The closest I come to artistic action is writing serviceable non-fiction prose, so you might say I'm more to the craft side of the spectrum than the fine arts side.
One thing I'd like to explore is the effect of feedback on performance, both artistic and programmatic. Gordon Pask, an early cybernetician, made something called a Musicolor machine, which tied a keyboard into a light show with not-quite-predictable delays in between (if I recall right - I'm getting a paper on it Monday). It would be neat to build an emulator, but I doubt I have time before OOPSLA.
(I can find no web description of the Musicolor.)
Programmers get feedback in what may be an oddly different way than artists. A performance artist gets immediate feedback. A novelist gets feedback at wide intervals. Programmers get feedback somewhere in between. How can the different types of artistic feedback inform programmer feedback?
Another thought: Dick Gabriel talks a lot about artistic selection: generate a zillion tries and pick the best. You could map that onto a typical iterative project, in which you could say each successive iteration is a try at a structure that's both flexible and useful. The obvious difference is that each iteration builds on the last, but surely that's also true - perhaps to a lesser extent - of Dickey's huge sets of draft poems? Where could a comparison take us?