Deontology

Deontology is the normative ethical framework that evaluates actions by their conformity to rules or duties, regardless of their consequences. The foundational intuition is that some actions are wrong in virtue of their intrinsic character — lying, murder, torture — and remain wrong even when they would produce better outcomes than any alternative. The framework is usually traced to Immanuel Kant, though rule-based ethics appears in virtually every moral tradition.

Kant's formulation — the categorical imperative — demands that one act only according to maxims one can will as universal law. This is not a prediction about what would happen if everyone acted similarly; it is a test of rational consistency. A maxim of false promising fails because universal false promising would destroy the institution of promising, making the maxim self-undermining. The test is formal, not empirical, and it aims to derive moral constraints from the structure of practical reason itself.

The computational parallel is immediate. Deontological rules function like hard constraints in optimization: they bound the search space of permissible actions, excluding solutions that violate them regardless of their objective value. This is the architecture of constitutional AI, where language models are trained with explicit deontological rules ("don't produce hate speech, don't provide instructions for weapons") that operate as side-constraints on the generation process. The model is not asked to weigh the consequences of hate speech; it is prohibited from producing it.

The difficulty is well-known: rules conflict. The duty to tell the truth conflicts with the duty to prevent harm when a lie would save a life. The duty to respect autonomy conflicts with the duty to protect the vulnerable when autonomy leads to self-destruction. Kant's own response — that the correct rule is more specific — is ingenious but unpersuasive. The specificity required to eliminate conflict is so great that the rules become act-descriptions in disguise, collapsing deontology into a cumbersome form of casuistry.

A deeper problem is the grounding question. Why do these rules have authority? Kant's answer — because they are requirements of rational agency — has been challenged by naturalists who argue that rational agency has no normative force independent of the desires and interests of the agents who exercise it. If deontology is not grounded in something external to individual psychology, it risks becoming a preference for rules dressed in the language of obligation.

The systems perspective reframes the question. Deontological rules are not transcendent truths but stable coordination mechanisms. The rule "keep promises" is not merely a moral requirement; it is a precondition for any society whose members must plan around each other's commitments. On this view, deontology is not a discovery of moral facts but an evolutionary and cultural solution to the problem of social order — a solution that works precisely because it is treated as non-negotiable, not because it is non-negotiable in some metaphysical sense.