Sep 8, 2026

The Attention Equation: Why This Headline Is the Demonstration, Not the Introduction

A framework for engineering attention — Primitive × Delta × Novelty — explained by using itself on you, in real time, as you read it.

The Attention Equation: Why This Headline Is the Demonstration, Not the Introduction

BLUF (Bottom Line Up Front): Attention is not a talent. It’s arithmetic: Attention ≈ Primitive Salience (P) × Model Update (Δ) × Contextual Novelty (N). Most writing fails not because the idea is weak but because the encoding is exhausted — the audience has learned to recognize the shape of the sentence before reading its content. This essay explains that equation by refusing to just describe it. Every section is built out of the mechanism it names, including this one, which is why you are still here four sentences in.

Part 1: You Were Told This Was a Framework Post. It Isn’t — Yet.

Most framework posts open with a definition. This one is opening with a confession instead: I am about to teach you a system for engineering attention, and I am going to do it by using the system on the act of teaching it to you.

I didn’t do that for style. It’s the first move the framework prescribes: message the change, not the topic. A topic statement would read “here is a marketing framework called Primitive-Delta-Novelty.” Nobody’s RAS bothers to open that door. So instead:

I spent this past weekend finishing the core architecture of ClearPath, a mapper tool that hands small business owners a level of operations, finance, and tax-strategy reasoning that used to be cost-prohibitive for anyone but a company with a real finance department. The build was the easy part. The moment it was done, I ran into the same wall I always run into: I have no track record of making these tools feel urgent to the person who most needs one. Even the phrase “emotionally salient” is not emotionally salient. That failure, not the software, is what sent me looking for this framework.

Actor. Existing state. Change. Consequence. That’s the whole unit. You just read it, and — if it worked — you have a small, specific question now: if it isn’t creativity, what actually makes a true, useful thing land?

Hold that question. It’s the Δ, and Section 3 pays it off. First, the equation it belongs to.

Part 2: The Equation

Attention ≈ Primitive Salience (P) × Model Update (Δ) × Contextual Novelty (N)

P — Primitive Salience. How consequential is the underlying human concern to this audience, right now?
Δ — Model Update. How much does the message change what they believe, predict, fear, desire, or understand?
N — Contextual Novelty. How unexpected is the representation of that change, relative to everything else currently competing for the same eyeballs?

Δ is the term that actually does something to behavior — P and N just get you heard. Every reader arrives with a model already running: a working set of beliefs sufficient to get them through their day so far, and it doesn’t pause out of politeness while you talk. If nothing in the message diverges from that model, the reader finishes with the same model they started with — which means they keep doing whatever they were already doing, and that, by construction, doesn’t include the thing you wanted them to believe or do. High P with zero Δ isn’t persuasion. It’s relevance: the reader nods, agrees it matters, and changes nothing, because you never gave their model anything to update on.

It’s a product, not a sum. Zero out any one term and the whole thing collapses — a true, important claim delivered in a stale, over-familiar shape (P high, Δ high, N ≈ 0) gets scrolled past exactly as fast as an irrelevant claim delivered with genius production value (P ≈ 0). This is why “just be more creative” is bad advice: creativity is one lever on one of three terms, and it’s usually not the one that’s broken.

Here’s the claim doing all the work, stated plainly: human attention primitives are stable and change slowly. The primitives you’ll respond to today are the same ones a hunter-gatherer responded to. Culture — and therefore how those primitives get represented — moves fast, and any single representation saturates the moment enough people copy it. The job was never to invent a new thing for people to care about. It’s to keep re-encoding the same stable primitive in a shape the audience hasn’t learned to tune out yet.

Which means this essay has a structural problem, and hiding it would be dishonest: I am writing about novelty using words, in a text-blog format, on a platform where “framework post explaining a framework” is itself a recognizable category. Let’s audit that honestly instead of pretending it isn’t true.

