I can learn many new things for fun but I already know a lot of things I'm fond of, which one should I give away given that time is finite? I didn't know anything when I was little and much less than now when I was half my age. Furthermore there are a few things I like and I became reasonably good at, professionally or for fun. I won't give them away unless I have to.
I think the multi-armed bandit problem would assume that the probability distribution, for each choice, would remain a constant for the duration of the problem.
Diminishing returns would imply that some subset of those choices would have decreasing probability of a positive pay-off the more they are used. This would be more realistic, and much harder to model. With the constant probabilities with the payoff being for a single score, it makes sense to just search for the highest expected value score. But that expected value relies on the probability distribution remaining the same!
Perhaps it's just a habit I've stumbled into because of ADHD, but I find that in the long run choosing to spend time on learning something very different doesn't feel like a zero sum game. It's because the different new things to learn compliment the old things and you can leverage that.
https://en.wikipedia.org/wiki/Diminishing_returns
I can learn many new things for fun but I already know a lot of things I'm fond of, which one should I give away given that time is finite? I didn't know anything when I was little and much less than now when I was half my age. Furthermore there are a few things I like and I became reasonably good at, professionally or for fun. I won't give them away unless I have to.