Analysis Lenses — A New Way to Examine Text Beyond Fallacies

TellDear can detect 535 distinct argument patterns across six analytical dimensions. That's powerful — but it answers only one question: What patterns are present in this text? There's a second question that matters just as much: How does this text operate on a meta-level? That's what Analysis Lenses are for.

The Problem: Dimensions Alone Aren't Enough

Imagine you're analyzing a politician's response to a crisis. TellDear's dimension analysis might flag an Appeal to Authority (D1), some Loaded Language (D2), and a dash of Optimism Bias (D3). Useful findings, each one. But they miss something that any attentive reader would notice immediately: the entire statement is hollow. It sounds like it says something without actually saying anything at all.

"We take this matter very seriously. Our thoughts are with those affected. We are on a good path, and we will continue to work tirelessly to ensure that the highest standards are met." Recognize this? It could come from virtually any press conference, any corporate crisis response, any politician dodging a question. The words fill space. They simulate concern. They deliver precisely nothing.

Or consider a news article that's factually accurate, logically sound, and avoids every propaganda technique in the book — yet frames an entire ethnic group through stereotypes so subtle they barely register consciously. The dimension analysis comes back clean. The article is anything but.

The problem isn't that dimensions are wrong. They're excellent at what they do: classifying what an argument pattern is. A logical fallacy, a cognitive bias, a propaganda technique, a statistical error, an argumentation scheme, a discourse mechanic. Six categories, clearly defined. But dimensions are vertical — they cut the world into types. They don't tell you about cross-cutting properties that span all types.

The Solution: Analysis Lenses

This is where lenses come in. An Analysis Lens is an additive perspective that cuts across all six dimensions. It doesn't replace the dimension analysis — it augments it.

Think of it this way: if dimensions tell you what ingredients are in the dish (flour, sugar, eggs, butter), then lenses tell you what dietary profile the dish has (high-sugar, gluten-containing, vegetarian). The ingredient list and the dietary profile answer different questions about the same dish. You need both.

Technically, each of TellDear's 535 aspects can belong to zero or more lenses. A single aspect — say, Thought-Terminating Cliché — lives in Dimension 2 (Manipulation & Propaganda) but might also be tagged as Hollow Rhetoric and flagged for Discourse Quality analysis. The dimension is its home address; the lenses are the clubs it belongs to.

The First Two Lenses

💨 Hollow Rhetoric — 15 Aspects

The Hollow Rhetoric lens identifies language that simulates meaning without delivering content. These are phrases and patterns that sound like communication while carefully avoiding it.

Harry Frankfurt, in his delightfully titled philosophical essay On Bullshit (2005), drew a crucial distinction: a liar cares about the truth (and deliberately inverts it), while a bullshitter is indifferent to truth entirely. The bullshitter's words are chosen for effect, not accuracy. That indifference is what makes hollow rhetoric so corrosive — and so common.

George Orwell saw it coming. In "Politics and the English Language" (1946), he described political language as "designed to make lies sound truthful and murder respectable, and to give an appearance of solidity to pure wind." Eighty years later, the wind blows harder than ever.

The 15 aspects in this lens include patterns like:

• Performative Concern — "We take this very seriously" (without any action following)
• Aspirational Vagueness — "We're committed to excellence" (what does that mean, concretely?)
• Non-Denial Denial — "I don't recall" (technically not a lie, functionally a dodge)
• Responsibility Diffusion Language — "Mistakes were made" (by whom?)

These patterns span multiple dimensions. A Non-Denial Denial is a logical evasion (D1), but it's also a manipulation technique (D2) and a discourse mechanic (D6). The Hollow Rhetoric lens ties them together under a single analytical question: Is this text producing meaning or merely simulating it?

🎯 Discrimination Detection — 10 Aspects

The Discrimination Detection lens identifies patterns of discriminatory language and framing — from overt racial stereotyping to subtle dog whistles, from ableist metaphors to tokenistic representation.

This lens operates on a spectrum, not a binary. "That neighborhood is dangerous" might be straightforward safety advice or a racial dog whistle — context determines which. The lens doesn't deliver verdicts; it flags patterns and lets analysts evaluate the context.

The 10 aspects include:

• Racial Stereotyping — Attributing characteristics to individuals based on racial categories
• Dog Whistle Language — Coded language that carries a secondary, discriminatory meaning for certain audiences
• Ableist Framing — Using disability as metaphor ("blind to the facts", "lame excuse") or framing disability as deficit
• Tokenism — Presenting superficial inclusion as proof of equality

What makes this lens distinctive is its cross-dimensional reach. Discriminatory patterns can manifest as logical fallacies (Hasty Generalization applied to groups), propaganda techniques (Scapegoating), cognitive biases (In-Group Bias), or discourse mechanics (Othering). No single dimension captures the full picture. The lens does.

What's Coming Next

Hollow Rhetoric and Discrimination Detection are the first two lenses, but they won't be the last. We're exploring several additional perspectives:

• Media Bias — Drawing on 38 bias types from the Table of Media Bias Elements to assess how reporting deviates from neutral coverage
• Discourse Quality — Scoring argumentative quality using Pragma-Dialectics (after van Eemeren) to measure how productively a discussion unfolds
• Emotional Framing — Mapping the emotional architecture of a text beyond simple sentiment analysis
• Fact-Checkability — Identifying which claims in a text are empirically verifiable and which are opinion dressed as fact
• Ideology Detection — Tracing underlying ideological assumptions without reducing complex positions to a left-right spectrum

Each lens adds a new question you can ask of any text, and each question reveals patterns that dimension analysis alone would miss.

The Architecture: Dimensions + Lenses

TellDear now operates on a two-layer model:

• Dimensions (vertical, categorical) — The six traditional categories that classify what kind of pattern something is
• Lenses (horizontal, cross-cutting) — Additive perspectives that group aspects by how they function, regardless of dimension

In the Analyzer, you'll find toggle buttons for each available lens. Activate a lens and the analysis focuses on the aspects relevant to that perspective. In the Aspect Directory, you can filter by lens to explore which aspects contribute to each perspective. And on individual aspect detail pages, badges show which lenses an aspect belongs to.

For developers and researchers, the lens data is available through the API at /api?action=lenses, returning the full mapping of aspects to lenses.

Why This Matters

The world doesn't need another fallacy detector. What it needs is a richer vocabulary for analyzing discourse.

A politician's speech might be logically valid, free of obvious propaganda, and statistically accurate — and still be a masterpiece of hollow rhetoric that communicates nothing while appearing to communicate everything. A news article might avoid every fallacy in the catalog but frame its subject through a discriminatory lens so consistent that it reshapes the reader's perception of an entire community.

These aren't edge cases. They're the norm. Most problematic discourse doesn't announce itself with obvious fallacies. It operates at the meta-level — through patterns that no single dimension captures but that lenses make visible.

Critical thinking has long focused on identifying errors in reasoning. That's necessary but not sufficient. We also need to identify patterns in how language is used — patterns that cross categorical boundaries, that operate through accumulation rather than individual instances, that shape discourse at a level dimensions weren't designed to reach.

That's what lenses do. They give you new questions to ask of any text. And better questions lead to better understanding.

Try It Yourself

Head to the TellDear Analyzer, paste any text — a political speech, a corporate press release, a news article, a social media thread — and toggle the lens filters. Watch how the same text reveals different patterns depending on which question you're asking.

Because the most important thing about analyzing a text isn't finding the answer. It's knowing which questions to ask.