Relevance Logic: A Critique of Classical Entailment

Introduction

Welcome back. In our last few lessons, we've navigated several deep challenges to classical logic, including the Liar paradox and the Sorites paradox. We evaluated responses like epistemicism and supervaluationism, which attempt to resolve these issues by either positing unknowable truths or by modifying the semantics of vague predicates. These challenges, while significant, often focus on specific phenomena like self-reference or vagueness.

Today, we turn to a more fundamental and sweeping critique. Relevance logic does not merely target a specific paradox; it challenges the very definition of logical consequence, or entailment, that lies at the heart of classical logic. The core claim is that for an argument to be valid, its premises must be relevant to its conclusion—a condition that many classically valid arguments spectacularly fail to meet.

Your learning outcome for this lesson is to analyse the philosophical motivations for relevance logic and its critique of classical entailment. We will first diagnose the problem by examining the so-called "paradoxes of material implication." We will then explore the formal and philosophical proposals that relevance logicians offer as a remedy, moving from a simple formal criterion to deep, competing theories about the nature of proof, meaning, and inference.

1. The Critique: Paradoxes of the Material Conditional

As you know from your studies in mathematical logic, classical propositional logic defines the conditional A → B truth-functionally as being equivalent to ¬A ∨ B. This is known as the material conditional. While this definition is powerful and simple, it leads to several "paradoxes" when we try to align it with our intuitive understanding of "if...then..." or logical entailment.

These are not paradoxes in the sense of formal contradictions, but rather conclusions that are classically valid yet seem intuitively fallacious due to a lack of relevance between the antecedent and consequent.

Flowchart of Logical Systems and Their Principles — This diagram shows how different non-classical logics diverge from classical logic by rejecting certain principles. Relevance logic is distinguished by its rejection of the material conditional in favour of a 'Relevant implication'.

To see exactly what relevance logicians object to, please watch the following video segment. It provides a clear overview of the material conditional and the counter-intuitive consequences that arise from its truth-functional definition.

Conditionals: the material conditional

The YouTube video 'Conditionals: the material conditional' by Kane B provides an excellent diagnosis of the problem that motivates relevance logic. This section focuses on the specific 'paradoxes' that arise from the classical definition.

Please watch from the beginning until 06:49. The first part (until 01:45) is a quick review of the material conditional. The key part for us starts at 05:00, where the speaker details the 'paradoxes of material implication'. Focus on why these classically true statements feel wrong.

As the video explains, the material conditional gives us the following valid schemas, which relevance logicians find objectionable:

A false antecedent implies any consequent: ¬A → (A → B)
A true consequent is implied by any antecedent: B → (A → B)

From these, we can derive other classic "fallacies of relevance":

(A ∧ ¬A) → B (Principle of Explosion: a contradiction implies anything)
A → (B ∨ ¬B) (Anything implies a tautology)

The common thread is a failure of connection. Intuitively, the truth of "the moon is made of cheese" has nothing to do with the truth of "2 + 2 = 5". Yet, the material conditional If the moon is made of cheese, then 2 + 2 = 5 is true, simply because the antecedent is false. Relevance logicians argue that a proper account of implication or entailment must capture this missing element of relevance.

2. A Formal Diagnosis: The Variable-Sharing Principle

How can we formally capture this intuition about relevance? The first and most famous step is a simple syntactic constraint.

Relevance Logic - Stanford Encyclopedia of Philosophy

The Stanford Encyclopedia of Philosophy entry on 'Relevance Logic' introduces the formal criterion that relevance logicians use to enforce a connection between premises and conclusions.

Please read the introduction and the short section just below it on the 'variable sharing principle'. This will explain the basic formal move made by relevance logicians.

As you just read, the variable-sharing principle is a necessary (though not sufficient) condition for a logic to be considered a relevance logic. It states:

No formula of the form A → B can be a theorem of a relevance logic if A and B do not share at least one propositional variable.

This simple principle effectively rules out the primary paradoxes. For example:

p → (q → q) is invalid because the antecedent p shares no variable with the consequent (q → q).
p → (q → p) is invalid in many systems because after discharging the assumption q, the resulting implication from p to p would be fine, but the implication from q to p lacks shared variables. The derivation itself is blocked in a proof system, which we'll see next.

While simple, the variable-sharing principle is just a symptom of a deeper philosophical diagnosis. The core of relevance logic lies in the philosophical motivations for why such a principle should hold.

3. Philosophical Motivations for Relevance

Why should we demand relevance? What is implication, if not material implication? Philosophers have proposed several deep, competing answers. The following paper provides an excellent survey of these "routes to relevance."

[PDF] Routes to relevance: Philosophies of relevant logics

Shawn Standefer's paper 'Routes to relevance' systematically surveys the main philosophical motivations that have been offered for adopting a relevant logic. We will focus on the two most historically prominent and philosophically distinct motivations.

Please read the 'BACKGROUND' (Section 1), 'USE CRITERION' (Section 2), and 'SUFFICIENCY' (Section 3). In the Background, note the distinction between contractive and non-contractive logics. For the Use Criterion, focus on how the subscripting system in the natural deduction proof formally captures the idea that a premise must be used. For Sufficiency, grasp the model-theoretic idea that an antecedent must be sufficient for the consequent even in 'impossible' worlds or situations.

