I had virtually the same thought while showering this morning. I was running late for an extremely important exam and found that I was sphexishly staring at nothing.
I said to my brain, "We're late. Finish showering."
My brain replied, "Hold on, I'm optimizing something unrelated to this."
"This is kind of urgent," I suggested.
"It's not very rational of you to ask me to stop thinking about this and thus risk allowing a potential cognitive error to pass unchecked," my brain said haughtily.
"I think it is instrumentally rational for us to not be late to the exam."
"Perhaps, let us consider the value of sacrificing global correctness for local instrumental rationality - "
"Shut up," I said out loud, and finished showering. But of course I continued thinking and realized how often I allow meta-optimization to take priority over actually doing things, and how it only really becomes apparent when I am frantic about a deadline. This also explains why deadlines actually do motivate me despite the fact that I hate them.
I was reading the NY Times article on Decision Fatigue, when I came upon a hypothesis I would like everyone's feedback on.
I take as a premise that there seems to be a high prevalence of akrasia in the lesswrong community.
I also take as a premise that the sequences give us a more-than-usual detailed model of the world, one that presents us with more possible trade-offs we could be making in every day life.
So the conjecture that by trying to reduce bias and perform a lot of cognitive calculation, we effectively spend large parts of our days in a decision fatigued state, leading to akrasia problems.
Does this sound (un)reasonable? Why? How would you go about turning this into a testable proposition?
UPDATE: Anna Salamon has put up a detailed poll here that may shed some light on the situation. Please take some time to fill it in.