Compartmental Decision Pathology (CDP): A General Framework for Context-Dependent Disease Architecture

Central thesis Neither disease labels, named exposures nor final phenotypes are sufficient to reconstruct mechanism. CDP asks which physiological mapping problem a specified compartment must solve, which causal roles are actually demonstrated in the analysed episode, and where the evidence stops. Abstract Compartmental Decision Pathology (CDP) proposes a restricted causal class in which a physiological architecture localized within a biological compartment must resolve an unavoidable trade-off by mapping input, context and prior state onto qualitatively distinct functional programmes. Pathology can emerge when that architecture processes atypical input or context, inherits an altered prior state, or is modified by an identifiable mode of interference; intrinsic defect of the architecture is not required. The programme-level grammar is Perturbation → [Access when required] → branch-specific core operation → Functional configuration → Amplification → Phenotype, with modes of interference acting orthogonally and with maintenance, effector execution and one-way propagation adjudicated separately. The four current branches occupy distinct causal domains: Recognition (ARA) for already-visible normal self, Discrimination (ADA) for altered self, Calibration (ACA) for developmental functional organization, and Fate (AFA) for genuine non-self after antigen-specific engagement. All four have positive branch-specific operation-level support, but closure is asymmetric. ARA and ADA retain qualified fixed-specificity core-operation anchors; ACA has qualified Calibration, operation-derived Functional-configuration and recurrent-Calibration anchors; AFA has strong Fate support and a qualified four-test recall-state anchor whose independently isolated Fate provenance remains unresolved. No branch currently closes the full causal chain. The framework includes a multiple-sclerosis stress-test showing how distinct candidate trajectories can converge on one phenotype without implying a shared branch or causal chain; convergence depth is treated as a comparative property, not a new causal layer. Neither a named input nor a shared phenotype identifies the intervening architecture; the named exposure may not be the causal object; mechanisms are classified by their demonstrated role in a specified episode rather than by molecular identity. Access remains optional and is relational to object, physical/formulation state, route or interface, and target compartment. Binding is not automatically a core operation, persistence is not automatically Functional configuration, effector execution is not Amplification, and seriality is not recurrence. Diseases may contain nested branch-specific episodes, and interventions themselves belong inside the causal map: treatment can suppress, deplete, restore, reopen Access or alter later prior state, while withdrawal can become a new Perturbation when treatment has changed the system. Therapeutic response therefore identifies a manipulable dependency at a specified phase, not causal origin by itself.