“Don’t know” is a valid output
Some heads are harder to read than others. Bone thickness varies, and for some people there is no usable window at all. A product that pretends otherwise, and shows a confident number anyway, is worse than no product.
So the device has a third output next to a value and its uncertainty: “don’t know.” On a bad window it says so, and the model above it works with what remains. Not-knowing is priced in everywhere in the system: a wider interval, a value for the missing information, and the right to stay silent.
We think this is the single most important design rule for anything that reads a mind. If it cannot say when it is blind, nothing else it says can be trusted.