Systems & evidence
Hypothesis and evidence monitoring with SeamWatch
SeamWatch starts with a question: what observation would actually change a judgment? It brings the explanation, its evidence and the expected signals into one explicit watch, so that a stream of new information can be assessed against a stated model.
The watch is the working object
A watch keeps a question, competing hypotheses and their supporting material together. It also records what each hypothesis would lead us to expect. An alert is more useful when its significance can be understood in that context.
The approach is relevant when an explanation depends on events that continue to unfold. Instead of treating each new report as a fresh conclusion, the record shows which part of the existing explanation the report strengthens, weakens or leaves unchanged.
Look for a discriminating observation
Two explanations can fit the same set of observations. Collecting more examples of that shared pattern may add volume without separating them. A discriminating test identifies an observation that the explanations predict differently.
For example, if two accounts predict different publication dates for a document, the document’s dated appearance may be more useful than another commentary about the dispute. This is an illustration of the method, not a claim about a particular investigation.
Absence and collection failure are different
A missing item may be confirmed absent, or it may simply have escaped collection. SeamWatch’s design preserves that distinction alongside provenance and audit history. A failed source check must not silently become evidence that an event did not happen.
SeamWatch is presented here as a project and a method for governed early warning. Human judgment remains responsible for interpreting evidence and deciding what to publish or do.