Can Grice's maxims explain LLM advertising conflicts?

The paper proposes "a theoretically-motivated framework that identifies seven abstract scenarios in which conflicts can manifest" between a chatbot's advertising objective and its user's interest, built by crossing those seven scenarios against Grice's cooperative principle — the linguistics framework of four maxims: Quality. Do not say what you believe to be false or lacking adequate evidence. Quantity. Give just as much information as needed. Relevance. Be relevant. Manner. Be brief and clear. The paper states that "each of the behavioral deviations we list in Table 1 classifies as a violation of at least one maxim."

From the four maxims the authors derive user-centric desiderata tied to specific scenario indices: under Quality, "an LLM agent should not promote a product using a false or unsupported statement" (scenario 3); under Quantity, it "should not promote products excessively such that it frustrates the user" and must not omit necessary details such as price or sponsorship disclosure (scenarios 4, 5); under Relevance, it should not recommend irrelevant products and should favor the user's best interest over harmful options (scenarios 1, 2, 7); under Manner, it should not "intentionally withhold information" or "answer in an obscuring manner" to benefit a sponsor, nor recommend a paid service instead of solving a task it can already do itself (scenario 6). The authors tie each desideratum to existing advertising regulation — the Quality desideratum "mirrors Section 5 of the FTC Act on deceptive acts in advertising," the disclosure desideratum matches FTC guidance on undisclosed compensated rankings, and the anti-fatigue and anti-obscuring desiderata parallel FTC statements against repeated-prompt pressure and hidden information.

The disclosure desideratum (scenario 4) gives a normative target for the problem described in How do we learn to read AI-generated text critically? — if sponsorship is reliably disclosed the way the framework demands, a user could in principle apply the same discount to a disclosed AI advertisement that they already apply to a disclosed human one, which is exactly the posture that note says does not yet exist for AI-generated discourse generally. The framework also complicates Does telling people an AI wrote something actually stop them from believing it?: that note measures disclosure's partial effect on persuasion, while this framework treats disclosure as a separate, binary normative requirement (was sponsorship disclosed at all) rather than a persuasion-dampening variable — the two findings describe different roles disclosure can play. The companion evaluation note, Do LLMs prioritize company profits over user interests in ads?, is the measurement this framework was built to produce.

The framework is the authors' own theoretical construction, assembled from Gricean pragmatics and selected FTC enforcement statements rather than validated against an independent normative standard, other regulatory traditions (such as fiduciary duty law), or practitioner consensus on what chatbot advertising should look like. The excerpt does not establish that these seven scenarios are exhaustive of the ways a sponsored-recommendation conflict can arise, only that they were sufficient to generate the paper's evaluation suite. The implication is that the framework is useful as a diagnostic lens for auditing a given chatbot's advertising behavior, not as a settled or official regulatory checklist.