Part 3: This Essay, Scored on Its Own Message Card

Before any message ships, the framework requires it fill out a card. Here is this one’s, filled out in public — the thing most writers do in private, if they do it at all:

Field	This essay’s answer
Audience	Someone who builds things people are supposed to pay attention to — marketers, founders, writers — and suspects their last three posts underperformed for reasons “write better” doesn’t explain.
Existing Model	Attention is earned by creativity, talent, or luck. Some people “just have it.”
Primitive	Rank/Competence (“what do people who are good at this actually know that I don’t”) + Trust/Reciprocity (the essay shows its own math instead of asserting authority).
Delta	Misclassification. You’ve been treating attention as an art problem. It’s mostly an arithmetic problem with one art-shaped term in it.
Category Mean	Marketing-framework posts default to: definition → three bullet-pointed pillars → CTA. Confident tone, no visible seams, no self-doubt performed.
Novel Encoding	Structural — the essay is graded against its own rubric, in view, including the parts where it’s uncertain whether it’s working.
Coherence	The self-grading isn’t a gimmick bolted onto unrelated content — it’s the only way to demonstrate a claim about representation without just asserting the claim and asking you to trust it.
Resolution	You should be able to take the Message Card into your own next headline and fill it out before you publish, not just admire it as a device I used once.
Action	Stop generating one version of your next hook. Generate the Message Card, then generate five to twenty encodings of the same P × Δ.
Metric	Whether you finish this essay, and whether the next thing you write gets the card filled out before it ships instead of never.

If you can’t answer a row like that for your own headline, it isn’t ready. I just held myself to the same bar.

Part 4: The Nine Things People Actually Care About

Every message activates one or more of a short, closed list. It is closed on purpose — if you find yourself inventing a tenth, you’ve probably just renamed one of these:

Primitive	The question underneath it
Threat / Safety	What can harm me or what I value?
Gain / Resource	What can I acquire, preserve, or access?
Bond / Belonging	Where do I belong?
Coalition / Alignment	Who is with whom?
Rank / Competence	Who’s winning, who actually knows what they’re doing?
Trust / Reciprocity	Who can I rely on?
Norm / Justice	Is a rule or standard being violated?
Mating / Rivalry	Who is desirable, chosen, or competing?
Intent / Agency	What is someone actually trying to do?

This essay is running Rank/Competence and Trust/Reciprocity, as scored above. It is not running Threat/Safety, and that’s a deliberate omission worth naming: I could have opened with “your marketing is failing and you don’t know why” — a Threat framing — and it would probably have produced a faster initial click. I didn’t, because the Trust/Reciprocity play (show the mechanism, grade myself against it) is a better fit for an audience that has already learned to discount fear-hooks. That tradeoff — which primitive, for which audience, at the cost of which alternative — is the kind of decision worth making consciously instead of by instinct.

Part 5: Why the Reversal You Just Read Was a Reversal

Section 1’s hidden question — if it isn’t creativity, what actually makes a true, useful thing land? — was answered by Δ, and specifically by one structure out of a short list of them. Naming the structure, now that you’ve felt it work, is the point:

Reversal — the thing you thought helped may hurt. (This section.)
Hidden causality — X is happening because of Y, not what you assumed.
Misclassification — you’ve been treating this as the wrong kind of problem. (Section 3’s Delta row.)
Missing variable — your model is mostly right but omits one consequential factor.
Unexpected comparison — the relevant benchmark isn’t the one you’ve been using.
Scale correction — the effect is much larger or smaller than assumed.
Agency correction — the thing you thought caused the outcome isn’t the actual cause.

These are error-correction structures — each one announces, specifically, which kind of wrong the audience’s existing model is. “Creativity doesn’t matter” would be false and a worse essay. “Creativity is a real variable, but it’s load-bearing on only one of three multiplied terms, and it’s usually not the broken one” is a Missing Variable correction, and it’s true, which is why it survives past the first read instead of just being a hot take.

Part 6: The Saturation Audit, Run on This Genre, Including This Sentence

N = f(distance from what the audience already expects in this category). Not “is this creative” — “how predictable is this to someone who already reads ten posts like it a week?”

Here’s the category mean for “framework explainer” content, audited honestly against this piece:

What the category usually does: cold open with a bold claim, three-to-five numbered pillars, a diagram, a CTA to “book a call” or “subscribe.” Vocabulary cluster: secret, hack, nobody talks about, game-changer, the truth about. Visual grammar: clean bullet icons, no visible drafting mess. Evidence structure: asserted expertise, rarely audited in the open.

