Does rational cooperation actually describe how AI communication works?
Gricean models assume good-faith rational agents coordinating meaning. But do AI systems designed to persuade—using credibility, emotion, and non-rational appeals—really operate under these assumptions? What happens when we drop the rationality premise?
The Rhetorical XAI paper makes a theoretical move that matters beyond XAI. It notes that Grice's maxims assume "people engaged in communicative interaction will do their best to get their message across, and in doing so will abide by a number of conversational conventions." In practice, communication often departs from these ideals. Rhetoric foregrounds what pragmatics idealizes away — credibility (ethos), affect (pathos), and non-rational influence — and treats them as constitutive of how communication actually works rather than as failure modes to be corrected. Pragmatic models of HCI communication, built on cooperative assumptions, cannot capture systems whose interfaces are designed to persuade.
This is a foundational point for any communication-centric account of AI. Pragmatic models treat language as a coordination instrument among rational agents trying to share understanding. Rhetoric treats language as a strategic instrument among situated agents trying to bring about adoption, change, action — and grants that affect, credibility, and non-rational appeals are first-class mechanisms, not noise. The two pictures are not on a continuum; they make different claims about what communication is. Treating AI systems through Gricean lenses presumes a cooperative interlocutor where there is, at minimum, a designed artifact with adoption-shaped incentives.
This is a theoretical sibling to the quasi prefix fails for communicative states because communication is constitutively intersubjective — you cannot weaken communication you can only eliminate it — both insights argue that imported philosophical frames (cooperative pragmatics, qualified mental-state language) miscarry when applied to AI communication because the underlying constitutive assumptions don't hold. And it is in productive tension with Does chain-of-thought reasoning reflect genuine thinking or performance?: the performative-CoT result shows that even within an apparently logical artifact (chain-of-thought), the rhetorical/performative dimension dominates on easy cases. Logos and pathos do not separate cleanly; performance bleeds into reasoning even at the token level.
For the Conversation Glossary project, this is foundational vocabulary for the tension between Habermas's ideal speech and Goffman/Bakhtin's situated communication. Language as event is rhetorical, not propositional, and AI systems live in event-time.
Inquiring lines that read this note 13
This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.
How do LLMs distinguish causal reasoning from temporal and semantic associations? How do formal dialogue structures reveal conversation coherence mechanisms? What makes AI persuasion effective and how can we counter it?- How do ethos logos and pathos shape AI persuasion under scrutiny?
- Why do logic-based arguments make AI persuasion feel objective and impartial?
Related concepts in this collection 2
This note in its neighbourhood — explore the map, then jump to a related concept in the list below.
Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph
-
Does chain-of-thought reasoning reflect genuine thinking or performance?
When language models generate step-by-step reasoning, are they actually thinking through problems or just producing text that looks like reasoning? This matters for understanding whether extended reasoning tokens add real computational value.
productive tension; rhetorical/performative dimension dominates even within ostensibly logical artifacts
-
What if XAI is fundamentally a communication problem?
Does explanation effectiveness depend on who delivers it, how it's framed, and who uses it? This challenges the dominant technical view that treats explanations as context-independent outputs.
sibling; the rhetorical situation is what idealized-rationality models cannot represent
Related papers in this collection 8
Papers most semantically related to this note, ranked by cosine similarity in the embedding space.
- GenAI as a Power Persuader: How Professionals Get Persuasion Bombed When They Attempt to Validate LLMs
- A meta-analysis of the persuasive power of large language models
- Rhetorical XAI: Explaining AI’s Benefits as well as its Use via Rhetorical Design
- Machine Bullshit: Characterizing the Emergent Disregard for Truth in Large Language Models
- How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs
- Collaborative Rational Speech Act: Pragmatic Reasoning for Multi-Turn Dialog
- Exploring the Role of Prior Beliefs for Argument Persuasion
- Large Language Models are as persuasive as humans, but how? About the cognitive effort and moral-emotional language of LLM arguments
Original note title
rhetoric breaks the idealized-rationality assumption baked into Gricean and pragmatic models of HCI communication