E.18:6.1 - Bias-Annotation (per E.8 SG-bias slot)

• Acyclic-bias risk. Tooling accustomed to DAGs may discourage admissible feedback loops; E.18 explicitly permits loops with budget and sentinel controls (CC-E18-13, -18).
• Scalarization-bias risk. Cultural defaults to single-score rankings can suppress Pareto fronts and QD archives; E.18 keeps declared order relations and return sets visible (CC-E18-10, CC-E18-12).
• Interop-dominance risk. File and format ecosystems (CWL, RO-Crate, and lineage) can be mistaken for semantic sources; E.18 places them in InteropCard and keeps the applicable semantic definitions and tests explicit at the loci and gates.
• Over-formalization risk. Category-theoretic formalisms can obscure operational guard-rails; ordinary E.18 states each crossing in readable prose, while replay-facing use keeps exact positions, per-binding accounts, one separate A.21 gate decision, and CrossingRef (CC-E18-11, -23).
• Retrospective rewrite risk. Global rewrites break replay; E.18 confines them to edition bumps and slice-local refresh (CC-E18-16).

Mitigations. Select checks from the current use before applying a profile. Require edition pins when a current comparison or publication depends on them, inspect the GateDecisionResult for a current decision and an optional DecisionLog only when audit, history, replay, or reuse depends on it, and keep refresh tests within the affected PathSlice.