All that's necessary for a proof system in a blockchain, as far as I know, is that mining be hard to do and easy to validate. Creating works of art is hard to do, but determining their value is relatively easy. The only issue I, who knows very little about blockchains, can find with this is that it would be quite slow. What are some issues that might come up with this, and is it generally plausible?
These are not the only requirements.
The most fundamental requirement is that the PoW actually verifies something. It's not enough to do work and prove that you've done it - the work actually has to express acceptance of a specific branch of the block tree. Otherwise, it's useless in preventing double spending.
It follows that the problem to be solved must be generated from the details of a specific branch/block, and the proof must show that work was done with this specific block in mind. The work cannot be reused to signal acceptance of a different block. The block is baked into the work in the most fundamental level.
How would you do that for creative work? How do you generate creative assignments that are tied exclusively to some specific block? That's completely far-fetched. If the work is, say, a painting - once the work is done and the painting is drawn, it can just as easily be assigned to block A as it can to block B. It therefore verifies nothing.
Another requirement is that the work is not only verified easily, but also in a digital and decentralized way. Your software can, without any additional knowledge, computationally verify that some amount of hashing was put into a block. It cannot verify that someone who voted is indeed a human - it needs to rely on external trusted sources for that.
Which leads to the final point, which virtually everyone who proposes alternative PoW mechanisms miss - if there was a way to verify the identity of humans and collect their votes, there would not be a need to have proof of work at all. Instead of voting on works of art, they can just vote on which branch they think is right. This way they can synchronize transactions and prevent double spending. The whole reason we don't do that in Bitcoin, is that there is no way to verify identities in a decentralized, digital way, with the standards sufficient to reliably administer the voting. This is why we rely on independently verifiable proof of computational work.
Only if you mean "something like" very, very loosely.
The entire logic of proof of work is that it is expensive to make, can be irrevocably, provably attached to a block quickly, and is cheap to rapidly verify automatically. How would human creativity voted on by other humans be irrevocably, provably attached to a block?