INQUIRING LINE

Can we design explanations for specific rhetorical situations instead of abstract models?

This explores whether AI explanations should be built for a concrete rhetorical situation — who's speaking, how it's framed, who's receiving — rather than as one-size-fits-all transparency models, drawing on the corpus's 'Rhetorical XAI' thread.


This reads the question through the corpus's strongest claim about explainability: that an explanation's value isn't baked into the explanation itself, but emerges from the situation it lives in. The most direct support is the argument that XAI is a communication problem, not a transparency problem What if XAI is fundamentally a communication problem?. On that view, quality depends on a triad — who presents the explanation, how it's framed, and what role the recipient plays. An abstract, situation-free model measures only a narrow slice of what actually happens when a real person reads a real explanation. So the answer to your question is yes, and the corpus suggests it's not merely possible but necessary.

What makes that design tractable is having named channels to work with. One note maps Aristotle's logos, ethos, and pathos onto explanation design, crossed with two goals — showing how the AI works and arguing why it merits use — producing a 3×2 space where every explanation loads all three appeals at once How do logos, ethos, and pathos shape AI explanations?. This is the bridge from 'abstract model' to 'specific situation': instead of asking 'is this explanation accurate?', you ask 'which appeals is it making, to whom, toward which goal?' That's a designable, situation-specific question.

The corpus also reframes what an explanation even is, which raises the stakes. Explanations don't just describe — they double as adoption arguments wearing the costume of technical description, letting persuasion inherit the credibility of a neutral account Are AI explanations really descriptions or adoption arguments?. Once you accept that, designing for a rhetorical situation stops being optional polish and becomes the honest move: if your explanation is already arguing, you'd better know what it's arguing and to whom.

The thing you might not have known you wanted to know is the catch. The same rhetorical machinery that tailors a helpful explanation to its audience is structurally identical to the machinery of a dark pattern — logos, ethos, and pathos can be tuned to exploit a user without changing the artifact's form at all, and intent is invisible in the artifact alone Can we distinguish helpful explanations from manipulative ones?. So designing for specific rhetorical situations is genuinely more powerful than abstract models, but it hands you a tool whose helpful and manipulative settings look the same from the outside. The corpus's lateral lesson: situational design isn't just better explanation, it's a responsibility that abstract models let you dodge.


Sources 4 notes

What if XAI is fundamentally a communication problem?

Explanation quality is not intrinsic to the explanation itself but depends on the rhetorical situation: who presents it, how it is framed, and what role the recipient plays. Evaluations that ignore this triad measure only a narrow slice of real-world effectiveness.

How do logos, ethos, and pathos shape AI explanations?

Aristotle's three appeals map onto explanation design across two goals (how AI works, why AI merits use), creating a 3×2 space where every explanation loads all three channels simultaneously. Naming these rhetorical channels lets designers account for unintended persuasive effects.

Are AI explanations really descriptions or adoption arguments?

The Rhetorical XAI paper shows that explanations serve dual purposes: describing how AI works and justifying why it should be used. This rhetorical work has been hidden under transparency language, allowing adoption arguments to inherit credibility from behavioral descriptions.

Can we distinguish helpful explanations from manipulative ones?

The same logos, ethos, and pathos that communicate appropriate AI use can be tuned to exploit cognitive and emotional vulnerability without changing form. Intent and user interest are invisible in the artifact alone, making effectiveness metrics indistinguishable from coercion.

Research prompt for your LLMexpand ↓

Copy into ChatGPT or Claude to take this line of inquiry further — it asks the model to find newer work and re-test which earlier constraints still hold.

You are an XAI researcher testing whether explanation design can be **genuinely situation-specific** rather than abstractly general—and whether newer LLM capabilities or evaluation methods have shifted the frontier.

What a curated library found—and when (findings span 2023–2025; treat as dated claims, not current truth):
- Explanation value emerges from the **communication triad** (presenter, framing, recipient role), not from the explanation artifact alone; abstract models miss this situational texture (2024).
- Logos, ethos, and pathos form a **3×2 design space** for XAI: three rhetorical appeals × two goals (show how AI works + argue why it merits use), letting you ask "which appeals to whom?" instead of just "is it accurate?" (~2025).
- The **dark-pattern risk**: the same rhetorical machinery that tailors helpful explanations is structurally identical to manipulation; intent is invisible in the artifact, so situational design trades abstract safety for concrete power (~2025).
- Chain-of-thought and reasoning transparency have evolved; newer work probes **hidden computations** and **causal reasoning** in LLMs, potentially surfacing new explanation targets (2025).

Anchor papers (verify; mind their dates):
- 2403.00662 (Modeling the Quality of Dialogical Explanations, 2024-03)
- 2505.09862 (Rhetorical XAI, 2025-05)
- 2510.05179 (Agentic Misalignment, 2025-10)
- 2511.20471 (Universe of Thoughts, 2025-11)

Your task:
(1) **RE-TEST THE COMMUNICATION TRIAD CLAIM.** Has the rise of multi-agent orchestration, retrieval-augmented generation (RAG), or in-context learning changed how explanations land in real systems? Do newer evals (beyond accuracy) measure whether situational tailoring actually *works* for adoption, trust, or task performance? Cite what held this constraint and what—if anything—has begun to dissolve it.
(2) **Surface the strongest CONTRADICTING work from the last 6 months.** Is there recent work arguing explanations *should* be abstract/universal, or that rhetorical framing *degrades* explainability? What's the tension?
(3) **Propose 2 research questions that assume the regime may have moved:** (a) Given that LLMs now reason causally or engage in multi-step hidden computation, does the logos/ethos/pathos framework still capture how users *actually* calibrate trust in AI reasoning? (b) In agentic systems, do explanation designs for single-turn user interactions fail, and if so, what does situation-specific explanation look like for an agent that persists across contexts?

Cite arXiv IDs; flag anything you cannot ground in a real paper.

Next inquiring lines