↩ ExploRational · A Catalogue of Divergent Rationality
explorational · a divergent move
Every plan is the visible tip of a pyramid of premises nobody said out loud. Dig the layers out and tag them — empirical, normative, definitional — and you usually find the disagreement living somewhere no one was arguing. Mason & Mitroff built the move for strategy rooms: trace rival plans to their rival assumptions, and debate the assumptions instead.
Richard Mason and Ian Mitroff spent the 1970s watching strategy arguments stall the same way: two camps, two plans, and a debate conducted entirely at the level of conclusions. Their diagnosis was structural. A plan is the last line of an argument nobody wrote down. Beneath "raise the price" or "close the plant" sits a stack of premises — about customers, regulators, rivals, the meaning of words like growth — that were never stated, so never examined, so never actually disputed. Their method, strategic assumption surfacing, made each camp excavate the stack under its own plan and then set the debate where the stacks differed. In their canonical case, a drug company at war over pricing found the rival plans standing on opposite pictures of the prescribing physician — price-blind in one, price-driven in the other. The loud argument was about price. The fork was a claim about doctors, and nobody had been arguing it.
The excavation has a grammar. One standing question — what must be true for this to hold? — is put to the plan, then to each answer, then to each answer's answers, and every reply becomes a block in the next layer down. Each block takes a tag, because premises fail differently: an empirical block could be checked, a normative block can only be chosen, a definitional block dissolves once the word is pinned. The classic finding is a mix shift: plans argued in empirical vocabulary go normative two or three layers down. The surface talks numbers; the foundation holds values. The bench below computes your pyramid's version of that, live.
the signature instrument
State the plan at the apex. Put the standing question to any block with ⌄ dig; each answer becomes a block in the layer beneath. Tag every block and record your own confidence in it. The pyramid, the layer counts, the tag mix, and the load arithmetic all compute live from what you enter. Nothing is graded — the bench counts, and the judgments stay yours.
Return to the office — a worked excavation
click any block in the pyramid to mark it as the crux — the block the dispute is actually about.
the standing question — "what must be true for this to hold?" · ⌄ dig puts it to any block, and your answer becomes a block in the layer beneath. tags and confidence are yours to assign; ◎ marks the crux.
Pick the single block where the real disagreement lives, and the bench restates the dispute as a dispute about that block — in your own recorded words.
the pyramid under the plan
An assumption, in Mason and Mitroff's sense, is unstated rather than hidden: nobody buried it, it simply never needed saying among people who all took it for granted — until two plans collided and the taking-for-granted diverged. The pyramid was always there, holding the plan up. Excavation adds nothing to it; excavation makes it visible, and visibility is what converts a premise from load-bearing background into something a room can actually inspect.
The grammar keeps the dig honest. Every block must answer the standing question about the block above it, so every layer is strictly load-bearing — no context, no background, only what must be true. And every block takes a tag, because the three kinds fail in three ways and are settled by three different kinds of work. Run in a group, the move goes camp by camp: each side digs the pyramid under its own plan, then the pyramids are compared block against block. Agreement usually reaches further down than anyone expected, and the genuine forks turn out to be few, deep, and oddly specific.
what to try
Read the worked pyramid's tag mix. Layer 1 is pure empirical — the plan argues like a lab report. By layer 3 the mix is majority normative: "work that can't be seen being done counts for less" is a value wearing a productivity claim as a coat. Now excavate a plan you are for, tagging honestly, and watch your own mix shift. Empirical paint over normative timber is the classic finding, and the readout counts your version of it.
Check the low-confidence load-bearers. The list ranks every block you tagged shaky by how many blocks stand on it — counted straight off the structure you drew. In the preset the champions carry three each. That is where to aim your checking first: if a low-confidence block three layers deep gives way, everything in its chain needs a new leg by Monday.
