If you go looking for an app in this space you'll find three kinds: streak counters with neuroscience vocabulary, detox challenges with a timer, and a small number of things trying to model something.
Here are the questions that tell them apart. They apply to us too, and we answer them at the end.
1. Does it claim to measure dopamine?
If yes, stop there.
No consumer product measures dopamine. Not an app, not a wearable, not a ring, not a headband. Dopamine varies by brain region, receptor type and timescale, and there is no non-invasive consumer method for reading it. A product claiming otherwise is either using the word loosely or making it up, and either way you can't trust the rest of the numbers.
The honest version of the claim is modelling — inferring reward-system state from behaviour you logged. That's a real thing to do. It's just not measurement, and the difference should be stated by them, not discovered by you.
2. Does it show you its model?
A number is only as good as the thing that produced it.
If an app gives you a score with no account of how it was computed, you can't tell whether it encodes published theory or someone's intuition about what feels bad. Ask what the inputs are, what they're weighted by, and where the weights came from.
Almost nobody in this category will tell you. It's worth noticing how unusual that is for something asking you to change your life based on its output.
3. Does it distinguish established findings from its own guesses?
Every model of this kind contains both — published theory, and parameterization the builder chose.
That's fine and unavoidable. What isn't fine is presenting the second as the first. "Neuroscience-backed" applied uniformly across a product usually means nobody separated them, and you have no way to know which claims are load-bearing.
4. Does it count volume, or structure?
Streaks, hours avoided, days clean — these are volume metrics, and volume is the weakest variable in the literature.
The things that predict a sliding baseline are structural: how predictable a reward is, how much effort gates it, how frequently it recurs. Schedule beats duration, and effort changes what a reward costs you. A tracker that only counts can't see either.
5. Is it built on abstinence?
If the core loop is "avoid things and maintain a streak," it inherits every problem of the detox framing — including the wrong timescale and severity as the only dial.
Worse, it makes effortful rewards look like failures. Under a load model those are protective. An app that scores your run and your scroll the same way is missing the variable that matters most.
Our own answers
Fair to hold us to it.
Measure dopamine? No, and we say so in the app, on the landing page, and on the science page. Myo models reward-system state from what you log.
Show the model? Yes — every mechanism is published with its source and an honest label, not described in the abstract.
Separate established from ours? Each mechanic carries a visible tag: whether the code is a direct statement of published theory, or directionally grounded and our own choice. The effort-protection function, the most load-bearing assumption in the product, is labelled as ours.
Volume or structure? Structure. Schedule, effort, novelty and frequency all enter the model; there are no streaks anywhere in it.
Abstinence? No. The engine treats effort-gated reward as protective, which is why it never tells you to want less.
What we're not: not a clinical instrument, not a diagnosis, not a validated model. It's a faithful implementation of published theory with our own parameterization, and we label which is which.
The short version
The bar in this category is low enough that "tells you how it works and admits what it doesn't know" is a differentiator.
It shouldn't be. Use the five questions on anything you're considering, including this one.