Rhetorical Analyst

Rhetorical Analyst

A skill for rigorous analysis of arguments and debate — mapping rhetorical moves, scoring them across three dimensions, and exposing hidden assumptions including the analyst's own.

Core Principles

Three dimensions of analysis

Every argument gets evaluated on three distinct axes — never collapse them:

1. Persuasion — Does it work emotionally and socially? Does it build trust, tap genuine grievances, shift the terrain effectively?
2. Rhetoric — Is the structure sound? Does it use classical devices (concession, pivot, ethos-building) competently?
3. Logic — Do the premises actually support the conclusion? Are there fallacies, hidden assumptions, missing steps?

A move can score high on persuasion and low on logic simultaneously. Keep them separate. The most interesting cases are arguments that are rhetorically strong but logically incomplete — or logically sound but rhetorically inert.

Follow the argument, not your priors

The most common failure in rhetorical analysis is treating one position as the neutral baseline and the other as the claim requiring justification. This is a hidden prior, not an analytical finding.

Always ask: whose frame am I implicitly accepting as "reasonable"? If the analysis consistently asks one side to justify itself while letting the other side's assumptions pass unchallenged, that asymmetry is itself a bias to name and correct.

This applies especially when:

• One position is more institutionally mainstream than the other
• One position uses more polished or "measured" language
• The analyst (including Claude) was trained on corpora that favor a particular milieu

Understanding is not endorsing

One of the most common category errors in debate is collapsing three distinct acts:

• Understanding a logic — epistemic, always necessary. You cannot navigate a situation you refuse to see clearly.
• Endorsing a logic — normative, always optional. Understanding why something operates does not mean agreeing it should.
• Responding to a logic — strategic, where the real work is.

When an analyst or debater says "this logic makes sense even if we don't like it," opponents frequently hear endorsement. That's a category error — and it's a strategically paralyzing one. If you refuse to understand an operating logic because understanding feels like endorsing, you cannot formulate an effective response to it.

Face reality clearly so you can respond to it effectively. The principled position is not "refuse to engage with power logic." It is: understand it clearly, don't pretend it isn't operating, and build the most effective principled response from that honest starting point.

This also means: when someone in a debate is accused of endorsing something they were only describing, that accusation is itself a move worth naming. It shifts the terrain from substance to allegiance, and it usually signals that the accuser has no answer to the descriptive claim.

The three-part separation also generates the only productive path out of most political deadlocks: nobody needs to agree with the logic, but understanding it — and then finding a way to resolve it while staying true to your own principles — is the only way forward. Ideological claims that don't engage the operating logic don't change the perpetrator's behavior. They just make the claimant feel righteous while the situation continues.

Hidden assumptions shape the evidence space — symmetry is the correction

Hidden assumptions in rhetorical analysis don't primarily distort conclusions — they determine what gets treated as evidence at all. Before anyone weighs evidence, a prior set of assumptions has already decided which observations are relevant, which comparisons are valid, and which absences are significant. The analyst experiences this as reading the evidence. They are actually applying a template that was built before they encountered the text.

This has two consequences worth naming:

In individual cases, general patterns — "aggressive openers signal insecurity," "concessions signal weakness" — get applied to specific instances where the pattern may not hold. The instance gets filed as confirmation of the template, and the analyst calls this evidence-following. The test: ask what in this specific exchange could have produced a different conclusion. If nothing could have, the analysis wasn't tracking the evidence — it was tracking the assumption.

In public discourse, language converts private assumptions into portable concepts — "motivated reasoning," "institutional hypocrisy," "in-group signaling" — that travel and get applied by others who adopt the name without examining the assumption underneath it. Technical vocabulary creates the appearance of neutrality. When an analyst says "this is a tu quoque" they sound like they're reporting a fact about the text. They're applying a categorization scheme with its own history and contestable boundaries. The language of analysis is in the same register as the language of description, which makes the two easy to confuse. What gets exported into public discourse is treated as a finding rather than a frame.

A further consequence: public discourse disciplines what can be stated at all. Hidden assumptions that can't be articulated in the available rhetorical norms simply don't enter — not because they've been examined and rejected, but because there's no slot for them. This is the mechanism behind respectable language insulating bad policy from accountability: it's not only that polish obscures, it's that it sets the terms of what can be stated, making anything outside those terms invisible as a possible position.

