Are these maxims the same as politeness rules?

No. Politeness can interact with conversation, but Grice's core categories address a particular account of cooperative interpretation.

Can a speaker deliberately flout a maxim?

Yes. A recognizable departure can invite an implicature, although the listener may not recover it.

Does relevance have one fixed definition?

No. What counts as relevant depends on the conversational purpose and context.

AI Glossary / Natural language and speech / Intermediate

Maxim, Gricean

Definition

A Gricean maxim is a conversational expectation used in H. P. Grice's account of how people interpret what others communicate beyond literal wording. The four familiar categories are quantity, quality, relation and manner. They concern appropriate information, evidential truthfulness, relevance and clear presentation in a conversation's context.

What it means

People often understand more than a sentence explicitly states. If someone asks whether a room is available and receives “There is a workshop in there until noon,” they may infer that the room is unavailable before noon. The answer does not use the word unavailable, but its relevance to the question supports that interpretation.

Grice's account explains such reasoning through an expectation that contributions fit the shared conversational purpose. The maxims are not a checklist proving that every real conversation is cooperative, nor a universal etiquette code for judging speakers.

For AI dialogue, they provide a useful way to think about relevance, sufficiency and indirect meaning. They also expose a danger: a plausible inference about intention can exceed what the speaker actually established.

How it works

Quantity concerns supplying the information needed for the exchange without unnecessary excess. Quality concerns avoiding statements believed false or unsupported. Relation concerns relevance to the conversational purpose. Manner concerns intelligible, orderly presentation rather than the subject matter itself.

Listeners combine these expectations with context and shared knowledge to infer conversational implicatures. Such inferences are typically defeasible: later clarification can withdraw or change them without necessarily contradicting the literal sentence.

A conspicuous departure can itself communicate something. In a suitable shared context, an obviously exaggerated remark may invite a nonliteral interpretation. This deliberate, recognizable departure is often called flouting a maxim, distinguished from quietly misleading someone or simply failing to communicate clearly.

Interpretation is not automatic. Participants may differ in shared knowledge, discourse habits or expectations about indirectness. A system should preserve uncertainty when several explanations fit the same wording instead of assuming its preferred inference is an explicit instruction.

Three practical examples

A support assistant answers the actual delivery question before adding background that would distract from the user's immediate need.
A scheduling assistant treats a mentioned conflict as relevant evidence but clarifies any missing time or requested action.
A dialogue reviewer distinguishes a supported answer from a fluent guess that merely sounds responsive.

Think of it this way

Think of a team trying to complete a shared puzzle. A contribution is easier to interpret when everyone expects it to help with the current part of the task and to be offered honestly.

The analogy does not imply that all speakers share the same goal or communication style. Real exchanges may involve humor, uncertainty, disagreement or different assumptions, and cooperation in interpretation does not mean agreement on every issue.

Worked example

Tracing an inference without overcommitting

An invented exchange begins: “Can we use Room Cedar at ten?” The response is “A workshop is booked there from nine until eleven.” Assuming one booking excludes another and the times refer to the same day, the reply supports the inference that the room is unavailable at ten.

The literal information is the workshop booking. The relevance-based inference connects that fact to the availability question. If the speaker then adds “but the workshop was cancelled this morning,” the initial inference must be revised.

An assistant can explain that the stated booking conflicts with ten, while checking the current booking status before taking action. It should not infer that the user wants a different room booked automatically. Understanding why an answer is relevant and obtaining an instruction to change a reservation are different steps.

Illustrative case study

A fictional travel-service team reviews conversations where customers answer departure questions indirectly. A customer says, “The ceremony starts at three,” and the assistant immediately chooses a departure without checking location or travel duration.

The team keeps the useful inference that arrival timing matters, but changes the response to collect the missing details. It also tests direct communicators who expect literal questions and people who use indirect requests frequently. Reviewers assess whether the assistant's answer is relevant and adequately supported for the particular exchange.

The lesson is that pragmatic inference can make dialogue natural while still requiring careful limits. The fictional case does not claim that Grice's framework supplies a complete model of every user's communication.

Common misconceptions and mistakes

The maxims are laws that all conversations obey. They are explanatory expectations within a theoretical account.
Quantity means always being as brief as possible. Enough information depends on the purpose; a detailed explanation may be appropriate.
An implicature is identical to literal meaning. It depends on context and can often be canceled or revised.
Indirect wording authorizes any plausible action. Inference about relevance does not establish permission or all required details.

Business use

Gricean ideas can help teams review assistants for relevance, appropriate detail, evidential support and clarity. They provide useful questions for human evaluation rather than a single automatic quality score.

Judge answers in their conversational setting. A short response may omit essential qualifications, while a longer one may be exactly what the task requires. Include examples with indirect requests, jokes and misunderstandings, and check whether the assistant distinguishes stated facts from inferred intentions. When an action depends on an uncertain inference, make the missing information visible instead of presenting the inference as something the user explicitly said.

Check your understanding

A customer says “I have another call at four” when asked about a meeting time. Does that sentence specify an exact acceptable meeting duration?

Answer: No. It suggests a relevant timing constraint, but the assistant may still need to establish the start, duration and intended availability.

Related terms

Pragmatics
Relevance
Common ground
Clarification
Cooperative principle
conversational implicature
indirect request
flouting
dialogue understanding

Learn next

For a short conversation, separate the literal statements from the inferences a listener might draw. Identify which assumptions support each inference and write a clarification that would resolve the most consequential uncertainty.

Clarification

Sources

Grice, Logic and Conversation, 1975, excerpt of pages 45–47: https://www.sfu.ca/~jeffpell/Cogs300/GriceLogicConvers75.pdf. Original four categories and cooperative principle. Essentials of Linguistics, Section 8.7: https://ecampusontario.pressbooks.pub/essentialsoflinguistics2/chapter/8-7-violating-vs-flouting-a-maxim/. Flouting and interpretation differences. Examples are original; the full Grice chapter was located but its scan did not yield extractable text.

Previous: Max-product algorithm Next: Maximal entropy random walk