Reading your description though, the only way I can imagine such a machine working would be by analyzing what is said against an objective truth - and that is what the false confidence value is measuring?

If I’m confident of something that is wrong, my false confidence number would be very high - especially if I believed it, right?