You now own a latticework. You can name a dozen models, you know the three ways they combine, and you’ve watched a case study bloom under several lenses at once. So here’s the uncomfortable question that ends this course: in the actual moment of decision — tired, rushed, emotionally invested, with a number due by Friday — will you use any of it?
Probably not. Not because you’re weak, but because that’s exactly when the lollapalooza effect you studied earlier turns its guns on you. Several biases — commitment, social proof, deprivation-reaction, authority — stack up and quietly capture your judgment. A latticework that only lives in your head is a tool you’ll conveniently forget to reach for at the one moment it matters.
The fix isn’t more willpower. It’s paper.
Before you read — take a guess
You've built a rich mental latticework. Why might a written checklist still beat 'just remembering to think it through' in the moment of a real decision?
Turn the lattice into a written checklist
Here’s the analogy pilots and surgeons already live by. A veteran surgeon knows how to wash their hands and count the sponges — and still, hospitals cut infection and retained-instrument rates by handing them a written checklist to run out loud, every time. Not because surgeons forgot medicine. Because a checklist defeats the specific failure of knowing something and skipping it anyway under pressure. Atul Gawande built a career on this: expertise fails not from ignorance but from ineptitude — knowing the right thing and not doing it.
Definition. A latticework checklist is a short, fixed, written list of model-prompts you run deliberately — before you commit — so that your best models get deployed on purpose instead of remembered by luck. It converts “I have a latticework” into “I used my latticework, here, today, on this.”
Steal this one. Read it top to bottom on any decision worth more than an afternoon:
| Prompt | The model behind it | What it forces you to see |
|---|---|---|
| Incentives — who profits from each outcome? | Incentive-caused bias | The behaviour you’re puzzled by usually pays someone. |
| Base rate — what usually happens to cases like this? | Base rates / outside view | Your specific story is probably less special than it feels. |
| Second-order — and then what? | Second-order thinking | The reaction to your action, the reaction to that. |
| Feedback — what reinforces itself here? | Feedback loops | Small edges compound; small rot compounds too. |
| Players — who else is deciding, and how do they respond to me? | Game theory | You’re not choosing against nature; you’re choosing against people who choose back. |
| Inversion — how could this fail? | Inversion | It’s often easier to avoid stupidity than to be brilliant. |
| The gap — what model am I not using? | Circle of competence | The lens you’re blind to is the one that gets you. |
Worked example — the checklist catches what enthusiasm missed
You’re about to sign a two-year lease on a bigger office because the team is growing and morale wants it. Feels obviously right. Run the list.
- Incentives: the broker earns a percentage of a longer, bigger lease — their advice is not neutral. Noted.
- Base rate: most early-stage teams that “grew into” fixed space regretted the fixed cost when growth wobbled. The base rate on locked long leases is unkind.
- Second-order: signing raises your monthly burn → shortens your runway → and then forces a fundraise on a worse timeline than you’d choose freely.
- Feedback: a bigger fixed cost is a self-reinforcing pressure — it doesn’t shrink when revenue dips.
- Players: the landlord holds the pen; if you’re desperate to close by month-end, they respond by holding firm on terms.
- Inversion: how does this fail? Growth stalls for two quarters and you’re paying for empty desks. Plausible.
- The gap: you were running one model — “morale / we deserve nice things” — and none of the others.
Same facts, thirty seconds of checklist, and the decision flips from “obviously yes” to “negotiate a shorter term with an expansion option.” The models were always in your head. The list is what got them out.
Keep the checklist short and fixed. Seven prompts you actually run beat forty you admire and skip. The value is in the ritual, not the length.
The pre-mortem — inversion, systematized
There’s a second tool that belongs on the same page: the pre-mortem, from decision researcher Gary Klein. A post-mortem asks why a dead project died. A pre-mortem does it in advance: gather the people, and say — “It is a year from now. This decision has already failed, spectacularly. Everyone knows it. Write down why.”
This is inversion wearing a lab coat. By assuming failure as a given, it gives people permission to voice the doubts that “will this work?” politely suppresses. Nobody wants to be the naysayer before the vote; everybody’s happy to explain a failure that’s already stipulated. Klein’s research found this “prospective hindsight” makes people markedly better at naming concrete reasons a plan could break.
Worked example. Before launching a new product, the room runs a pre-mortem. “It’s next year and the launch bombed. Why?” Out come: we shipped before the onboarding was ready (feedback loop — bad first impressions compound), a competitor cut price the week we launched (players responding), we counted on a channel that throttled us (single point of failure). None of these surfaced in the optimistic “why this will work” meeting. Each is now a risk you can pre-empt. The pre-mortem is how you run inversion across the whole lattice at once, with a group, before the money’s spent.
When to use it
Use the checklist on any decision that is consequential and hard to reverse — a hire, a big spend, a strategy commitment, a lease. Skip it on cheap, reversible calls; running a seven-lens audit to pick a lunch spot is its own failure mode (we’ll get there). Use the pre-mortem specifically when a group is converging too fast toward a plan everyone already likes — its whole job is to unlock the dissent that consensus is smothering.
Now the honest part. A latticework is a tool, and tools cut the user. Here are the four ways the lattice quietly lies — even to people who’ve read every earlier lesson. Learn to spot each one on yourself.
Pitfall 1 — Forcing a model where it doesn’t fit
Definition. This is the man-with-a-hammer problem, reborn inside the latticework. You didn’t stop loving your favourite model just because you collected others. So now you reach for it because it’s yours and it’s satisfying — not because the situation’s actual assumptions call for it. It’s the hammer, dressed up in the respectability of “using a mental model.”
