the open model · judgement calls
Every number in here that is a judgement
There is no validated instrument for domestic cognitive labour in English or French. So some of these numbers are findings, some are directions taken from findings, and some are choices. This page says which is which.
The distinction this page is built on
A weight can have three kinds of backing, and they are not interchangeable:
- Measured — a number that comes out of a study, with a sample size attached.
- A direction with a chosen magnitude — the research says large, or near zero, and somebody had to pick how large. The direction is evidence; the magnitude is a judgement.
- A choice — nothing in the literature bears on it; it was picked because something had to be, and it is defended on product grounds.
Most of the constants below are in the middle category, which is the honest place for them to be and an uncomfortable place to leave them. All of them get calibrated against real households before any of this is called finished.
The constants
| Value | What it does | What kind of number it is | |
|---|---|---|---|
| K | 1.0 | Weights the cognitive half against the execution half | A choice, and the most consequential one. See below |
| Cognitive capacity coefficient | 0.20 | How much long paid hours reduce what you are expected to carry | Direction measured (near zero), magnitude chosen. One US survey, no European replication |
| Execution capacity coefficient | 0.70 | How much long paid hours reduce what you are expected to do | Direction strongly measured (7,535 couples, ten countries); the coefficient itself is a guess |
| The 28–72% clamp | 0.28 / 0.72 | Stops extreme hours producing you owe nothing | A choice, with external cover from what people rate as fair — the numbers themselves are not measured |
| Intensity ceiling | 1.55× | The most that effort and disruption can mark up a task's execution weight | A choice, deliberately the smaller of two candidates: between two unaudited multipliers, prefer the one with less argument surface |
| The vigilance exponent | ^1.5 | Makes high-vigilance tasks pull away from the pack rather than sitting in a line | A choice about shape |
| The frequency term | ln(1 + f) | Something carried constantly costs more than something carried once a year, with steeply diminishing returns | A choice about shape |
| Gap bands | 5 / 14 / 24 | Where balanced becomes tilted becomes significantly tilted | A choice, to be tuned against real households |
| The tolerance band | 5 percentage points, per axis | How close to its reference a split can sit before the app stops offering moves on it — the reference being what your hours suggest, or before hours, for the carrying, the even split | A guess, to be recalibrated against real households |
| The minimum effect | 3 percentage points | The smallest change to a printed split that a set of moves must make to be shown at all | A guess from sampling: halving a list at random moved a share by up to 2.5 percentage points. The first value to revisit once real list lengths are known |
| Library minutes | per task | The marginal time a task costs that its neighbours don't already claim | Estimates, scored against anchors and rank-checked — not measurements of your household |
K, in detail, because it explains the whole product
K sets how much of the headline is cognition rather than execution. It is the single most consequential number in the model and nothing in the literature determines it.
It was 1.5 as a placeholder, before a library existed to test it against. Measuring the finished library showed 1.5 put 72–82% of every household's headline into the cognitive half — stable across household types, because both halves scale with how many rows a household has. Nobody chose that. It fell out of a placeholder.
At K = 1.0 the cognitive share is 63–75%, centring on about 65%.
Why 1.0 and not something else. The evidence gives a direction, not a magnitude: cognitive labour predicts depression, stress, burnout and relationship quality where physical labour predicts none of them but relationship quality; and employment cuts mothers' physical labour by 17–22% and their cognitive labour by zero. Cognition has to dominate. Against that, three product arguments for not letting it dominate completely:
- The sceptic has to be able to move the needle. At a 75% cognitive share, a partner doing substantial physical work watches the number barely respond when he does more — and a partner who concludes he cannot affect the number stops engaging, which is the failure that kills this kind of app.
- It hollows out the capacity model. The whole good-faith concession about long hours would be adjusting a number four-fifths untouched by it.
- 65% is still an emphatic claim. The mental load is nearly two thirds of the reading. Nobody will mistake this for a chore-splitter.
A reader who understands why K is 1.0 understands this product.
Three conflicts we have not resolved
Places where the research points one way and a decision has gone another. They are written down as conflicts rather than quietly settled, and they stay here until something settles them properly.
- Independent scoring should probably come earlier than it does. The perception gap — two people scoring the same household separately and comparing — is the most reliably measured phenomenon in this field and the least accusatory output available, because it indicts neither person. The app asks the two of you to go through the list together, on one shared device, and keeps one picture you agree on. That is not the same thing, and we know it.
- A floor for rare, high-stakes deadlines. One reading of the research asks for a cognitive floor under annual hard-deadline tasks — the tax return, the passport renewal — on the grounds that the frequency term crushes them. The arithmetic argues back that at an annual frequency the term is 1.019, so a vigilance-5 annual task already earns about 98% of K and needs no floor. Both are on the record; the conflict was deliberately not used to settle the wider question it touches.
- Simultaneity is not measured at all. Mothers multitask roughly ten hours a week more than fathers, and it is associated with distress; nothing on the market measures it, and neither do we, because it proved too situational to score per task. Recorded as a conflict, not as a plan.
And the one nobody has an answer to
No study anywhere tests whether measuring a division of labour changes it. Not ours, not anyone's. The best-known product in this category has a single published evaluation with no comparison group.
That is this product's central untested assumption, and it is the reason nothing on this site claims an outcome. Champignon says what it measures. It does not say what it achieves, and it will not until there is a controlled result to point at.