Responsible use
What the readings are never for, what your app takes on when it uses them, and what Neurologyca does on its side.
The indicators and the interpretation are estimates of a person’s context. They are meant to support the person and must never be used on their own to decide about them. This page sums up what that means for your app; the binding text is in the DevLab Terms and the Acceptable Use Policy.
What the readings are never for
Section titled “What the readings are never for”- Diagnosing, screening or treating a health condition, or presenting a result as doing so.
- Deciding about the person on the readings alone: hiring, dismissing, grading, credit, insurance or any decision with legal or similarly significant effects.
- Sharing a reading with anyone other than the person it was written for without their consent.
- Detecting lies, sincerity or intentions.
- Comparing or ranking people, or scoring them over time.
- Measuring without telling the person first, or monitoring anyone covertly.
- Manipulating: using how the person feels to push a purchase, a subscription or any other decision.
- Inferring emotions at work or in education, recruitment included, if your organization is established in the EU or the EEA, or the readings are used there, unless for a medical or safety reason the law accepts. There, and wherever the law forbids it, no reading for an interviewer who evaluates a candidate. Practising an interview is for the person themself, outside any selection process.
The other uses prohibited everywhere, such as identifying people or categorizing them by sensitive traits, are in section 3 of the Acceptable Use Policy.
What your app must do
Section titled “What your app must do”- Tell the person before measuring and collect their consent: what is measured, what for, that an AI writes the reading, and that they can stop at any time. A model notice to adapt is in §14 of the Biometric Addendum.
- Show each reading to its declared reader: every context is written for one reader. A reading for the person goes to nobody else without their consent.
- A person reviews every decision that uses the readings and affects someone: they can change or stop it, and the person can contest it. Do the same with Preview features: never use them to decide about people without that review.
- In health, wellbeing, mental health or emotional support, a route to help always visible: a crisis line or the emergency number of the person’s country, and a way to contact you. The service is not for emergencies, and your procedure for serious distress does not depend on the readings.
- Measure as often as the purpose needs, no more: never use the readings to make the person stay longer, come back more often or depend on your app, and let them pause or stop at any time.
- Explain the limits: they are probabilistic estimates, not a diagnosis nor direct access to anyone’s mind.
- Delete on request: Deletion and retention.
What Neurologyca does
Section titled “What Neurologyca does”- It designs each context with a declared reader and limits on what the reading may say.
- It gives the language model that writes the reading drafting rules that tell it not to diagnose, not to pass moral judgment on the person and not to attribute intentions to them.
- It receives reports of readings or uses that may harm a person, examines them and corrects what it finds. How to report: «Human-Centered AI» in the DevLab Terms.
These commitments do not replace yours, nor do they make the results suitable for every person or use.