Interesting. It's a fixed point solution concept, where you do like a PD controller-style iterative re-adjustment based on the current sleep efficiency score. J

Interesting. It's a fixed point solution concept, where you do like a PD controller-style iterative re-adjustment based on the current sleep efficiency score. Just happens I was talking to my dad about fixed point approaches to sleep just a few hours ago. When people change their wakeup time, one of two things happens: they either stick to their new habit successfully, or they break it, ignore it's broken, leave the alarm on but just ignore it. There's little opportunity to break free from the cycle because to re-enter the cycle is harder than exiting it. So you need some mechanism that provides sporadic reward for re-entering participation into the desired behavior. Whatever your fixed point of behavior becomes, it becomes what you stick to. So you have to 'jam' the fixed point nature. Jostle it with some noise. Basically - change the meta. My proposed solution models bedtime as a game, played by "night me", "morning me", and a randomized alarm clock. Alarm clock: either goes off at e.g. 5am (I want to wake up at 6am, so 5am is like a kind of punishment) with probability p, or it doesn't go off at all Night me: I won't know whether I will just 'sleep in' (relative to 5am), or whether I will be rudely awoken. The hope is it makes me hedge my bets - maybe I'd strategically adapt my nighttime activity, prudently leaning towards winding down earlier if I'm uncertain about my rest that night. Or I'll just be tired anyway because I will have repeatedly already demonstrated that I haven't yet learned the lesson to hedge one's time.. (ie got woken up because I got greedy about bedtime several times already) ..and go to sleep earlier anyway.. Morning me: the subconscious me, really- I only have one move - to wake up - but in some sense it's out of my control until I've 'trained' myself via this routine. The score of this move is a function f(whether alarm actually went off, time I would have naturally woken up): - the alarm doesn't go off, I get to sleep in, yay, bc if I'm sleeping in I must have needed it - the alarm doesn't go off, I woke up early anyway, nice, that's the desired behavior - the alarm goes off, I don't get to sleep in, boo, I should have gone to sleep earlier, oh well, I was or will be tired tonight, anyway, probably - this is the only real 'negative' quadrant - the alarm goes off, but I was already awake around that time anyhow, meaning I've become 'over-trained' for my desired target - meaning, I can shut this whole game down for a few days! By sort of anticipating a kind of homeostatic control that will make my wakeup time drift away from my overshoot target (5am) back to my actual target (6am), I will probably wake up later than this unusually punishing wakeup time. And, as a bonus, it will be perceived as sleeping in, simultaneously :) It's gotta be that I could possibly get so wiped that I'd even be willing to go to bed not just like 30 minutes sooner but e.g. 3 hours sooner. That level of signal in behavior change is what I'm after, if possible This is kind of like the "folk theorem" in game theory - you can kind of fudge the payoffs in infinite games Nash equilibria to basically be whatever you like. You only need to dish out as little punishment as is necessary to encourage positive reinforcement becoming the Nash equilibrium. So, in this context - pausing the 'game' a few days once I naturally overachieve my wakeup target = sleeping in more = higher ratio of the game payoffs being good ones = Nash equilibrium to pursue current mode of 'play'. I hope by reaching a kind of a Bayesian Nash Equilibrium in this weird bedtime 'game', I jam my own meta, and avoid fixed point patterns where I just ignore my original goal. Though, to be honest, that's likely what will end up happening in the long term haha. ps: you could also have a kind of stochastic bedtime approach to night time distractions too. Basically, somehow, set it up so that there's a p probability risk that you lose all your open tabs, unsaved files, internet connection, active phone calls, etc., at some uniformly unpredictable time e.g. up to 3 hours before bedtime. Hopefully you plan ahead, save all your work, and wind down far in advance, so you don't get burned by just surfing the net too long or something like that.