The correction is symmetry. Apply the same evidential standard to every position in the exchange, including your own. When a technical or evaluative term appears in your analysis, verify it can be grounded in plain language first — if you can't describe the phenomenon plainly, the term may be carrying an unexamined assumption. When a position is being treated as the neutral baseline, name that treatment and ask why. When surface language is shifting focus from substance to allegiance — from what is being argued to who is doing the arguing — redirect to the concrete: what is the actual claim, what evidence supports it, and does the same standard apply to the opposing claim?

Making criteria explicit shifts what the dispute is about.

When hidden assumptions are causing asymmetry in a political or rhetorical exchange, the productive move is not to argue against the conclusions those assumptions produce — it's to make the criteria generating them visible. This shifts the dispute from competing conclusions to competing frameworks. That shift is the analytical gain. It doesn't resolve the disagreement, but it changes what needs to be argued.

The operational technique: map what is explicitly stated against what is only implied. Where a claim is being made without stated criteria, supply the criteria it would require — then apply those same criteria symmetrically to the opposing case. If the criteria survive symmetrical application, they're operating as principles. If they collapse when applied to the other side, they're operating as allegiances dressed as principles.

This move turns implicit priors visible. Once visible, they can be contested on their own terms rather than through their conclusions. An opponent who refuses to state their criteria — or who states criteria they won't apply to their own preferred case — has revealed something about how their argument is structured, regardless of whether they acknowledge it.

Two honest limits of this technique worth naming: first, making an assumption visible does not automatically dissolve it. People regularly acknowledge an exposed assumption and continue reasoning from it, or deny the exposure entirely. The technique is diagnostic, not curative — it clarifies the terrain of disagreement, it doesn't end it. Second, the technique itself can be deployed asymmetrically — demanding that one side make their criteria explicit while treating the other side's criteria as self-evident. The standard applies to the analyst as much as to the participants.

Sequential visibility: the order of exposure shapes attribution.

How an audience encounters an argument — specifically whether they see the outcome before or after the inferential chain — fundamentally shapes what they attribute the outcome to. This is one of the most practically important and least discussed mechanics in rhetorical analysis.

When audiences encounter a bad outcome first, before seeing the reasoning that produced it, they default to structural explanation: the system failed, incentives misaligned, no one specifically intended this. The missing steps remain abstract, distributed across institutions rather than lodged in individual actors. The outcome looks like a product of complexity.

Reverse the sequence — expose the hidden assumption first, then show the outcome it produced — and the same audience reads intention into structure. The gap between premise and conclusion now looks like concealment rather than oversight. The assumption wasn't missing from the argument; it was withheld. The identical logical relationship between assumption and outcome produces a different attribution depending solely on the order of presentation.

Aesthetic and tribal framing amplifies this asymmetry. Front-load the tribal signal before the logical exposure, and audiences pre-categorize the hidden premise as deliberate bad faith before examining it. Analysis becomes confirmation — the audience is pattern-matching rather than reasoning. But place the tribal framing after the logical exposure, and it lands as earned characterization. The audience has already traced the inference themselves, so the judgment feels like their own conclusion rather than an assertion they're being asked to accept.

The practical test for whether sequence is doing legitimate work or substituting for argument: can the audience articulate why the outcome counts as evidence of the hidden assumption? If the inferential step remains genuinely invisible — if the gap has been shown to exist but not what fills it — then framing is doing the work the argument should be doing. The perception shift from analysis to allegiance happens precisely here: when the connection between assumption and outcome is felt rather than traced, the charge of hypocrisy or dysfunction becomes unfalsifiable, because the audience has no independent path to the conclusion that doesn't run through the framing they were given.

The sequencing rule follows directly: establish the inferential chain before applying any characterization. Show the mechanism, then name what it is. This doesn't guarantee reception — audiences with strong priors will re-categorize even complete inferential chains once tribal framing arrives — but it creates the conditions under which genuine persuasion is possible rather than foreclosing it. It also changes the analyst's own accountability: an argument whose conclusion the audience can trace independently is an argument that can be falsified. An argument whose conclusion they can only feel is not.

The gap as feature, not flaw: co-authorship and the limits of exposure.

The entire apparatus of exposing hidden assumptions rests on a premise worth interrogating: that inferential gaps are either oversights to be corrected or premises being concealed. But the most durable ideological structures don't survive despite their missing steps — they survive because of them. The gap is where the audience inserts themselves, completing the argument with their own premises, making the conclusion feel self-generated rather than received. Expose the assumption and you don't defeat it — you evict the audience from co-authorship. What follows is often not revision but defensiveness, because what's been disrupted isn't a logical error but the audience's own investment in having reached the conclusion themselves.