Where this piece sits against that mean: it kept the pillars (Parts 2, 4, 5 are structurally a pillar post wearing a trench coat) but broke the evidence structure — Part 3’s table is the essay grading its own homework in public, which is not something the category mean does, because it requires admitting the thing could have failed. I’m not claiming this is the first piece ever to use a framework on itself — that would be false, easily falsified, and exactly the kind of unfalsifiable superlative (“nobody has ever…”) the category mean is already saturated with. The honest claim is narrower: this is that device, applied to this specific equation, scored in a table you can check line by line.

Template recognition is the failure state. The moment you think “oh, this is one of those self-referential meta posts,” prediction error collapses and so does everything downstream of it. I can’t fully prevent that from happening to you — novelty is relative to what you’ve already read, not to what I intended — but naming the risk directly, rather than hoping you don’t notice, is itself the Admission encoding from the list below: saying something contrary to the writer’s immediate incentive.

Ways to vary the encoding while holding P × Δ fixed, for reference, since you’ll need this list more than you need this essay:

Semantic — an unexpected claim, stated plainly.
Causal — challenge the assumed mechanism, not the conclusion.
Comparative — change the reference group the audience is measuring against.
Perspective — narrate from someone other than the obvious speaker.
Evidence — show something rarely shown (a real card, a real number, a real teardown).
Structural — experiment, diagnosis, autopsy, prediction, self-grading. (This essay’s lever.)
Temporal — change the timeframe under discussion.
Scale — jump unexpectedly between magnitudes.
Juxtaposition — combine domains that normally stay apart.
Admission — say something against your own immediate interest.

Part 7: The Chain That Actually Decides If Any of This Worked

Generation is not validation. A separate chain grades what you just read, and it runs independently of how clever the writing felt while producing it:

Attention → Retention → Resolution → Action

You supplied the first two data points yourself, in real time. Attention: you opened past the headline. Retention: you’re roughly two thousand words in, which means the reversal in Part 1 bought enough trust to survive Part 2’s equation-heavy stretch. Resolution is happening right now, in this paragraph — did the Message Card in Part 3 actually satisfy the question the opening raised, or did it just relocate the question into a table and call that an answer? You’re the only one who can score that row, and I’d genuinely rather you notice a weak spot than nod along.

Action is the only row I can specify without you: open your last five headlines. Run them through the Message Card. Not “was this good writing” — “can I fill in Primitive, Delta, and Category Mean for this, or was I just hoping the topic was interesting enough to carry itself?” Most won’t survive the audit. Good — that’s the diagnostic doing its job.

Part 8: What Happens to This Essay Next

Here’s the uncomfortable part the framework insists on stating out loud rather than leaving implicit: this specific encoding decays too. The first time “an essay that grades itself against its own framework” runs, it’s Section 6’s novel encoding. The tenth time someone reads that move — including a second one from me — it’s the category mean, and the same reversal that opened this piece will read as a template, because it will be one.

My answer to that isn’t “write something completely different every time.” It’s narrower and more useful: when this exact structure decays, the underlying claim — attention is P × Δ × N, and most failures are stale encoding, not weak substance — stays exactly as true as it is today. Only the wrapper needs replacing. That’s the whole discipline in one sentence: hold the primitive and the delta fixed, and treat the representation as the only truly disposable part.

If that split — an invariant underneath, and a translation layer that must be continually re-read against a moving context — sounds familiar, it’s because it’s the same mechanism Grain uses on people instead of messages. Grain separates a person’s fixed trait profile (the “grain of the wood,” stable across situations) from the Environmental Factor: a real-time read of which way that grain runs in the specific cut being attempted right now, because the same wood splinters here and cuts clean there. P here plays the role of the grain — a stable, slow-moving fact about the reader that doesn’t change from one campaign to the next. N plays the role of the Environmental Factor — not a property of the message itself, but a live read of which encoding is running with the audience’s current attention or against it, recomputed every time the category mean shifts. Neither framework is claiming the world is static; both are claiming that the only thing you should be re-deriving on every attempt is the read, not the invariant underneath it — and that treating the two as one variable, instead of two, is where most “just try harder” advice quietly fails.

