Clinical decision support increasingly places clinicians in a distinctive moral position: they remain answerable for decisions while acting in an informational environment partly structured by algorithmic recommendations. The ethical problem therefore begins not only when an algorithm makes a decision, but after advice has been received and a clinician must follow, modify, delay, seek review of, override, or ignore it. This normative-responsibility analysis examines how those responses alter the grounds on which responsibility should be attributed and how unresolved responsibility may leave moral residue even without proven wrongdoing. It argues that formal decision authority is insufficient to determine moral responsibility because responsibility also depends on epistemic access, meaningful control, causal contribution, professional role, available alternatives, institutional conditions, and the capacity to question algorithmic advice. The article proposes a relational account in which responsibility is distributed across actors according to their actual relationship to the decision rather than assigned automatically to the final human decision-maker. Moral residue is subsequently conceptualized as a morally salient remainder that may persist when value conflict, uncertainty, constrained choice, or unresolved accountability survives an otherwise defensible decision. The proposed framework distinguishes prospective accountability from retrospective blame and treats documentation, explanation, review, acknowledgement, and learning as potential resources for moral repair rather than demonstrated remedies. Its claims are bounded by variation among clinical systems, professions, institutions, patient populations, and jurisdictions. The framework is normative and theory-generating; it does not establish an AI-specific syndrome, causal pathway, liability rule, validated intervention, or universal allocation of responsibility.
