Adopting an Inference Rule: Accepting It or Learning to Use It?

If someone acknowledges that "all universal statements entail each of their instances," yet still cannot infer "this crow is black" from "all crows are black," should we say they have adopted universal instantiation, or merely accepted a sentence about it? This distinction lies at the cutting edge of the "adoption problem" in the philosophy of logic.

Viviane Fairbank and Ulf Hlobil break logical adoption into two steps: first, rationally determine that a certain logic is correct; second, change one's intellectual practice based solely on that determination and begin using the logic. "Adoption" here means not only treating the principle as true, but also putting it to work in actual reasoning. Their survey of the adoption problem frames it as a gap between logica docens and logica utens: the logic that is explicitly taught is not necessarily the logic that actually operates in reasoning. 1

The difficulty emerges at the second step. Imagine a person who cannot use universal instantiation, who first hears "universal statements entail each of their instances" and reasonably believes this principle is correct. Faced with "all crows are black" and a crow in a box, they might still fail to infer that the crow in the box is black. Because they still need to classify "this universal statement" as falling within the principle's scope of application, and then apply the principle to this concrete instance — and this principled application, in turn, looks suspiciously like universal instantiation itself. 2

This yields the classic dilemma: if the subject could already perform this kind of reasoning, adoption does no work; if the subject cannot, accepting the principle cannot enable them to start doing it. Replacing "accept" with "believe" or "know" does not automatically dissolve the dependency. Fairbank and Hlobil's contribution lies in reconstructing this set of pressures as four propositions that cannot all be true: rational adoption requires that the subject could not originally use the principle; accepting the principle should suffice to initiate use; some principles are basic, and adopting them requires reasoning at the same level as the principle itself; all logical principles can be rationally adopted. 2

William Nava's "Adopting an Inference Rule: A How-To Guide," published in Mind (2025, pp. 621–646), makes a specific core claim: adopting an unfamiliar inference rule is a diachronic process. During this process, the subject is guided by reasoning about the rule's statement but does not use the very rule being adopted; the ability to use it is a capacity acquired only at the end of the process. 3

This move changes the temporal structure of the problem. Classical arguments treat "accepting a principle" and "being able to reason by the principle" as adjacent states and demand that the former immediately produce the latter; Nava brings the stage between them into the theory. The rule statement functions here like an operating instruction: it participates in organizing the process without being treated as a complete reasoning ability the process already possesses. Adoption, then, is not "believe it and immediately apply it," but rather "while not yet capable of using the target rule, follow statements about the rule to acquire the ability to use it." From this perspective, the regress is not defeated by a more basic rule, but is dispersed across different temporal stages.

Nava also offers an important constraint: adoption requires six inference rules with special status — the author's public summary lists universal instantiation, modus ponens, and the naïve truth rule. But these rules need not be possessed in fully general form in advance; as long as "reasonably robust restricted versions" are available, the adoption process is not blocked, and there is no single minimal set of restrictions that must be in place for all adoptions. 4

The theoretical move here deserves careful differentiation. Nava is not declaring six rules as an uninterpretable logical bedrock. Privileged rules play a special functional role in adoption, but they are not for that reason inadoptable in a fully general sense. This weakens the regress argument's most powerful premise — "if a rule is basic, adopting it must already use its complete form." The adoption process still needs structured inferential resources, but it does not need a fixed minimal core that already packs all logical abilities inside.

This also clarifies how Nava's conclusion should be understood. It suffices to refute the universal claim that "no unfamiliar logical rule can be adopted," but it does not thereby prove that any subject under any psychological conditions can learn any arbitrary rule. The "reasonably robust" restricted versions mentioned in the summary remain a technical condition that needs unpacking: how they guide a subject who has not yet mastered the target rule, and what counts as a successful process, must be checked against the paper's full argument. A safer assessment is that Nava resolves the structural regress of adoption, rather than turning adoption into an unconditional psychological fact.

The normative pressure of the rule-following problem cannot be borne automatically by capacity formation. The SEP entry on rule-following and intentionality points out that rules, linguistic meaning, and mental content all involve correct and incorrect application; especially in the case of meaning, normative conditions require a theory to explain why a certain application is correct, and how this correctness constitutes an ought, a permission, or a reason for the agent — rather than merely describing what the agent will in fact do. 5

This leaves Nava's diachronic model with a precise question: when the rule statement "guides" the subject through the process, is this normative guidance or causal guidance in the training sense? A procedure can gradually induce stable responses in a person, but stable responses by themselves do not explain why a particular application is correct, nor why the subject has reason to take this rule as their own standard of reasoning. If "guidance" only means informational input and capacity shaping during the process, the model explains how the subject goes from being unable to being able; if "guidance" also means the subject has reason to reason this way, a normative link between the rule statement and correct application must be supplied.

This is not a simple rejection of Nava's proposal, but rather situates it at the right level of the adoption problem. Fairbank and Hlobil discuss whether one can move from logica docens to logica utens; Nava mainly rewrites the mechanism of that transition. The SEP's normativity asks a further question: when utens has already been formed, what makes it rule-following rather than merely a behavioral disposition accompanying some training outcome? The former question can have a process-theoretic answer; the latter still involves reasons, correctness, and rule-identity.

The adoption problem's most misleading feature is the compression of "knowing a rule," "being able to use a rule," and "having reason to use a rule" into the same thing. Fairbank and Hlobil's four-claim reconstruction exposes the conflicts this compression creates; Nava, through temporal stratification, shows that accepting a principle does not require immediately possessing the ability to use the target principle — the regress is therefore not a logical necessity.

But the boundary of this advance is equally clear: the diachronic process explains how ability emerges, but "ability is the result of a process" cannot on its own yield normativity. The real difficulty of rule-following is not just in making responses stable, but in explaining why a particular response is correct, why it is governed by the rule, and why the subject can treat the rule statement as a reason for reasoning. Only by separating these three layers can the adoption problem advance from "does accepting a rule equal being able to use it?" to a more testable question: when is a process merely learning, and when does it constitute rationally adopting a rule?