What algorithm could be used for discovering the next best intervention one can make to improve oneself?
Trying a bit of this, a bit of that, and comparing results? I doubt it can get any more precise, because the interventions on different levels can be, well, different.
For small changes I would recommend trying each strategy one week (to filter out work-day cycle and other noise), and having a set of similar tasks, randomly assigned to those weeks; or one repetitive task. But some level of change would probably disrupt such setting. As an example, if my task is to "motivate myself to clean my room" and the intervention is "move to a different environment", then of course, when I move to a new room, cleaning it is a different task than cleaning my old room, so it is not completely fair to compare my efficiency in those tasks.
Could the high level be discovered gradually, instead of making the first correct guess? Such as: start with some low-level improvement, and when you find that something is stopping you, analyze it, and do a meta-action. So instead of starting at the right level (and risking going too meta), we could instead start at the right place (where the outcome is measurable) and gradually find the necessary level of change.
But even this cannot be done exactly, because at a sufficiently high level I may choose a different outcome. For example, I work at a paperclip factory, and my initial goal is to make more and better paperclips. First step: I try doing overtime, but then I find I am too tired to continue this way. Second step: I get regular sleep, exercise and eat healthy food. Third step: I attend paperclip-making lessons. So far my progress is measurable. Fourth step: I realize I actually don't care about paperclips, I just do it to make money; so I change a job to something that pays better. Oops, my first metric (paperclip) just broke; I need to replace it with money. Fifth step: After having enough money I realize that more money does not bring me most happiness, so I would prefer having more free time while making the same amount of money; or maybe less money but also less expense. Oops, my second metric broke too, and no replacement is precise enough... I could try some psychological questionaires for measuring happiness, but that seems to easy to cheat.
The "What is Rationality?" page on the new CFAR website contains an illuminating story about Intel:
I presume Andy and Gordon had considered intervening at many different levels of action: in middle management, in projects, in products, in details, etc. They had probably implemented some of these plans, too. But the problem with Intel — it was in the wrong market! — was so deep that the place to intervene was at a very low level, the foundations of the entire company. It's possible that in this situation, no change they could have made at higher levels of action would have made that big of a difference compared to changing the company's market and mission.
In 1997, system analyst Donella Meadows wrote Places to Intervene in a System, in which she outlined twelve leverage points at which one could intervene in a system. Different levels of action, she claimed, would have effects of different magnitudes.
This got me thinking about levels of action and self-improvement. "I want to improve myself: where should I intervene in my own system next?"
My bet is that if the next greatest leverage point you can push on is something like neurofeedback, then you're pretty damn self-optimized already.
In fact, I suspect almost nobody is that self-optimized. We do things like neurofeedback because (1) we don't think enough about choosing the highest-leverage self-interventions, (2) in any case, we don't know how to figure out which interventions would be higher leverage for ourselves, (3) even if there are higher-leverage interventions to be had, we might not successfully carry them through, but neurofeedback or whatever happens to be fun and engaging for us, and (3) sometimes, you gotta stop analyzing your situation and just do some stuff that looks like it might help.
Anyway, how can one figure out what the next highest-leverage self-interventions are for oneself? Maybe I just haven't yet found the right keywords, but I don't think there's been much research on this topic.
Intuitively, it seems like hacking one's motivational system is among the highest leverage interventions one can make, because high motivation allows on to carry through with lots of other interventions, and without sufficient motivation one can't follow through with many interventions.
But if you've got a crippling emotional or physical condition, I suppose you've got to take care of that first — at least well enough to embark on the project of hacking your motivation system.
Or, if you're in a crippling environment like North Korea or Nigeria or Detroit, then perhaps the highest level intervention for you is to get up and move someplace better. Only then will you be able to fix your emotions or hack your motivational system or whatever.
Maybe there's something of a system to this that hasn't been discovered, or maybe there's no system at all because humans are too complex. I'm still in brainstorm mode on this topic.