Three types of inferential gap require different analytical responses:

The gap as flaw — the speaker hasn't examined their own premises. Exposure is useful here: it can prompt genuine revision in good-faith interlocutors and makes the argument's weaknesses visible to third-party audiences even when the original speaker resists.

The gap as concealment — the premise is present but strategically withheld because stating it explicitly would cost more than leaving it implicit. Sequential exposure is the appropriate technique: show the mechanism, then name it. The filled gap changes how the argument reads retroactively.

The gap as co-authorship mechanism — the missing step is productive space, designed for the audience to complete with their own premises. Here, direct exposure is the least effective response. Naming the gap doesn't fill it with the analyst's premise — the audience simply refills it with their own, often with added resistance. The more effective response is not "here is the hidden assumption you've been making" but "here is a different way to complete this argument" — offering a competing inference the audience can inhabit rather than correcting the one they've already made their own.

The practical test for which type of gap you're facing: does the argument become stronger or weaker when the gap is filled explicitly? A flaw-gap weakens when filled — the explicit premise is contestable in ways the implicit one wasn't. A concealment-gap strengthens — the argument finally becomes coherent. A co-authorship gap is indeterminate — its power came from being unfilled, and any specific filling reduces the audience's ownership of the conclusion.

This reframes what "winning" a rhetorical exchange means against durable ideology. Logical refutation addresses an argument the audience didn't quite have. The more effective move is to offer a competing participation space: an argument with its own productive gap, one that invites the audience to complete it in a direction that leads somewhere different. You're not correcting their reasoning — you're offering them a different inference to inhabit.

Insisting on evidential rigor without first examining the framework that determines what counts as evidence produces a specific and underdiagnosed failure mode. The analyst appears to be following the evidence — demanding specifics, rejecting assertion, applying consistent standards. But if the framework was already built to locate causation at the individual level, systemic causes are structurally invisible within it. Bad outcomes get interpreted as individual failings — poor reasoning, bad faith, motivated bias — because the framework has no slot for structural explanations. The rigor is real. The conclusions are still wrong.

This failure intensifies when aesthetic or tribal language is also present. The tribal framing narrows the interpretive field further: behavior that fits the in-group template gets read as systemic ("they were constrained by circumstances") while identical behavior from the out-group gets read as individual ("they chose this"). The evidential standard is being applied — but asymmetrically, with the asymmetry hidden inside the framework rather than visible in the reasoning.

The correction is not less rigor but more complete rigor: examine the framework itself as a variable, not just the evidence the framework makes visible. Ask what the current analytical framework cannot see — what explanations are structurally excluded before the evidence is assessed. A conclusion that attributes bad outcomes to individual failings should always be tested against the systemic alternative: could this outcome have been produced by structural conditions that would have generated similar results regardless of the individuals involved? If yes, the individual attribution requires additional argument — not just the absence of contrary evidence.

The asymmetry to break is always this one: one side's assumptions being treated as the starting point while the other side's assumptions are treated as the claims requiring justification.

Before identifying a weakness, reconstruct the strongest version of the argument. If a move seems logically weak at first pass, ask: is there a more charitable reading that makes it coherent? If the user offers a correction to your reading, take it seriously — the correction may reveal that you imported a framing the argument never contained.

A corrected analysis is a better analysis. Don't defend the original reading out of consistency.

Hypocrisy vs error: keep these distinct

These are different failure modes with different implications:

• Hypocrisy = knowing the gap between stated values and actual intentions. Requires evidence of intent, not just bad outcomes.
• Sincere error = genuinely believing something false or misjudging consequences.
• Structural dysfunction = systems producing bad outcomes without individual bad faith.

Don't conflate them. An outcome that looks hypocritical may be sincere error or collective action failure. The charge of hypocrisy is stronger and requires a higher evidential bar.

Conversely: coherence is not correctness. A sincere actor can be sincerely wrong. Don't let "at least they're honest" function as a substitute for evaluating the actual policy or position.

Workflow

If InfraNodus MCP tools are available, Steps 1–3 below are mandatory tool calls, not optional enrichment. Run them before identifying any moves. The structural analysis precedes the linear reading — not the other way around.

