A notational stumble it took me a while to solve when reading this—what’s M(q)|H(q,M) supposed to mean? My guess at the answer, so that others can get through this quicker, or so that evhub can tell me I’m wrong:
M(q) is supposed to be M’s distribution over answers to questions q. If δ is a distribution over set S and s∈S, then δ|s is the probability mass δ assigns to s. Finally, H(q,M) is how the human would answer question q when given access to M. So, M(q)|H(q,M) is the probability that M assigns to the answer that human H would give if H could use M to answer q.
Yeah, that’s right—for a probability distribution P:Δ(X), I mean for P|x to be the value of the probability density function at x. I edited the post to clarify.
A notational stumble it took me a while to solve when reading this—what’s M(q)|H(q,M) supposed to mean? My guess at the answer, so that others can get through this quicker, or so that evhub can tell me I’m wrong:
M(q) is supposed to be M’s distribution over answers to questions q. If δ is a distribution over set S and s∈S, then δ|s is the probability mass δ assigns to s. Finally, H(q,M) is how the human would answer question q when given access to M. So, M(q)|H(q,M) is the probability that M assigns to the answer that human H would give if H could use M to answer q.
Yeah, that’s right—for a probability distribution P:Δ(X), I mean for P|x to be the value of the probability density function at x. I edited the post to clarify.