Principles

What guides us.

These principles apply to everything we build at Allostatic Labs — regardless of which project is currently being worked on.

Full locality

What we build processes data, wherever possible, directly on the user’s device — not on a server, not in a cloud. This is not a privacy policy added after the fact, but an architectural decision from the start: where local processing is possible, there is simply no channel through which data could flow away. Less transmission means less risk and more control for the person concerned.

Patterns, not content

We never capture what someone writes, reads or says. Where our software observes behaviour, it is abstract patterns — rhythm, pace, pauses — never the content behind them.

Observation, not judgment

The states our software recognises are neither diagnoses nor performance verdicts. They are observations from which users draw their own conclusions — not judgments imposed on them.

Transparency, not a black box

Where our systems infer states or patterns, we prefer transparent, rule-based methods over opaque models. What a system does should be explainable.

Accessible to everyone

We build multilingually (German/English) and with accessibility in mind from the outset — not as an afterthought, but as a basic stance.

These principles derive directly from our understanding of research — more on the “Research” and “Transparency” pages.