Epistemic Uncertainty

Epistemic uncertainty measures how confidently an artifact reflects the truth it claims to represent at the moment of consumption. It is independent of committed/uncommitted status and of atomicity level (L0–L3). An artifact is fully characterised only when both dimensions are known — for example, L1-U-med: a committed L1 invariant aggregate with medium uncertainty. See L–U shorthand for the combined notation.

Three components must be characterised before appropriate use: temporal uncertainty (staleness since ground truth was established), probabilistic computation loss (information efficiency ? — information lost by the processing that produced the artifact), and cross-stream consistency (whether a combined view corresponds to a state that could have existed together — via as-of alignment, bounded clock skew, causal order, and where applicable serializability or linearizability guarantees from distributed systems theory). Shorthand levels U-low, U-med, and U-high summarise the composite for dependency decisions; residual cross-stream error is often negligible inside a trusted clock domain with a declared snapshot or linearizable read, but never assumed away without disclosure.

Impossible simultaneity is the failure mode when cross-stream inconsistency is high and concealed — a dashboard mixing streams of different vintages or alignment policies without disclosure, presented as one moment. High-U L3 renderings are prohibited for Class A consumers regardless of committed status, because the output looks authoritative while inconsistency is invisible to the reader.