Reconstruction from traces – MetaScience, Scientific Inquiry, and Agents

Epistemic purpose

Establish what happened in a past event or history, and when, from the physical traces it left, by finding the account that independent traces support and rival histories do not.

The event is unique and cannot be rerun: a past earthquake, an impact, an ice-sheet retreat. Conditions are found, not set. Reconstruction can still involve new sampling, coring, or experiments on the traces.

List rival histories before reading the traces as support for one; record each trace’s provenance and preservation; judge which traces are independent (different processes, different measurement chains); look for the trace that discriminates between rivals, Cleland’s “smoking gun”; date the events through estimation.
Inputs: the question, the traces and their provenance, the rival histories, dating constraints, the budget.
Outputs: a trace table with independence judged, the rival histories with what each predicts, the retained history with its status, and a claim record.

A history is retained as the best explanation among the stated rivals, supported by traces that are independent of each other; it is not confirmed by repetition. An event may be unique while its traces admit independent checks.
Distinguish a warranted decision at the time from a claim’s later assessed adequacy; append follow-up without rewriting the original record.

Stop when independent traces converge on one history and the remaining rivals are excluded or stated. Use estimation for dates and magnitudes; hypothesis assessment for rival histories; return to exploration when no rival accounts for the traces.
On graph v0.4.2: the trace-series composition, conditions found at F, O over independent traces in place of V, a best explanation at K. In the decision tree, question 6b when nature does not repeat the event.

Treat correlated traces as independent; replace primary records with retellings; read chronology from the order of articles; keep one history without naming its rivals.

Defined philosophically by Cleland’s account of historical science, the asymmetry of overdetermination and the search for a smoking gun (Cleland 2001, 2002), and by Turner’s analysis of how prehistory is made from traces (Turner 2007).
Geoscience cases: Cascadia’s great earthquakes from buried marsh soils and turbidites (Atwater 1987; Nelson et al. 1996); the end-Cretaceous impact from the iridium anomaly and the Chicxulub crater (Alvarez et al. 1980; Hildebrand et al. 1991; Schulte et al. 2010).
The actions above are design hypotheses derived from these sources, not established optimal policies. Missing process chronology remains unknown.

Given the traces available at a past date, retain the history the field later accepted, name its rivals, and identify which trace discriminated. Include correlated traces planted as independent and a plausible rival history.
Assessment definitions · Agent implementation