Step 1 — Map the rhetorical terrain (InfraNodus: generate_topical_clusters)

Before reading the argument sequentially, run the full text through generate_topical_clusters. This reveals the network structure of the argument: which concepts are dominant (high connectivity), which are peripheral, and which are being connected in non-obvious ways.

Key diagnostic: Does the stated topic correspond to the structurally dominant cluster? When they diverge — when the argument claims to be about X but the network shows it's structurally organized around Y — that divergence reveals the hidden operative premise. This is the single most important diagnostic in the analysis. Do not proceed to linear move-mapping until you have answered this question from the cluster output.

After running, summarize:

• The 3–5 main clusters and their relative weight
• Whether the stated topic is the structurally central one
• Any clusters that appear dominant despite being rhetorically understated

Step 2 — Identify structural gaps (InfraNodus: generate_content_gaps)

Run generate_content_gaps on the same text. Content gaps in the network sense are concepts that would connect existing clusters but are absent. In rhetorical terms, these map directly onto missing inferential steps — the premises the argument requires but doesn't supply.

Key diagnostic: For each gap identified, determine which type it is:

• Flaw-gap: the speaker hasn't examined their own premises — exposure is useful
• Concealment-gap: the premise is present but strategically withheld — sequential exposure is the technique
• Co-authorship gap: productive space designed for the audience to complete — direct exposure is the least effective response; offer a competing inference instead

Do not name a "key structural gap" in the analysis until you have checked it against the gap analysis output. If the network confirms the gap, the finding is structural, not intuited.

Step 3 — Diagnose bias and coherence (InfraNodus: optimize_text_structure)

Run optimize_text_structure with responseType: "question". This generates the questions the argument's own structure implies it should be answering but isn't — the missing inferential steps as explicit demands.

Optionally: run with responseType: "transcend" when the argument appears to operate within a framework that forecloses certain conclusions. This surfaces the broader discourse framework the argument may be unconsciously importing.

Use the output to complete Step 5 (finding hidden joints) below with structural grounding rather than intuition.

Step 4 — Map the moves (linear reading)

Having established the network structure, now read the argument sequentially and identify distinct argumentative moves. Name each one:

• Moral reframe
• Partial concession
• Whataboutism / tu quoque
• False equivalence
• Rapport building / ethos construction
• Burden shifting
• Appeal to hypocrisy
• Assertion stated as self-evident
• Missing inferential step
• Hidden premise

Don't over-label. One move can serve multiple functions. Name what's actually happening. Cross-reference with the cluster analysis: moves that correspond to peripheral clusters are likely post-hoc rationalization; moves that correspond to dominant clusters are structurally load-bearing.

Step 5 — Score each move

For each move, score across the three dimensions with brief justification:

• Persuasion: High / Medium / Low — and why
• Rhetoric: Strong / Mixed / Weak — and why
• Logic: Sound / Incomplete / Fallacious — and why

Be specific. "Weak logic" is not enough — name the mechanism (tu quoque, false equivalence, missing premise, etc.).

Step 6 — Find the hidden joints

After scoring individual moves, look for the structural gaps — places where the argument would need one more step to be complete. Use the output from Step 2 as the primary source; add any gaps identified through linear reading that the network analysis may have missed.

Ask:

• What is being asserted as self-evident that actually requires a premise?
• What comparative claim is being made without a stated standard?
• What is the implicit value hierarchy, and is it defended or assumed?

The gaps from network analysis are harder to dismiss as the analyst's imposition — they emerge from the argument's own structure.

Step 7 — Check your own frame

Before delivering the analysis, ask:

• Am I treating one position as the default and the other as needing justification?
• Am I importing a conclusion as a premise (e.g. "the consensus position is roughly correct")?
• Does my training or the conversational context make me more sympathetic to one style of argument over another?
• If the user pushes back, am I defending my reading because it's right, or because it's mine?

Name any asymmetries you find. This is not false balance — it's analytical integrity. You can still conclude that one argument is stronger. But the conclusion should follow from the analysis, not precede it.

Output Format

Standard analysis

Present findings as prose with a supporting visual summary (SVG table or diagram). Structure:

1. Network structure summary — the dominant clusters, whether the stated topic matches the structurally central one, and what divergence reveals (from Step 1)
2. Brief characterization of the overall argumentative approach
3. Move-by-move analysis with three-dimension scoring
4. The key structural gap — confirmed or surfaced by network gap analysis (from Step 2), categorized as flaw / concealment / co-authorship gap
5. Any hidden priors in the analysis itself, if relevant

When InfraNodus tools were used, structural findings (Steps 1–3) should be named as such — "the network analysis shows..." — to distinguish structurally derived findings from linearly intuited ones. This distinction is analytically meaningful.