Run the relocation. Pick the block where the disagreement actually lives and read the restated dispute out loud. It will sound like a different argument — smaller, stranger, more answerable. It also ends the dig: the pyramid's job was to find that block, and it has.
where disagreement actually lives
At the apex, every underlying difference is compressed into two words: for and against. Two people can hold the same apex for incompatible reasons, or oppose each other while sharing nine premises out of ten — and apex-level debate cannot tell those situations apart, which is why it can run for hours and move no one. Three layers down, the same dispute is usually one block wide: checkable, choosable, or definable.
The tag tells you what settling it would even mean. An empirical fork calls for evidence, and the argument becomes a search. A normative fork calls for a decision about values — and shelling a chosen value with measurements is a category error, dimensional analysis for arguments. A definitional fork needs the word fixed before anything can be true of it. The relocation template restates your dispute as a dispute about one tagged block, in your own recorded words, with a count of what stands on it. That sentence is the deliverable. Everything above it was acoustics.
the honest bench
Every figure on the bench is structural. Blocks per layer, the tag mix, the empirical share, the load count — all of it is arithmetic over the diagram you drew. "Carries 3" means exactly that three blocks sit in the chain above that block, no more. The tags and the confidence grades are your own judgments, recorded and counted, never second-guessed.
What the bench cannot do is worth saying just as plainly. It cannot tell a true premise from a false one. It cannot notice the block you failed to dig — a pyramid is a map of your articulation, and its blind spots are your blind spots. And it does not score the quality of your plan, because no honest instrument could: a tidy pyramid over false blocks is still a tidy pyramid. The bench makes the stack visible. Checking the blocks is work it refuses to fake.
the move ↔ the bench
| The move | The bench |
|---|---|
| the apex | the plan exactly as argued — one line, standing in for everything underneath it. |
| a block | one thing that must be true for the block above it to hold. Nothing else qualifies for the layer. |
| an empirical tag | checkable: some observation could settle it, which makes the fork a search. |
| a normative tag | chosen: a value is being asserted, and no measurement will settle it. |
| a definitional tag | assumed in the words: pin the meaning and the block may dissolve entirely. |
| the low-confidence load-bearer | the shaky block holding the most weight — first to test, and the best place to relocate a debate. |
how this opening fails
risk · endless deferral
Every block has blocks beneath it — dig long enough and you reach epistemology, and the plan still has to ship. The ghost layer under your pyramid is honest: there is always another layer, which means excavation can become the most respectable procrastination in the room, everyone surfacing and no one deciding. The bench builds the exit in. The relocation step ends the dig the moment you can name the one block the dispute is actually about: you were excavating to find that block, and no deeper. Once it is named, the discipline is declining to visit the layer below it.
risk · weaponized surfacing
Demanding that opponents surface their assumptions while your own stay buried is interrogation passed off as methodology. It feels rigorous and is merely aggressive — every pyramid looks embarrassing mid-excavation, because every plan, including good ones, including yours, stands on unproven blocks. Mason and Mitroff's format made each camp dig under its own plan first, and the bench keeps that discipline: it holds one pyramid at a time, and the first one should be yours. Surfaced symmetrically, the move is inquiry. Surfaced one-way, it is siege equipment.
The argument is at the apex; the fork is three layers down.
can you use it?
RECOGNITION — Which is surfacing the assumptions? A: naming the one belief that would change your mind. B: digging out every premise a claim stands on — empirical, normative, definitional — and tagging each layer. C: stating your conclusion more forcefully.
B. It maps the whole pyramid of unstated premises and labels their kinds, so you can see where disagreement actually lives. A is finding the crux (one premise); C adds no premises.
THE NEAREST NEIGHBOR — Finding the crux also probes beneath a claim. What separates surfacing the assumptions from it?
Breadth. Surfacing maps all the premises and their types; finding the crux isolates the single one whose flip would change your mind. One is a full excavation, the other a targeted dig.
PRODUCTION — Take a plan you support ('we should go remote-first'). List three premises it rests on, and tag each as empirical, normative, or definitional.
A version: 'the work doesn't need co-location' (empirical); 'autonomy matters more than serendipity' (normative); 'remote-first means default-remote, not no-office' (definitional). Yours works if the three are different kinds — three empirical claims means you haven't dug past one layer.