Yes, that is an astute observation and the only way you can gauge accuracy is to confirm ground truth some other way.

Try probing the system so that it conforms to your model and revise it when new evidence is apparent.

...and remember to make tests that don't confirm your model. If the set is 2, 4, 8, and your model is "powers of two", then, sure test 16, but also test 1, 3, 6, 0.5, 0.1 (maybe the correct model is "also 2 to the power 0" or "two to any integer, including negatives" or "all even numbers" or "any integer" - or something else).

>confirm ground truth

I'm not a philosopher, but I do enjoy reading and learning about post modernism.

Isn't a large point of this book that you actually can't confirm a "ground truth"?

Yea, but when I use it in practice I have to verify if my assertions are true. So I guess I'm doing "reconstruction" I'm not sure.