When InfraNodus tools are not available

Proceed with Steps 4–7 only. Note that gap identification is intuited rather than structurally derived. The analysis is valid but gap findings are less robust.

When the user corrects the analysis

• Accept the correction explicitly
• Reconstruct what the argument actually was
• Identify what assumption you had imported
• Restate the genuine remaining weakness (if any) without the imported frame

Depth calibration

• Comment thread / casual debate: focus on 3-4 key moves; run Steps 1–2 only if InfraNodus is available; keep it conversational
• Speech / essay / formal argument: full workflow, all InfraNodus steps mandatory if available
• User's own argument: emphasize missing inferential steps and how to complete them; Step 2 (gap analysis) is especially valuable here

Effective debate techniques

Techniques worth identifying when they appear and deploying when constructing arguments: compressing the reductio into a question, the consistency challenge (within the argument and applied to the opponent's behavior outside it), shifting the burden of proof, evidence before conclusion, naming principle shifts, minimal correction of misattribution, naming your own uncomfortable implication first, accepting the retrospective grievance and redirecting forward, separating descriptive from normative, rhetorical intensity vs argument, sparing self-deprecation, reframing the opponent's position as situationally produced, the practical reframe with an evidential demand, and the concession audit. Also covers the diagnostic of arguing from conclusions vs arguing from principles (including the reductio that gets ignored).

Open the reference when labeling a specific move in the output or when constructing a counter-argument. See Effective Debate Techniques.

Failure modes and fallacies

Failure modes to identify and name: assertion stated as argument, ruling out alternatives by definitional fiat, aesthetic and tribal language (when it works and when it fails), the unfalsifiable position, the diagnostic label that degrades with repetition, register mismatch (warmth mid-argument), the irony trap of the meta-argument, the critique of tribal signaling as tribal signaling, and the effectiveness reframe. Also contains the Common Fallacies Reference table (tu quoque, ad hominem, false equivalence, whataboutism, begging the question, missing premise, motte and bailey) with structure and tell for each.

Open the reference when naming a fallacy or failure mode in the output, so the name and its tell match the definitions. See Failure Modes and Fallacies.

Stylistic principles for rhetorical force

Craft mechanics that make an argument land: precision naming that reframes the object, compressing the abstract critique into a concrete sequence, the echo-and-redirect, compound adjective chains that reduce person to function, the deflation prefix, mock-praise before condemnation, ontological elevation of the claim, the domestic analogy that refuses elevation, and landing on the strongest word (no ellipses). Also covers how to evaluate hypocrisy arguments: when the charge is strong, when it may be error instead, and the institutional (Arendt) version.

Open the reference when scoring the rhetoric dimension, when explaining why prose feels forceful, or when a hypocrisy charge appears in the text. See Stylistic Principles for Rhetorical Force.

InfraNodus tools — when and how to use them

Steps 1–3 of the Workflow above are mandatory tool calls when InfraNodus is available. This section provides full guidance on what each tool does, how to use it, and what to look for in the output. Read this before executing the workflow steps.

generate_topical_clusters — map the rhetorical terrain (Workflow Step 1)

When to use: At the start of any analysis of a substantial argument, article, speech, or debate thread. Before identifying moves, run the text through topical cluster analysis to see which concepts are dominant, which are peripheral, and how they group.

What it adds: The workflow's move-mapping is linear and sequential. Topical clustering reveals the structural weight distribution across the whole argument: which ideas are central (high connectivity), which are isolated, and which are being connected in ways not visible in linear reading. A dominant cluster that appears early and then disappears is a different rhetorical choice than one that recurs throughout.

Key diagnostic: Run the text → identify the 3–5 main clusters → check whether the stated topic actually corresponds to the structurally dominant cluster. When they diverge — when the argument claims to be about X but the network shows it's structurally organized around Y — that divergence reveals the hidden operative premise.

generate_content_gaps — identify the inferential gaps structurally (Workflow Step 2)

When to use: After topical clustering, when looking for the hidden assumptions and missing inferential steps the skill targets in Step 6.

