Inferential estimation

Drafted human-editable specification · v0.4.0 · no agent-performance claim

Epistemic purpose

• Recover an unobserved quantity with uncertainty and identifiable limits.

When this mode helps

• A structural, temporal, or physical quantity is inferred from observations through a measurement/forward model. A standard inversion with a new target is still estimation.

Agent actions and representations

• Define target and units; specify forward relation; inspect identifiability and noise; fit or invert; assess resolution, regularization, uncertainty, and alternative assumptions.
• Inputs: the question, available evidence and provenance, constraints, prior claims, and remaining budget.
• Outputs: inspectable artifacts, updated question/evidence state, and a claim record when a substantive claim is made.

Evidence and claim scope

• Report what data constrain, what assumptions supply, and which quantities remain unresolved. Assess recovery on appropriate known cases and independent observations where available.
• Distinguish a warranted decision at the time from a claim’s later assessed adequacy; append follow-up without rewriting the original record.

Transitions and stopping

• Revisit instrumentation for response errors; enter methods for a new inversion procedure; use hypothesis assessment only when the estimates can distinguish explanations.
• Neighbouring profiles: instruments, methods, hypothesis, robustness.

Characteristic failure

• Treat a regularized image as a direct photograph; report narrow uncertainty while ignoring model error; infer a causal mechanism from a resolved parameter alone.

Human evidence and borrowing

• Tarantola (2006) motivates inverse-problem limits; Bogen and Woodward (1988) distinguishes data and inferred phenomena. Mai et al. (2016) provides earthquake-source inversion benchmarks, not a unique reference truth for every real event.
• Cross-field comparison: Measurement philosophy and statistical inference travel across domains; the forward model, identifiability conditions, units, and scales must be re-established in the target field.
• Science of process/impact: Woollam et al. (2022) supplies reusable seismological benchmark infrastructure. Population distributions of estimation advances have not been measured in this project.
• The actions and transitions above are design hypotheses derived from these sources, not established optimal policies. Missing process chronology remains unknown.

Evaluation against past work

• Improve a meaningful estimate at stated resolution and uncertainty, or establish a consequential non-identifiability. Test plausible model error and a known reference case.
• Use the historical evaluation to choose a source record, preserve evidence boundaries, and assess a meaningful gain. A synthetic control tests a constructed case, not a historical discovery.
• Assessment definitions · Agent implementation

Bogen, James, and James Woodward. 1988. “Saving the Phenomena.” The Philosophical Review 97 (3): 303. https://doi.org/10.2307/2185445.

Mai, P. Martin, Danijel Schorlemmer, Morgan Page, et al. 2016. “The Earthquake‐source Inversion Validation (SIV) Project.” Seismological Research Letters 87 (3): 690–708. https://doi.org/10.1785/0220150231.

Tarantola, Albert. 2006. “Popper, Bayes and the inverse problem.” Nature Physics 2 (8): 492–94. https://doi.org/10.1038/nphys375.

Woollam, Jack, Jannes Münchmeyer, Frederik Tilmann, et al. 2022. “SeisBench—a Toolbox for Machine Learning in Seismology.” Seismological Research Letters 93 (3): 1695–709. https://doi.org/10.1785/0220210324.