↩ ExploRational · A Catalogue of Divergent Rationality
explorational · a divergent move
You cannot weigh what you cannot yet state. What changes when you are made to build the strongest case for each side in turn — not as fairness theater, but as the training discipline of the Roman rhetoric schools, the drill Cicero credited with letting him see what was coming in court?
Protagoras taught, notoriously, that on every matter there are two logoi opposed to each other. An exercise book from around 400 BC survives under that banner — the dissoi logoi, "double arguments": paired cases, for and against, on the same questions. The Greek claim hardened into a Roman drill — declamation in utramque partem, "on both sides": the same student, the same case, argued first one way and then the other, whole each time. Not a debate between two students; one mind, made to build each case as if it meant it.
Cicero trained on the drill and kept it for life. Arguing every question on both sides, he said, let an advocate see what was coming in court — the opponent's best argument cannot surprise a man who has already built it himself. The drill ends in a verdict — the equality of the sides was never the claim. The claim is narrower and harder: you cannot weigh what you cannot yet state. The bench below runs the drill on any proposition you bring.
the signature instrument
Enter a proposition. Fill the FOR slots first; the AGAINST pane unlocks at 60 words of FOR — a plain count, shown as it fills, open for good once opened. Then build the AGAINST as if you were paid to win with it. The comparison beneath computes word counts, slots answered, and shared vocabulary — and nothing here judges which side won.
In utramque partem
one proposition · two cases, FOR then AGAINST · four slots a side
shorter ÷ longer word count. equal length is not equal merit — it only shows both sides got a full turn.
a slot counts as answered at 5+ words. five bad words still cover a slot.
needs words on both sides.
distinct content words shared ÷ distinct content words used (jaccard). a proxy for whether the cases met — not a score.
No readout grades an argument or picks a winner. The weighing is yours, and it comes after both cases stand.
the discipline, not the theater
"Of course there are arguments on the other side" costs one sentence and changes nothing — an empty chair, gestured at, never seated. Stating the other case is a different act with different costs: you must find its central claim, not the caricature easiest to hold; its strongest ground, the one you would least like to face; the rebuttal it would make to yours. These obligations are generative: grounds you could not have listed in advance appear only while you build the case that needs them. Construction produces knowledge that acknowledgment never touches.
The Roman form adds the sequence, and the sequence is most of the discipline. Objections handled as interruptions never get the floor: your side frames, the other side heckles. In utramque partem gives each side the whole floor in turn — one case finished before the other begins — which is why the bench locks the AGAINST pane until the FOR has substance. You can sit at it alone; the form is still dialogic — two speakers taking turns in one head.
what to try
Enter a proposition you actually hold and write the FOR case — it comes fast. Then watch the AGAINST slots resist: the central claim arrives as a caricature, and strongest ground is where the writing stalls. The stall is the exercise — the distance between knowing objections exist and being able to state one.
When both cases stand, check the shared-vocabulary readout. Low overlap means your two cases never met — the AGAINST argued some adjacent proposition instead of engaging yours. Revise it until it works with the FOR case's own matter: its evidence, its terms, its ground.
Load the exemplar and read the two anticipated-rebuttal slots against each other — each case names the other's best argument and concedes nothing. Its readouts show a finished drill structurally: eight slots answered, near-even counts, enough shared vocabulary that the cases plainly met. Structure is all the numbers can see.
weighing needs stating
A verdict reached between a stated case and an unstated one is a ratification, not a weighing. The unstated side is represented by a summary you wrote for it — shorter than it would have written, and weaker; one pan of the scale loaded, the other holding a note that says objections exist. On anything worth deciding an opposed case can be built — Protagoras's point was about matters, not truth — and until it is, your decision has never faced it.
Stating is prior to weighing the way listing is prior to choosing. The drill defers the verdict; it does not dissolve it. Once both cases stand, weigh — that is the step the training exists to improve. What changes is what the verdict must now survive: the best thing the other side would have said, said already, by you, in its voice.
the honest bench
Every number on the bench is listed here. The word counts are whitespace-split tokens containing a letter or digit. The lock opens at 60 FOR words — an arbitrary, disclosed threshold that enforces sequence, nothing else; once open it stays open; clearing the bench re-locks it. A slot counts as answered at five words or more — presence, not adequacy.
The overlap readout reduces each case to its distinct content words — four letters or longer, lowercased, minus a fixed stoplist of function words (listed in the page source), with no stemming, so morning and mornings count as different words — and reports shared ÷ total distinct: a Jaccard index. It is a crude proxy for whether the AGAINST engages the FOR's matter or talks past it: it cannot tell engagement from quotation, and it cannot read merit. The bench structures practice; it does not grade arguments. If your browser allows it, drafts persist locally; nothing leaves the page.
the move ↔ the bench
| The move | The bench |
|---|---|
| the proposition | the matter at issue — one arguable sentence both cases must address, fixed before either side speaks. |
| the FOR pane | your side, made explicit — claim and grounds put on the record instead of resident in your head. |
| the AGAINST pane | the opposing case you must build in full — the opposing argument as a whole construction with its own best ground. |
| slot coverage | whether each case is complete — claim, ground, rebuttal, and a test it could fail; presence, never adequacy. |
| vocabulary overlap | whether the cases engage or talk past each other — two arguments about the same matter share its words. |
| the sequence lock | the discipline of finishing one side before starting the other — in utramque partem argues in turns, not in interruptions. |
how this opening fails
risk · flattening
The exercise proves the sides can be stated; it proves nothing about their weight. Read a finished bench as evidence that "reasonable people disagree" and the drill has curdled into both-sides-ism — journalistic false balance is this move with the final step deleted. The discipline is state both, then weigh; suspending the verdict forever is the catalogue's flattening — every view held equally, none engaged. The bench refuses to compute a winner because weighing the arguments requires a substantive judgment. A verdict may still exist.
risk · token opposition
The cheap way through is a limp AGAINST — a thin central claim, grounds nobody would pay for, a rebuttal your own case swats aside. The motions get performed; the conclusion never comes under fire. The readouts make the tokenism visible — a lopsided balance bar, uncovered slots, an overlap so low the two cases never met — but visible is all they can make it. A word count cannot make you want the other side to win for an hour. That want is the exercise, and no instrument can perform it for you.
State the case you fear before you weigh the case you favor.
can you use it?
RECOGNITION — Which is the drill? A: Two colleagues are assigned opposite sides of the vendor question. B: Your memo closes with “of course there are counterarguments.” C: You write the whole case for the switch, then the whole case against, each with its strongest ground.
C. In utramque partem is one mind building each case whole, in turn. A splits the drill between two heads; B gestures at an empty chair — acknowledging, never stating.
THE NEAREST NEIGHBOR — Steelmanning also builds the opposition’s best case. What does this drill add?
The whole floor, twice, then a verdict: each side gets a complete case — claim, ground, rebuttal, evidence — in sequence, and the weighing follows. State both, then judge.
PRODUCTION — Your team is split on “ship the redesign now,” and you favor shipping. Write the AGAINST case’s strongest ground — the one you would least like to face — before looking.
“The redesign has never carried a peak-season load; shipping now means discovering its limits in front of customers, when a rollback costs triple.” Yours works if it engages your own case’s matter — its evidence, its terms — and you would genuinely rather not face it.