LodeCraft insight · Decision quality
How geologists can keep uncertainty visible in exploration decisions.
Uncertainty becomes usable when evidence, interpretation, contradiction, missing support and the next investigation are recorded separately rather than compressed into one confident map.
Publication details & scope
Direct answer
Do not treat uncertainty as one number to hide at the end. Keep its sources attached to the interpretation and to the decision that follows.#
A geological team rarely needs a declaration that “the project is uncertain.” It needs to know which observations are reliable, which relationships are inferred, where evidence conflicts, what is missing and which next investigation could change the decision.
Separate observation from interpretation.#
Recorded lithology, assay values, survey response and mapped structures are not the same thing as a deposit model, target rationale or model-generated rank. Keeping those layers distinct allows a reviewer to challenge the interpretation without losing the observations.
Derived products should retain lineage back to the source and the transformations that created them.
Record support, contradiction and absence separately.#
- Support
- What points towards the hypothesis?
- Contradiction
- What points towards another explanation?
- Missing evidence
- What has not been tested?
For each target hypothesis, write down what supports it, what contradicts it and what has not been tested. Absence of a dataset is not negative evidence. Likewise, a broad regional dataset may support a regional idea while remaining incapable of resolving a prospect-scale feature.
Show model disagreement before averaging it away.#
When structural, geochemical, geophysical, ML and DL views differ, the disagreement can identify an assumption worth investigating. A single composite is useful for navigation, but it should not erase the component views or imply that every branch provides independent evidence.
Give abstention a legitimate place in the workflow.#
A system should be able to say that support is inadequate, coverage is inconsistent or the result depends too heavily on one uncertain input. Abstention prevents the need to rank every location merely because the computation can produce a value.
Connect uncertainty to a next action.#
Uncertainty is most useful when it changes the investigation plan. A team may choose higher-resolution mapping, infill sampling, reprocessing, field checking, a different geophysical survey or no further work. The recommendation should state which unresolved issue that action is intended to test.
A practical uncertainty ledger#
Use this as your own preparation record. Selections last while this page is open; they do not validate project readiness.
Uncertainty is not indecision.#
A decision can still be made with incomplete evidence. The objective is to make the basis and limitations visible to the accountable person. Public-reporting frameworks such as JORC emphasise transparency, materiality and competence; project decision-support work should retain the same discipline without implying that every internal analysis is a JORC report.
See the LodeCraft case studies for public examples where disagreement, warnings, limitations and the next human review remain attached to the result.
For your own project, the LodeCraft exploration analysis engagement explains how the evidence, uncertainty and next review priorities form part of an agreed handover.