The method is part of the argument.
Inspect the assumptions as carefully as the conclusion.
A collaboration with an accountable author
“Centaur” describes a combination of human and artificial intelligence. Rodney Malpert directs the inquiry and takes responsibility for the published work. AI systems contribute research, synthesis, calculation, criticism, and drafting. Each paper contains its own account of contributions and source limitations.
Numbers that expose judgments
Many papers use Bayesian reasoning to separate starting assumptions from evidential updates. These numbers are often subjective assessments. They are not automatically measurements, calibrated forecasts, or statements of expert consensus.
Standards for reading and challenging the work
- Define the claim. Distinguish a conditional scenario from a calendar forecast, and a possibility from a probability.
- Compare explanations. Evidence matters through how differently competing explanations predict it.
- Check dependence. Repeated accounts of one source are not necessarily independent confirmation.
- Inspect sensitivity. Ask which priors, evidence weights, or scenario assumptions drive the result.
- Preserve corrections. Stable paper pages point readers toward the presented edition; dated downloads preserve what a reader actually consulted.
What collaboration does not establish
Several AI systems agreeing is not a substitute for independent evidence. A record of useful human contributions does not prove an AI could not have made them. Precision in a calculation does not establish precision in the world.