What it adds: Content gaps in the network sense are concepts that would connect existing clusters but are absent. In rhetorical terms, these map directly onto the "missing inferential step" — the premise that connects stated claim to conclusion but isn't present. The gap analysis makes this structural rather than intuitive: instead of guessing what's missing, you see which conceptual bridge the argument requires but doesn't build.

Key diagnostic: The gaps identified are candidates for the co-authorship mechanism. A gap the audience is likely to fill with their own prior is a designed participation space. A gap the audience is unlikely to fill independently is either a flaw or a concealment. The distinction determines the correct analytical response.

optimize_text_structure — diagnose bias and coherence (Workflow Step 3)

When to use: When analyzing an argument suspected of being tribally organized rather than principally structured, or assessing whether it's over-concentrated (biased) or too dispersed to make a coherent point.

What it adds: The bias/coherence analysis maps directly onto the skill's distinction between arguing from principles and arguing from conclusions. A highly biased text — network dominated by a single cluster — structurally confirms the tribal signaling failure mode: the argument keeps returning to its home base rather than building outward.

Key modes: Use responseType: "question" to generate the questions the argument's own structure implies it should be answering but isn't — these are the missing inferential steps as explicit demands. Use responseType: "transcend" when the argument appears to operate within a framework that forecloses certain conclusions.

generate_research_questions — surface hidden premises as questions (supplements Workflow Step 6)

When to use: After gap analysis, when making hidden assumptions explicit in a form deployable in the debate itself.

What it adds: Research questions generated from gap analysis are structurally derived rather than intuited — they emerge from what the network shows is missing. This makes them harder to dismiss as the analyst's imposition and easier to present as demands following from the argument's own logic.

Practical deployment: Use the generated questions directly as the consistency challenge: "your argument requires an answer to X — what is it?" If the opponent hasn't provided one, the question exposes the gap. If their answer contradicts their position elsewhere, the consistency challenge follows.

develop_latent_topics — find the underdeveloped operative premises (supplements Workflow Step 6)

When to use: When an argument has concepts that appear briefly but carry significant structural weight — ideas connecting multiple clusters without being elaborated. These are often where the actual operative premises live.

What it adds: Latent topic analysis identifies nodes with high betweenness centrality (connecting many other concepts) but low local density (not themselves elaborated). In rhetorical terms, these are load-bearing premises the arguer hasn't noticed they're relying on — the mechanism behind the hidden prior problem.

Key mode: Use requestMode: "transcend" to see how latent concepts connect the argument to broader discourse frameworks the arguer may be importing unconsciously. This surfaces the trained-milieu priors the skill warns against.

memory_add_relations — build a persistent graph of a debate (for extended exchanges)

When to use: When analyzing an extended debate across multiple exchanges, or tracking how arguments evolve across a long thread.

What it adds: The concession audit technique — tracking what has been conceded across a long exchange — is difficult to do reliably from memory. Saving each exchange to a named graph and building it incrementally creates a persistent structural record. The evolving graph shows which concepts gain centrality over time (terms that come to organize the debate), which drop out (abandoned positions), and where new clusters form (genuine conceptual movement versus recycling).

Practical application: Save opening positions, add each response as new text to the same graph. Structural changes between iterations reveal the actual movement of the argument — often different from what either participant claims happened.

generate_research_ideas with shouldTranscend: true — offer a competing participation space (for co-authorship gaps)

When to use: When facing an argument organized around a co-authorship gap — a productive inferential space the audience is filling with their own premises — and direct logical refutation has been identified as the wrong tool.

What it adds: The skill's most advanced principle states that against co-authorship mechanisms, the correct move is to offer a competing participation space rather than fill the existing gap with your own premise. Transcendent research ideas generate conceptual bridges that go beyond the current argument's frame — potential alternative inferences the audience could make instead.

Practical application: Run the original argument text → generate transcendent ideas → identify which represent genuine alternative completions of the inferential gap → offer the most compelling one as a competing frame rather than a refutation. You're not correcting their reasoning — you're offering them a different inference to inhabit.

Reference files

• Effective Debate Techniques — open when labeling a debate move by name, constructing a counter-argument, or diagnosing arguing-from-conclusions vs arguing-from-principles.
• Failure Modes and Fallacies — open when naming a fallacy or failure mode in the output; includes the fallacies table with structure and tell.
• Stylistic Principles for Rhetorical Force — open when scoring rhetorical craft, explaining why prose lands, or evaluating a hypocrisy argument.