There is a whole category of app now that will read your messages and hand you back a number.
An interest score, telling you how much the other person really cares. A ghost risk rating. An effort balance across the last month, showing who has been putting more in. Hot and cold tracking. Several of them will now draft your next reply for you.
I want to be clear at the outset that I have no objection to being handed a number. I build tools that produce them. Measurement is how somebody gets out of an unwinnable argument with themselves and onto firmer ground, and anybody who has spent months going round the same conversation and coming out less certain each time has earned firmer ground.
The problem with these apps is not that they measure. It is what they have chosen to measure.
YOU CANNOT SCORE A RELATIONSHIP
An interest score is a verdict on a relationship. A ghost risk rating is a prediction about another person’s intentions. An effort balance is a summary judgement about two people’s inner commitment.
Look at what each of those is asking the software to do. It has a partial text record, selected by one participant, covering some fraction of the exchanges between them. From that it produces a figure about the whole thing.
Not just the messages. The whole thing. What happened in the room. What happened on the phone. What happened over eleven years. What was said in the car on the way home and never typed anywhere.
The issue is not that the figure comes out inaccurate. It is that there was never a measurable object there in the first place. A relationship is not a quantity. Another person’s inner state is not observable from outside, by software or by anybody else, and a number attached to either is a guess wearing the costume of a measurement.
WHAT IS ACTUALLY MEASURABLE
Behaviour is.
Who said what. What happened immediately before it and immediately after it. Whether an account of an event was later replaced with a different account of the same event. Whether an apology arrived with a condition attached. Whether a request was met with a response about something else entirely. Whether the same shape appeared again three weeks later in different words.
Every one of those either happened or did not happen. They can be located, dated, sequenced, and compared against established descriptions of specific behaviours that researchers have spent decades documenting.
That produces a completely different kind of number. It does not say your relationship is a four out of ten. It says this particular behaviour occurred, here is where it occurred, here is how severe it appears to be, and here is how confident that reading is given how much material it had to work with.
One of those is a verdict. The other is a finding, and a finding is something you can do something with.
THE SECOND QUESTION, WHICH IS THE MORE IMPORTANT ONE
Measured against what?
This is the question almost nobody asks of a number, and it does more work than any other question in this domain. A score with no framework behind it is an opinion with a decimal point on the end.
When something hands you a figure, ask what standard produced it. If the answer is a proprietary model, trained on unspecified material, calibrated by nobody in particular and validated against nothing you can inspect, then the precision of the number is doing work its evidence cannot support. Two decimal places is a rhetorical device, not a claim about accuracy.
If the answer names an established body of research, and shows you which specific words in your own material produced which specific reading, then something entirely different is on the table. You can check it. You can disagree with it. You can look at the marker it flagged and say no, that one was a joke, you have misread that. You can take it to a professional and they can see what it saw.
That is the difference between being told an answer and being shown the working.
WHY THIS MATTERS MORE HERE THAN ALMOST ANYWHERE
In most domains a rough summary figure is harmless. Here it is actively counterproductive, and for a structural reason.
Manipulation of the serious kind is built out of individually deniable moments. That is not a side effect of how it works, it is the design. Each incident, examined alone, is defensible. It was a bad day. It was taken out of context. It was a joke. It was your reaction that made it into something.
Which means the harm is only ever visible in the sequence. And a whole-relationship average is precisely the operation that destroys a sequence. Good week, difficult week, quiet fortnight, it all nets out to something unremarkable in the middle. The very thing you needed to see is the thing the averaging removed.
Behaviour-level analysis across the whole record does the opposite. It keeps the sequence intact, which is where the shape lives.
WHAT CHANGES WHEN THE UNIT CHANGES
The finding becomes checkable. You are no longer being asked to accept a conclusion on trust. You can look at what triggered it and form your own view, which is the entire point.
The unanswerable questions drop away. Is he a narcissist. Does he actually love me. Neither of those has an available answer and both keep people circling for years. What did he do, and does that behaviour have a name, are answerable.
Your own judgement comes back into play instead of being replaced. This matters more than it sounds. By the time somebody is analysing their own relationship, what they have usually lost is not information but confidence in their reading of it. A verdict substitutes for that confidence. Evidence rebuilds it, because you are the one doing the reading again, with something solid in front of you.
You end up with something you can take to somebody. A screenshot of a score is not much use to a therapist, a solicitor or a friend. A documented pattern of specific named behaviours, with the material that evidenced each one, is a different conversation entirely.
THREE QUESTIONS FOR ANY TOOL
Before you trust a number about your own life, ask three things.
What is it measuring. If the answer is you, or him, or the relationship, it has picked a target that does not hold still long enough to be measured.
What is it measuring against. If there is no named framework, there is no standard, and without a standard a number is decoration.
What does it hand back. A conclusion invites you to stop thinking. A finding, with the evidence attached, invites you to think properly for the first time in months.
WHERE THIS LEAVES YOU
The instinct that sends somebody to one of these apps is sound. Something in you has registered that this situation warrants examination, and that registration is your judgement working rather than failing.
The instinct just deserves a better instrument than a rating out of ten.
Validate was built for exactly this. It reads real communication against peer-reviewed frameworks for coercive control, gaslighting, emotional manipulation and related patterns, and returns four things together: the name of the behaviour, how severe it appears, how confident that reading is given the material, and the exact wording that produced it. It does not diagnose anybody, it does not score the relationship, and it does not tell you what to do. It shows you what is there.
If you want somewhere to start before the app goes live, the Relationship Clarity Check is a short set of questions that will tell you whether what you are experiencing has a name.
https://validate-quiz.scoreapp.com
Nobody can hand you a number that tells you what your relationship is worth. What can be handed to you is an accurate account of what has been happening, which is the thing you were actually looking for on the night you first typed it into a search bar.
