The acceptance gate, in one calculation

Companion page to the manuscript. It takes about five minutes and there is a slider halfway down.

A surgeon, one day

At nine in the morning she is in an operating theatre. A junior colleague tells her that her incision is wrong. She stops, looks, says “you are right,” and changes what she is doing.

Six hours later her husband says almost the same words: she is being unfair to their daughter. This time she does not stop. She explains why he is mistaken, and then why he has always been mistaken.

Her intelligence did not change between those two scenes, and neither did her capacity for effort. What changed was whether the correction was allowed in.

The usual explanations do not fit. Stubbornness would have shown up in the morning too. Slowness is not the issue. And it is not two different people.

Three conditions

The proposal is that acting against your own strongest habit requires three things at once, and that each of them caps the result on its own. They are not stages you pass through in order.

All three building blocks are established and have been studied separately. The claim is about the wiring, not about the parts.

The obvious objection

It is this: the middle condition is just another name for intelligence, and people who can do more correct themselves more often. If that were right, raising someone's ability should shrink the gap between morning and afternoon.

The display below lets you try it. The slider changes one thing only, how far ahead she can project consequences, and that quantity enters the third condition. It does not touch the first two.

What that shows, and what it does not

Her ability works in both scenes and by the same factor. That is not an approximation: the slider never touches acceptance, so it cancels when you divide one autonomy by the other.

She still ends up high on one side and near the floor on the other. Nothing on the right is broken. That number rose too, by the same factor, but it rose under a ceiling.

The testable consequence is about that absolute gain. A very able person whose self-worth is staked on some domain should collapse there rather than spend ability as compensation. An account in which the three conditions add up predicts the opposite.

Deciding between the two accounts needs the rival account fitted to the same data, which is a study rather than a slider. The repository does not contain that rival yet.

The calculation

The core is a product of three numbers, so that none of them can stand in for another:

from math import exp

def sigmoid(x):
    return 1 / (1 + exp(-x))

def autonomy(detection, acceptance, override):
    """How far a person can act against their strongest schema."""
    return detection * acceptance * override

The case the manuscript works through is an able man with an excellent error detector and formidable modelling ability, meeting a criticism that lands exactly where his self-worth is staked:

>>> round(autonomy(sigmoid(2), sigmoid(-3), sigmoid(2)), 4)
0.0368
# Both faculties removed, the gate still shut:
>>> round(autonomy(0.5, sigmoid(-3), 0.5), 4)
0.0119

# His two faculties tripled the result. It is still three hundredths.

These come from the full logistic values. Rounding the inputs first gives a different answer, which is one way to get a worked example wrong. Units are arbitrary and illustrative throughout: they show an ordering, not a measurement, and none of them should be quoted as a measured value. The same holds for the numbers in the display above.

What would refute it

The manuscript names three forms the test can take and three patterns that would sink the model. The one it leans on: where acceptance is low, the effect of forecasting depth should fall below what a study of that size could detect. If it merely shrinks, the conditions are trading off and the architecture is wrong.

Before any of that, something else has to hold, and it is easy to overlook. The three conditions must be measurable separately, without using the outcome to do it. If acceptance is recognised by the fact that someone changed their behaviour, nothing has been explained, only renamed. That is a question about measurement, and the manuscript treats it in an appendix of its own.

Where this comes from

This page accompanies The Acceptance Gate: Three Necessary Conditions for Acting Deliberately on What One’s Schemata Propose by Nevzat Subasioglu, independent researcher, Switzerland. The manuscript carries the derivation, the demarcation against existing models, and a proposal for how the matter could be settled experimentally.