Short answer. The common methodological thread across Riff Systems is to define the claim and system boundary before decisive interpretation, use controls capable of defeating attractive explanations, distinguish structural evidence from causal evidence, and preserve failed or incomplete outcomes rather than converting them into retrospective successes.
Recurring methodological principles
- Declare the boundary. State what system or representation is being tested and what remains outside it.
- Freeze decisive criteria where possible. Later exploratory analysis can explain a result but should not silently change a confirmatory endpoint.
- Use strong conventional nulls. A proposed new mechanism should survive ordinary explanations such as confounding, persistence, filtering, seasonal structure or continuing forcing.
- Separate evidence levels. Association, structure, causal efficacy and human interpretation are not interchangeable.
- Preserve negative results. A method is useful partly when it can return an unwanted answer and leave that answer in the record.
How this appears across domains
In Titan T1, an intermediate prior-state signal was rejected after stronger seasonal and circulation controls. In Forest F1, superficially adaptive signatures stop at different causal gates. In NBS-001 / NBS-002, a confounded structural result is rejected, a corrected structural and causal sequence passes, and the prospective interpretation still fails.
Questions this work addresses
- How can cross-domain comparison be rigorous without assuming that different systems share a mechanism?
- What makes a null hypothesis strong enough to challenge a preferred explanation?
- How should confirmatory and exploratory analyses be separated?
- What does causal evidence add beyond prediction or decodability?
- How can negative and incomplete results improve a research programme?