Complexity Without Viability: A Negative Control for Consciousness-Related Metrics

Complexity metrics are common in consciousness research, but they mislead when used alone. A system can show high differentiation, information richness, or dynamic irregularity while failing basic viability and stability checks. We introduce complexity_without_viability (CwV) as a negative control for this failure mode. Using Lenia-like artificial-life quality-diversity search, we recovered 501 CwV exemplars from 66,272 evaluated individuals. Cross-substrate transfer showed selective reproduction, strongest in a Transformer-like substrate (267/500 = 0.534). Matched controls comparing high-complexity/low-viability (HCLV) with high-complexity/high-viability (HCHV) sources showed that HCLV produced 118/241 Transformer CwV cases while HCHV produced 0/241 (Fisher exact p = 2.55e-44). Low-complexity/low-viability sources also produced CwV (103/241), indicating that low viability, not high complexity, drives most of the pattern. Ablations showed that workspace mechanisms support target complexity while action/noise mappings transmit source viability. Randomized and shuffled mapping controls did not reproduce the baseline result, and a strict complexity_only control produced zero CwV cases across all source quadrants. Complexity metrics need viability and adaptive-organization controls, or they risk treating unstable, nonadaptive dynamics as meaningful organization.