Let's unpack the two central motivations you just read about. Your background in both proof theory and model theory will be valuable here, as these two motivations map neatly onto that distinction.

A. The Use Criterion: A Proof-Theoretic Motivation

Pioneered by Alan Anderson and Nuel Belnap, this is arguably the most famous motivation. The idea is simple and powerful: for A to genuinely imply B, a proof of B from the assumption A must actually use A.

As Standefer explains, this is formalized in a natural deduction system using subscripts to track which hypotheses are used to derive each line of the proof.

An implication A → B can only be introduced if the assumption A was actually used in the sub-proof that derived B.
The paradoxical derivation of p → (q → q) is blocked precisely because the initial assumption p is never used to prove q → q. The subscript tracking makes this formally explicit.

This approach defines relevance internally to the proof system. It's an argument about what constitutes a legitimate deduction. An inference is valid only if the premises are fulfilling their role in the deductive machinery.

B. Sufficiency: A Model-Theoretic Motivation

This approach, championed by Richard Sylvan (formerly Routley) and Val Plumwood, offers a semantic or model-theoretic motivation. The core idea is that genuine implication, A → B, means that A is totally sufficient for B.

What does "totally sufficient" mean? It means A guarantees B not just in the actual world, or in all classically 'possible' worlds, but across a maximal range of situations, including 'impossible' or non-normal worlds where the laws of logic themselves might differ.

In this framework, p → (q ∨ ¬q) fails to be a logical truth because one can imagine a situation (a world in the model) where p is true but the Law of Excluded Middle fails and (q ∨ ¬q) is not true. In such a situation, p is not sufficient for (q ∨ ¬q).
This approach naturally leads to the ternary relational semantics you saw mentioned in the SEP article. The three-place relation Rabc can be interpreted as "the information in a and b combines to yield the information in c". An implication A → B holds at a world a if for any world b where A holds, the combination of information guarantees B.

Where the Use Criterion focuses on the mechanics of proof, the Sufficiency view focuses on the range of scenarios over which an implication must hold. Relevance is a by-product: if A and B have nothing to do with each other, it will always be possible to construct a situation where A holds but B does not, demonstrating that A is not sufficient for B.

4. Is Logic Revisable? A Meta-Philosophical Perspective

The clash between classical and relevance logicians is not just a technical dispute. It raises a fundamental philosophical question: is logic an immutable, a priori discipline, or is it a theory like any other, subject to rational revision in the face of counter-intuitive data? Graham Priest offers a powerful framework for thinking about this.

Revising Logic - Graham Priest

In this classic talk, Graham Priest argues that logic is indeed revisable. His argument provides a powerful meta-philosophical lens through which to view the debate between classical and relevance logicians.

Please watch the segment from 03:02 to 20:40. Focus on his distinction between different senses of 'logic' and his core argument that 'logica docens' (theories of logic) are chosen based on standard criteria for theory choice, including adequacy to the data. Note his specific mention of the conditional as an area where classical logic might be weak.

Priest's argument gives us a way to frame the entire project of relevance logic. He distinguishes:

Logica docens: The logic we teach; the explicit theory of validity.
Logica utens: The logic we use; our implicit inferential practice.
Logica ens: The logic that is; the objective facts about validity (if any).

The debate over relevance is a debate about logica docens. Relevance logicians are proposing a new theory of validity. They argue that classical logic, while a powerful theory, is a poor fit for certain "data"—namely, our powerful intuitions about relevance in entailment and conditionals. They offer their systems as a better theory, one that is more adequate to this data.

On this view, choosing a logic is like choosing a scientific theory. We weigh competing theories (classical, intuitionistic, relevant) against criteria like simplicity, explanatory power, and, crucially, adequacy to the data. The paradoxes of material implication are the problematic data points that relevance logic claims to explain better than classical logic does.

Conclusion

Today we have explored the deep philosophical challenge that relevance logic poses to classical logic. We've seen that this is not a minor quibble but a fundamental disagreement about the nature of logical consequence.

Key Takeaways:

Relevance logic is motivated by the "paradoxes of material implication," where classically valid inferences violate the intuition that premises must be relevant to the conclusion.
The initial formal proposal is the variable-sharing principle, which provides a necessary but not sufficient condition for relevance.
The philosophical motivations are deeper and more diverse. We examined two major "routes to relevance":
- The Use Criterion (proof-theoretic): A premise must be formally used in a deduction of the conclusion.
- Sufficiency (model-theoretic): The antecedent must be sufficient for the consequent across a maximal range of situations, including non-classical ones.
The debate can be framed as a rational process of theory choice for logica docens (our theory of logic), where relevance logic is proposed as a theory that better accounts for the "data" of our inferential practices and intuitions than classical logic does.

This radical revision of entailment opens the door to even more profound questions. If logic is revisable, how far can that revision go?

Preview of the Next Lesson:

Our next lesson takes this question to its limit. We will analyze the Liar paradox again, but this time considering a response that is even more radical than relevance logic: dialetheism. Advanced by Graham Priest, this is the view that some contradictions are, in fact, true. We will evaluate the philosophical case for accepting true contradictions and explore the paraconsistent logics designed to contain them without leading to explosion.

Recommended resources