Mark Histed on Nostr: This framework is well-worked out in theory—we must choose an error criterion ...

This framework is well-worked out in theory—we must choose an error criterion ("wrong"), which comes w/ a cost in time/$.

This isn't ex it (need a cost term), bt same idea, fm David Heeger's notes.

Type 1/2 errors, sensitivity/specificity: same idea.

https://www.cns.nyu.edu/~david/handouts/sdt/sdt.html