How the Resilience Score is calculated
Model v1.2 · O*NET 30.3
This is an exposure index, not a prediction. It measures how much of the work you describe can be performed by systems available now or plausibly soon. It does not say what will happen to your job, and it has no opinion about dates.
The four components
Task automatability 40%
How your week divides across activities, weighted by how automatable each one is. The only part not based on self-report, which is why it carries the most.
Human moat 25%
Physical presence, licensing, whether clients come to you personally, and how well defined your problems are.
Accountability weight 20%
Who answers when the output is wrong. Roles that require a signature change shape rather than disappearing.
Adaptation velocity 15%
What you have already done about it. Weighted lowest on purpose — it should move your score, not decide it.
Where the activity data comes from
Occupations and their activities come from the O*NET database. O*NET publishes no exposure figure, so that number is ours: every activity is assigned a prior from the 41 activity groups it belongs to — the level at which the published research operates — then adjusted by signals in its description, capped so that wording cannot overturn the judgement about the kind of work. Nothing scores 0 or 100, because neither extreme is honest.
How the judgement stays current
The exposure judgement is re-examined every quarter against evidence of what AI systems demonstrably shipped — collected continuously, from primary sources, and reviewed by a person. Every revision ships as a new, numbered model version (this page is generated from the live one, above). A score you were shown is never rewritten: the new version is computed beside it, both remain readable, and a subscriber whose placement moves by five percentile points or more receives one plain note stating what changed in the model and why. No movement past the threshold means no note — silence is a stated part of the method, not an oversight.
Every round is on the public record: the methodology changelog lists each model version, how many activity indices it recorded, and when.
Limitations
- The activity index is a judgement, not a measurement. It is reproducible and shows its working, but reproducible is not the same as correct.
- Three of the four components rest on your own answers about your own role.
- O*NET describes American occupations, and transfers to other labour markets imperfectly.
- Automatable is not automated. Cost, regulation, liability and inertia are not in this model.
- People are poor at estimating where their week goes.
This page includes information from the O*NET 30.3 Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. Modaltech has modified all or some of this information. USDOL/ETA has not approved, endorsed, or tested these modifications.