How this works, and where it's wrong

A tool that reads something this sensitive should be able to explain itself. Here is exactly what happens, what it can see, and what it cannot.

Educational content, written from published research and recognized domestic-violence resources. How this content is developed. MendRii is general-public educational software — not a clinical service, assessment or professional advice.

What happens when you press analyze

1. Your text is sent over an encrypted connection to MendRii's server and passed to a large language model with a fixed instruction set.
2. That instruction set asks for one thing: which documented patterns the wording may resemble, the lines each observation relates to, what the text cannot establish, and context worth thinking about. It is explicitly told not to diagnose anyone, not to label a person, not to guess at anyone's intentions, not to predict anything, and to say plainly when it found nothing.
3. The structured result comes back, is checked, and is shown to you. Free analyses are not written to any database. Saved conversations exist only if you are signed in and chose to save them.

What it can't see

• Tone of voice, facial expression, and everything that happened before the text.
• Who is who. If you paste a conversation, it reads both sides — it does not know which person you are unless you say.
• Your history, your fear level, your culture, your family's normal.
• Sarcasm and in-jokes, which it will sometimes read as hostility.

It is wrong in both directions

False positives: a heated but ordinary argument can contain sentences that look like a pattern. One observation about one message means very little.

False negatives: the most controlling relationships often have the calmest-looking messages, because the pressure lives in what everyone already knows will happen. A clean result is not a clean bill of health.

This is why MendRii shows patterns rather than a score, and why it keeps pointing you at people rather than at itself.

Where the patterns come from

The eleven patterns are drawn from how coercive control is described in the established literature and in service guidance — Evan Stark's work on coercive control, UK Home Office statutory guidance on controlling or coercive behaviour, US domestic violence service material, and Dr. Jennifer J. Freyd's research on DARVO. Sources are listed on each lesson.