Worked example. An engineer adores compounding. Faced with a hiring decision, they model the candidate’s “compounding skill growth” over five years and project a superstar. But compounding’s home assumptions — a stable base, uninterrupted reinvestment, no shocks — don’t hold for a person’s career, which has jumps, plateaus, burnout, and moves. The candidate leaves in eighteen months and the elegant curve was fiction. A model applied outside its domain doesn’t just add nothing; it manufactures false precision.
Antidote: before you apply a model, ask do its home assumptions actually hold here? If the terrain violates the model’s premises, put the hammer down.
Pitfall 2 — Double-counting the same effect under two names
Definition. The lattice’s power comes from independent confirmation — several different forces pointing the same way (recall the STACK). But if two of your “models” are really one effect wearing two labels, you’re counting it twice, and your confidence inflates on air. Two hats, one head.
Worked example. You’re bullish on a stock. “Social proof says everyone’s buying it,” you note. “And herd behaviour says the crowd’s piling in.” Two lenses agreeing! Except social proof and herd behaviour are the same phenomenon — people copying people — described in two vocabularies. You haven’t found two independent reasons; you’ve found one reason and flattered yourself with a second name. Real independent confirmation would be, say, insiders are also buying (incentives) plus the base rate for this setup is good — genuinely different causes.
The tell: your two “confirmations” would always move together. If model A is true, is model B automatically true? Then they’re one force, and it should count once. Independent lenses can disagree; that’s what makes their agreement worth anything.
Antidote: for any two lenses that agree, ask are these really independent — or one effect wearing two hats?
Pitfall 3 — Confirmation-shopping for the flattering model
Definition. You already know the answer you want. So you stroll down the latticework like a supermarket aisle and pick up whichever model happens to endorse it, leaving the rest on the shelf. This is confirmation bias with a graduate degree — now armed with respectable-sounding frameworks to launder a predetermined conclusion. The models make you more dangerous, because your bad call now sounds rigorous.
Worked example. You want to keep a struggling project alive (you started it; you’re attached). You reach for compounding — “it just needs more time to pay off” — and circle of competence — “we’re the experts, we know something the doubters don’t.” Both flatter the answer you already had. You conveniently skip inversion (“how is this failing right now?”), base rates (“most projects this far behind don’t recover”), and sunk cost (“the money already spent is gone either way”). You didn’t reason to a conclusion; you shopped for a justification.
Antidote: name the answer you’re hoping for out loud, in writing, first — then run the lattice honestly, or better, hand it to someone told to red-team it. If your models only ever agree with your wishes, they’re decoration, not thinking.
Pitfall 4 — Paralysis-by-lattice
Definition. More models is not better past the point of decision. Every situation can be lensed forever; there’s always another angle, another framework, another “but have you considered…” You can analyse a choice straight into the grave — the decision-making version of a browser with two hundred tabs open and nothing read.
Worked example. Two candidates, roughly equal, and you’ve run incentives, base rates, game theory, second-order effects, personality models, and a pre-mortem — and now you’re building a weighted twelve-factor spreadsheet to break a tie that a coin could break. Meanwhile the stronger candidate, waiting a week for your over-lensed verdict, takes another offer. The analysis didn’t improve the decision; it destroyed it by making you too slow to have one.
Here’s the thing people forget about Charlie Munger, the patron saint of this whole course: he didn’t run a hundred models on everything. He ran a few big ones, hard, and then decided. The latticework is a small set of load-bearing ideas used with conviction — not an infinite checklist that guarantees you never have to be brave enough to choose.
Antidote: pick a few decisive models for the situation, run them honestly — then decide. Past the point where more analysis would change your action, more analysis is just fear in a lab coat.
There’s a symmetry worth noticing. Pitfalls 1–3 are the lattice being used too loosely (wrong tool, doubled tool, cherry-picked tool). Pitfall 4 is the lattice used too much. Good latticework thinking lives in the narrow band between sloppiness and paralysis — enough models to see clearly, few enough to still move.
Match each pitfall to its antidote
Pair each way the lattice lies with the one-line fix that defuses it.
Sort the moves: sound use vs. misuse
Drag each into the right bucket.
- I'm on my ninth framework for a reversible $40 decision and still haven't decided.
- We ran a pre-mortem and three concrete failure modes surfaced that the optimistic meeting never raised.
- I modelled a person's career with compounding even though careers have shocks and plateaus.
- Before signing, I ran a fixed seven-prompt checklist and it flagged an incentive I'd missed.
- For any two lenses that agreed, I checked whether one being true forces the other — and dropped the duplicate.
- The two candidates are close, so I ran my few decisive lenses and then just chose.
- Social proof AND herd behaviour both say buy, so that's two independent reasons.
- I wanted to keep the project, so I grabbed the two models that agreed and skipped inversion and base rates.
Recap quiz
The checklist and its failure modes
Why does a written checklist beat 'remembering to think it through' at the moment of decision?
Check your answer to continue.
Where this goes next
That’s the whole latticework: the hammer problem, the idea of the lattice, the three ways models combine, a worked case, and now the checklist habit plus the four honest limits. You have the tool and you know how it cuts.
One thing remains — proving it to yourself. The Final Exam is a single graded run across everything this course taught: spotting man-with-a-hammer thinking, stacking independent models, choosing lenses, and catching the lattice mid-lie. One question at a time, no going back. Bring the checklist in your head — then go decide.