Evidence and research standards

CosmosIntelligence is designed to explore ambitious questions without lowering the standard of proof.

The project succeeds when weak patterns are rejected quickly and promising candidates become easier to investigate.

Models should first learn ordinary stars, galaxies, quasars, nebulae, pulsars, transits, flares, supernovae, dust, lensing, radio interference, calibration effects and other known phenomena.

An anomaly score is not a discovery. It is a reason to review an observation more carefully.

A strong candidate should move through progressively harder checks:

1. Visible in the original or appropriately calibrated source data.
2. Provenance is known and reproducible.
3. Instrument artifacts and data processing errors are considered.
4. Known astrophysical explanations are tested.
5. Statistical controls show the pattern is meaningfully unusual.
6. Independent instruments, surveys or observations replicate the important feature where possible.
7. The hypothesis makes a testable prediction or produces a useful prospective check where possible.
8. Reviewers can inspect the evidence and disagree with the interpretation without losing access to the underlying data.

Every public research page should make it clear whether a statement is:

• an established measurement or observation
• a model dependent scientific interpretation
• an active research question
• a hypothesis being tested
• a speculative idea without strong evidence

This distinction is especially important for consciousness, technosignatures and other subjects that can be distorted by sensational reporting.

A failed hypothesis, unsuccessful connector test or anomaly that turns out to be ordinary physics can still be useful if the method and result are documented.

Publishing negative results helps train better systems and protects the project from repeatedly chasing the same false positives.

Prefer primary sources: NASA, JPL, ESA, observatories, official archives and peer reviewed papers. Secondary reporting can help with context, but technical claims should link to the underlying evidence whenever possible.

Sources and Evidence
Research Notes