Forecasting markets commission replication experiments

What the forecast means

Milestone stage: integrated service demonstration

replication forecasting demonstrated; funding-to-experiment integration not established

Timing window: 2028–2042. An editorial range, not a statistical confidence interval.

Completion criteria: A public platform uses preregistered prediction and budget rules to procure at least 20 independent replication experiments across two disciplines and settle forecasts against audited outcomes, disclosing unsuccessful and inconclusive runs. Resolve only from inspectable evidence meeting every stated criterion; a plan or extrapolation does not resolve it.

This page describes a forecast, rather than an achieved result. Read the methodology and assumptions.

Built on

• Cloud labsRecorded · 2014enables
• Scientific replication-market study publishedRecorded · 2015enables
• Interoperable workflow evidence packagesRecorded · 2024enables
• AI research partnersRecorded · 2025scales with
• Automatic replicationForecast · ~2031scales with
• AI theories survive adversarial experimentsForecast · ~2033influences

Led to

• All science machine-readableForecast · ~2062scales with

Sources & evidence

• Using prediction markets to estimate the reproducibility of scientific research 2015

  A 2015 publication evaluated markets forecasting replication of 44 psychology studies. It does not demonstrate an autonomous market that finances experiments.

• DARPA SCORE program

  Automated confidence in reproducibility and replicability was a research goal; the program does not validate autonomous science-financing markets.

• Recording provenance of workflow runs with RO-Crate 2024-09

  Workflow Run RO-Crate packages inputs, outputs, code and execution provenance; six workflow systems implement it. This establishes a research-software capability, not automatic truth checking.

Astra editorial synthesis; cited researchers retain credit for their results

Record reviewed 2026-10-18