How much time should you spend optimizing any particular process you might engage in? Even assuming that you’re optimizing for a value of overriding importance or value there is only a limited amount of time available.
If all available time is spent optimizing clearly that would be suboptimal since there would be no time left to actually engage in any particular process pursuant to what we value. So the optimal level of optimization is always suboptimal.
However, that might seem to be trivial and only operant at some kind of asympoptic limit we need not worry about in our lives. The problem, though, is deeper. That the optimal level of optimization is suboptimal is both a kind of trivial truth as our time is finite but also a statement about our epistemology; our capacity to know what level of optimization is optimal (even if suboptimal) is itself subject to the same kind of suboptimality that any other kind of knowledge-pursuit is.
This is not to say we should pursue single-mindedly the optimization of any value(s) we desire; rather, it is a guard against the view that optimization is synonymous with perfectionism.
I find this confusing because I think it’s (unintentionally) equivocating on what optimizing means.
Naively my response is that if you find you spent too much time trying to optimize for something, you did a suboptimal amount of optimizing. But this wouldn’t mean the “optimal level of optimization is suboptimal”, it would mean “optimizing only for goal B is suboptimal relative to goal A”.
Programmer often talk about this problem as premature optimization: writing code that is optimal for some local concern but comes at the cost of effecient allocation of effort towards the larger goal.
I think perhaps a better way to put what I interepret to be your point would be “the globally optimal amount of local optimization is locally suboptimal”.
This is further complicated by some people overestimating their ability to optimize things, and underestimating the costs (in time, but also psychological) of doing that optimization. For example, I worked in a company that had lots of meetings to make sure that we were doing the higher-priority task instead of a lower-priority task, but I often suspected that if we didn’t have these meetings, we would simply have enough time to do both. Maybe in the wrong order, but the overall outcome would still be better.
Some unsolicited feedback: I believe I would have a easier time understanding this blogpost if it focused on some tangible examples or referenced similar ideas which are already well documented (like the Secretary Problem?). The third paragraph is particularly hard to understand and I’m not sure I grasp it
(you’re saying that while there may exist a theoretical point where more ‘optimization’ is possible—that there is a law of diminishing returns as the cost of searching for that rapidly decreasing extra-output shrinks?)
What heuristic on evaluates this at is at peril too, applied heuristic of not bikeshedding and incessant yak shaving is probably fair bit applicable here.