Is formal logic the same as common sense?

No. It provides precise reasoning frameworks, while ordinary common-sense reasoning often relies on unstated context, expectations, and exceptions.

Can a logically correct conclusion still be wrong in practice?

Yes. The derivation may follow correctly from premises that are inaccurate or from a model that omits an important condition.

Does formal logic replace machine learning?

No. They address different needs and can be combined, for example when learned observations feed an explicitly specified reasoning process.

AI Glossary / Graphs and symbolic AI / Beginner

Formal logic

Definition

Formal logic is the study and use of precisely defined languages and rules for reasoning. It distinguishes how statements are written, what they mean, and which conclusions follow from given premises. In AI, formal logic supports explicit knowledge representation, inference, and checking whether a conclusion is justified within a stated model.

What it means

Everyday language leaves many details implicit. A sentence such as “approved requests can proceed” might describe permission, a normal practice, or a rule with exceptions. Formalization requires choosing an exact interpretation and representing it explicitly.

The resulting system can test relationships among statements without relying on the persuasive wording of an argument. This does not make the premises factually correct. Formal reasoning establishes what follows from assumptions; checking whether those assumptions accurately describe the intended situation is a separate task. Clear symbols cannot rescue an incorrect policy definition or unreliable input record.

How it works

A formal language has syntax specifying well-formed expressions. Its semantics specify how those expressions are interpreted. For example, propositional logic combines whole statements with operators such as “and,” “not,” and “if ... then.” First-order logic additionally represents objects, relations, variables, and quantifiers.

Semantic entailment means that a conclusion is true in every interpretation where the premises are true. A proof system provides permitted steps for deriving conclusions. A sound inference system derives only entailed conclusions; completeness means it can derive all entailed conclusions within its defined scope.

Different logics offer different expressive resources and computational properties. Formal logic is therefore a family of approaches, not a single program that automatically resolves every question expressed in words.

Three practical examples

1. A configuration tool represents rules such as “choosing module A requires module B.” Logical checks can identify incompatible combinations before a configuration is accepted.

2. A knowledge-based assistant applies explicit classification rules to recorded facts. Its explanation can identify which premises and inference steps support a particular classification.

3. A software-verification team writes properties that an implementation should satisfy. Formal reasoning can help establish those properties under the modeled assumptions and the verification method's scope.

Think of it this way

Think of formal logic as a carefully specified board game for arguments. The symbols are the pieces, the starting premises are the initial position, and inference rules define legal moves.

The analogy highlights disciplined transformations, but logic also has semantics: the pieces stand for statements interpreted in possible situations. A legal-looking sequence matters because the rules preserve the intended relationship between premises and conclusions. The game cannot independently certify that its starting description matches the real world.

Worked example

Let p mean “Request A has completed review” and q mean “Request A is eligible for release.” Suppose the premises are p and p implies q. By modus ponens, q follows.

To check semantically, consider truth assignments to p and q. The assignment p true and q false violates p implies q, so it cannot satisfy both premises. Assignments where p is false violate the premise p. The only assignment satisfying both premises has p true and q true. Therefore q is entailed.

Now change the premises to q and p implies q. Does p follow? No. The assignment p false and q true satisfies both new premises but makes p false. That counterexample disproves the proposed entailment. Eligibility might arise through another route unless the rule explicitly excludes it. The example shows why reversing an implication requires additional justification, not merely familiar wording.

Illustrative case study

Fictional case: Juniper Services turns a small approval checklist into explicit rules for an internal prototype. During review, staff notice that “reviewed implies eligible” does not capture a separate requirement concerning missing attachments.

They revise the rule and test cases with completed reviews, incomplete records, and conflicting inputs. The project distinguishes errors in the rule specification from errors in the inference procedure. A derivation can be correct relative to the old rule while still implementing the wrong business requirement. This fictional scenario illustrates model review; it reports no real compliance outcome or automation saving.

Common misconceptions and mistakes

Do not equate logical validity with factual truth of every premise. A valid argument preserves truth from premises to conclusion; premises can still be false.

Do not treat “not proved” as automatically equivalent to “false.” That depends on the reasoning framework and assumptions about missing information. Also avoid translating “only if,” “if,” and “if and only if” interchangeably.

Contradictory premises require care. In classical logic, an inconsistent knowledge base entails every statement. A production system therefore needs a deliberate approach to inconsistency rather than assuming more recorded assertions always improve reasoning.

Business use

Formal logic is useful when requirements can be made explicit and traceable reasoning matters. Start with a limited vocabulary, define each predicate or proposition, and test representative cases with domain specialists.

Keep rule versions, assumptions, and evidence provenance available alongside derived conclusions. Ask whether a problem also requires probabilities, defaults, time, or obligations that a basic propositional model does not express adequately. The practical value comes from matching the formalism to the decision and validating the model, rather than from adding symbols to ambiguous requirements.

Check your understanding

Why does knowing q and p implies q fail to establish p?

Answer: A situation with p false and q true satisfies both premises. Since at least one satisfying interpretation makes the proposed conclusion false, that conclusion is not entailed.

Related terms

Learn next

Practice building truth tables for implication and finding counterexamples to invalid arguments. Then study first-order logic to see how objects and quantifiers extend the language, while continuing to separate a proof's correctness from the accuracy of its premises.

First-order logic

Sources

Russell and Norvig, AIMA logical-agents slides, syntax, semantics, entailment, and inference: https://aima.cs.berkeley.edu/slides-pdf/chapter07.pdf

Russell and Norvig, first-order logic slides: https://aima.cs.berkeley.edu/slides-pdf/chapter08.pdf

The approval example, countermodel, and fictional case are original teaching illustrations.

Previous: Formal language Next: Formant