Identity–Action Gap · working notes
You can sincerely believe “I am someone who writes” and still not write today. This page is a hand-built machine for thinking about that space between who we say we are and what we actually do.
Metaphor The ink line is a hill. Its height at any point is the model’s probability of action. The red figure walks as far as your intention strength carries it, and you can see how steep the climb is from there.
The model treats “I am someone who…” as a source of push, and treats the gap as everything that sits between that push and a first visible act. It assumes behaviour is not a simple sum of willpower: there is a tipping region where small changes in conditions produce large changes in whether anything happens.
P(action) = 1 / (1 + e−k(I − θ))
θ = 2 + 0.45F + 0.40A + 0.45E − 0.50S
k = 0.55 + 0.09B
I is intention strength. θ is where the hill is halfway climbed. k is how sharply the hill rises. The weights are hand-picked so the sliders feel meaningful. They were not estimated from any data.
In this sketch feedback does not move the hill. It makes the rise sharper. With quick feedback, a bit more intention past the threshold turns into action quickly. With slow feedback, even strong intention climbs a long, shallow slope where progress is hard to feel.
It is not a diagnostic instrument, not a clinical tool, and not derived from a dataset. Nothing you enter leaves your browser. Sliders and the form are saved only in this browser’s local storage. A shared link carries slider positions in the URL fragment, and nothing else.
Five short steps. The result is a one-page card you can print or save as PDF. Your answers are kept in this browser so you can come back to them.
The Gap Engine · diagnostic card
Exploratory model, not validated. The number is an illustration of the idea, not a prediction. identityaction.sahiltayal.com
Speculation Each is written so that a specific result would count against it. None has been tested. They are offered as invitations to test, not as findings.
Among people with equal stated intention, the probability of starting rises non-linearly as friction falls: a fixed reduction in steps produces a larger change near the middle of the range than at either end.
Falsified if the relationship between number of required steps and start rate is linear, or its largest change is at an extreme.
Strengthening identity statements (“I am a writer”) without changing friction or support produces a smaller change in action than an equal-effort reduction in friction.
Falsified if identity-only interventions match or exceed friction-reduction interventions on observed behaviour in a pre-registered comparison.
Faster feedback raises action rates mainly for people whose intention is already near threshold, and has little effect on people far below it.
Falsified if the effect of faster feedback is the same size across low, middle and high baseline-intention groups.
Self-reported dread is higher in the minutes before a first action than during it, and falls after the first visible act.
Falsified if momentary dread ratings show no drop after initiation, or peak during the task rather than before it.
A fixed amount of structural support (a scheduled slot, a partner, a prepared space) can compensate for lower intention: people with lower intention and high support act at rates comparable to people with higher intention and low support.
Falsified if support raises action only for already-high-intention participants and never offsets low intention.
People can name, in advance, the moment they usually abandon an intention, and that named moment matches where abandonment is later observed more often than chance.
Falsified if pre-named abandonment points do not match observed ones above a chance baseline.
The Identity–Action Gap model sits next to established research on intentions and behaviour. It borrows questions from that work but has not earned its conclusions. The sources page lists adjacent literature that shaped the questions here. Listing a work does not mean its authors endorse, reviewed or know about this model.