Bounded Causal Cognition and Host-Relative Ethics: A Functional Framework

## Abstract This paper presents a functional, non-essentialist framework that separates the *generative explanation* of behavior from its *consequence evaluation*. On the explanatory side, agents are modeled as resource-bounded causal reasoners who act on truncated, subjective causal graphs (\(C_t\)) rather than objective reality. On the evaluative side, moral valence is defined not as an intrinsic property of actions, but as a relation between an act’s consequences and an analyst-specified host system (\(H\)) over a defined horizon (\(T\)) against a counterfactual alternative (\(A_0\)). The framework classifies moral failures into three distinct, non-exclusive error dimensions—causal error, scope error, and weighting error—each mapping to distinct systemic interventions. We present this as a descriptive, falsifiable framework designed to support quantitative modeling and experimental testing; it provides the *format* of evaluation, leaving host boundaries and normative weights as explicit inputs.