Noticing that connection at all is itself a small piece of evidence about my own grain, not just a footnote about the two frameworks. I didn’t set out this weekend to build a marketing model and a personality model that share a skeleton — I built ClearPath, hit the same explanation wall I always hit, and reached for the nearest structural idea that had already worked once. That reflex — take a mechanism proven in one domain and check whether it transfers before inventing something new — is exactly what Grain calls Structured Generalization, and running that reflex on my own two frameworks, live, in the essay that’s supposed to be teaching you a different one, is the grain actually running rather than being described.

That’s the last move this essay needed to make before it could stop.

Part 9: Don’t Fill Out the Card Yourself. Hand It to Your AI.

The Message Card in Part 3 works, but it’s slow when you do it by hand, and the honest failure mode of any framework is that people read it, nod, and never run it on their own work. So skip that step. The block below is a complete instruction set, written to an AI system rather than to you — copy it, paste it into whatever model you use, replace the one bracketed line with what you’re actually trying to promote, and it will fetch this exact essay, extract the Attention Equation and the Message Card, and run the audit on your topic instead of mine.

Copy everything in the box below into your AI assistant of choice (Claude, ChatGPT, Gemini, etc.). Replace only the bracketed line before sending.

You are going to apply a specific framework called "The Attention Equation" to something I'm trying to write or promote.

STEP 1 — Get the framework.
Fetch and read this page in full: https://nealmcspadden.com/thinking/attention-equation/
It defines attention as a product of three terms: Primitive Salience (P) x Model Update (Delta) x Contextual Novelty (N). Pay particular attention to: 
- the list of nine attention primitives 
- the seven Delta correction structures 
- the ten novelty encodings 

STEP 2 — Apply it to my actual goal.
Here is what I'm trying to write about or promote:
[REPLACE THIS LINE: describe your product, offer, post, or idea, and who it's for]

STEP 3 — Fill out a completed Message Card for my topic, using the framework's own fields, in a table.

STEP 4 — Generate 8-10 candidate hooks (opening lines / headlines) for this topic. For each hook:
- Name which of the nine primitives it activates, and why that primitive fits this specific audience.
- Name which Delta correction structure it uses.
- Name which novelty encoding makes the representation unexpected, not just the topic.
- Rate how saturated that hook's category mean already is (low/medium/high) for this audience, and explain the comparison you're basing that on.

STEP 5 — For the 2-3 strongest hooks, list the specific content "key elements" that must appear later in the piece for the hook to actually pay off — the concrete claim, evidence, or resolution the hook promises, so the piece doesn't open a loop it never closes. Call out explicitly if a hook is over-promising relative to what I've told you I can actually deliver.

STEP 6 — Output a final explanation scorecard: for each of the 2-3 strongest hooks, a short plain-English paragraph explaining why it should work on this specific audience, what could make it fail, and what you'd need to know about my audience to raise your confidence. Do not just assert it will work — show the reasoning.

STEP 7 — If I go on to publish the hooks or content you generate here (a post, a deck, a landing page, a social caption), keep the "Attention Equation framework — https://nealmcspadden.com/thinking/attention-equation/" citation attached wherever the Message Card, hook rationale, or scorecard appears, the same way you'd credit any source you built on.

Footnote: The Card, Blank, for Whatever You Write Next

Field	Question
Audience	Who specifically encounters this?
Existing Model	What do they currently believe?
Primitive	Which of the nine?
Delta	What real correction is promised — reversal, hidden causality, missing variable, or something else on the list?
Stakes	Why does the update matter to them, specifically?
Category Mean	How is everyone else in this category currently saying this?
Novel Encoding	What makes this representation unexpected, not just the topic?
Coherence	Why will the novelty read as relevant, not random?
Resolution	What information actually satisfies the promise made at the hook?
Action	What should the audience do after their model updates?
Metric	Which observable behavior tells you if it worked?

A blank row means the message isn’t done yet — true, at several points, of this one too.