation, public evidence, transparent funding records, and better science communication can help useful work receive more attention and support. This is an experimental hypothesis.

What AI can and cannot do

AI can help summarize, classify, compare, and retrieve work. It can also make serious mistakes. It cannot turn an unsupported claim into a validated result, and it cannot substitute for domain expertise where expert verification is required.

Why visibility matters

Research that cannot be found is less likely to be read, criticized, replicated, cited, funded, or used. Better metadata, open publication, structured evidence, and responsible science communication can reduce this problem.

Why visibility is not enough

Visibility can amplify both good and bad work. For that reason, Science DAO’s approach should connect discoverability to evidence: source documents, code, datasets, preregistrations, reproducible procedures, known limitations, and independent reviews.

Current status

AIIM is an experimental funding system in beta. Its evaluations are heuristic judgments rather than validated measurements of scientific value or causal impact. On-chain payment records are available, while parts of the current implementation and governance remain off-chain or administratively controlled.

Correction regarding Navier–Stokes

A previous claimed solution of the Navier–Stokes problem by Victor Porton was found to be incorrect. Science DAO does not treat that proof attempt as a valid result. This correction does not by itself resolve the status of separate mathematical work such as Discontinuous Analysis, which must be evaluated on its own definitions, proofs, publication record, and independent review.

What would count as evidence of progress?

• Preregistered evaluations and adversarial tests
• Public datasets and reproducible procedures where appropriate
• Independent reviews, including critical reviews
• Documented failures and corrections
• Evidence that allocation decisions are stable, fair enough for their stated purpose, and resistant to manipulation

See Testing & Evidence and Independent Review.