About a month ago I committed myself to an anti-akrasia resolution, inspired by Yvain and ZM. I won't repost the resolution here, but if you want to see it, click the first link. The essence of my resolution was to commit myself to practice math to prepare for graduate school in the fall. The resolution was valid from June 18 to 28 - I ended it there because the next day I was on a plane to Orange, CA to start an internship (in experimental economics, if anyone's interested). Now, I'm going to provide a lot of anecdotal evidence and speculation. Please do not fall prey to the typical mind fallacy. The structure of my brain may be sufficiently different from yours that nothing I say here generalizes. I also could be deluding myself. This is only anecdotal evidence, and I have no real way of knowing whether or not my perception of what happened is biased.
With that being said, there is some objective evidence, namely how much I deviated from my resolution. There is also what happened after my resolution while here at Chapman University. Here, again, the evidence is mostly subjective with the exception of how much I stuck to the old resolution.
To start with the beginnning, the first day my resolution was in effect I worked diligently for well over two hours - probably closer to 4 - working through multiple chapters of my old probability text... actually, that was a lie. I had planned a date with my girlfriend that day, and she assumed that the date day continued on through the evening. It was a fair assumption on her part, so I used my one free day on the first day of my resolution. Great start, eh?
From that day on, I did pretty well. It wasn't until near the end that I finally broke the resolution. Twice I did it because I realized I needed a second loophole - a way to trade an extra two hours one day for a free day at a later date. On the last day I broke the resolution because I didn't have the foresight to plan a full day with my girlfriend before leaving for six weeks. So I broke my resolution 33% of the non-free days (total days minus the one free day I gave myself). If we count the extra long sessions as compensating for the first two days, that reduces to 11%. Not bad, considering that this summer I had opened a textbook up exactly twice before my resolution.
When I originally posted my resolution, I said I would make a new one once I made it here to Chapman and got a feel for how much time I had to spare. Well, I never got around to that (Akrasia applies to plans to fight akrasia. It really is vicious). The fact that I didn't put that in my resolution may or may not be significant. I don't think it is. I did continue to do math, at first. My first two mornings here I spent a solid two hours apiece reviewing some analysis. There were a couple more days where I might have worked for about an hour. But for the most part, I haven't don't much studying in the past three weeks.
This meager data suggests that the resolution worked for the specific duties I mentioned, but there could be counfounders. I may have been gung ho about it because I had just worked myself up. As time went on, this might have waned, and caused me to not renew my resolution as well as not do as much math. This doesn't seem like what happen to me, but the data doesn't rule it out.
As far as the "maximize my utility function" clause, i have no objective data. From here on out, it's just my impressions. Take them for what they are worth.
It did seem like I did a better job of doing what I really wanted to do (i.e. maximizing my utility function) even when taking into account that my resolution was already binding me to do better for two hours out of the day. I spent less of my free time playing Settlers of Catan online1 and more hanging out with friends, getting paperwork filled out for graduate school, reading books, etc. I seemed to spend much less time, well, killing time. It wasn't perfect though, I still killed time doing mostly worthless things. I did much better with the specified things2 in my resolution than with the utility maximization clause. It helped, but not as much.
After the resolution expired, however, it was back to my old ways. Lots of flash games, little productivity. In fact, this is largely the reason that I haven't posted about The Simple Math of Everything yet (it's coming...) and why it took so long to write this followup. If MichaelHoward hadn't asked me how everything panned out, I may not have even written this.
I have an idea3 which I think explains why there was a difference at all and why the resolution worked at all. The resolution, being so formal, was in the forefront of my consciousness. Normally, it seems, my mind would query "what should I do" and faced with the daunting task of actually computing this, a null would pop out. Basically akrasia was preventing my efforts to stop akrasia. And so I would go on doing whatever worthless nonsense I was doing the day before. Playing cheesy computer games, watching sportscenter, etc. I wanted to do what I should do, but when I actually sat down at my desk and saw my computer in front of me, I wasn't even thinking about what I should do anymore. My resolution, in effect, kept this in my consciousness, which made it easier for me to stop akrasia before it starts. I would think, "what should I do?" and my mind would immediately go to my resolution.
This explains why it worked at all, but why the difference between specific activities and the utility maximization clause? The difference seems to lie what had to be computed after I thought of my resolution. With specific activities, I could just go down a checklist. When the checklist was completed, I actually had to think about what I really wanted to do, which just left the door open for akrasia to rear its ugly head again. I suspect I can improve my ability to just figure out what I want to do and then do it, but for now having specific things outlined in a resolution is more effective.
This also seems consistent with my previous experience. When a deadline is nearing, I simply think about whatever I should be doing more which makes it easier to conclude "I'm going to do that" and then actually go and do it. At work, I know I have specific duties, again seemingly with the same effect. Far away from a deadline, I'm simply not thinking about it. I'm not really sure why - it seems to be a subconscious process that I can't really control - but it happens. It might be related to stress - the resolution might be creating a mild amount of stress which, in turn, motivates me. It didn't feel like that - I really enjoyed the math I was doing, but it's worth looking into.
One other thing I should mention about my resolution. I used a green dry erase marker to write LW in large capital letters on the resolution as it hung in front of my desk. This is close to the symbol for LessWrong in Google Chrome (the W is lowercase), so it grabbed my attention and reminded me of what I should be doing. Again, it seemed to keep me conscious of what I needed to be doing. I probably should have made my computer's wallpaper a collage of these symbols or something to help keep it in my mind. (This may be part of the psychology behind chaos magic). After rereading the thread about chaos magic I linked to, I immediately thought "so I guess that's what priming feels like." Priming is probably a big part of the explanation I have above, but I don't think it is everything. Sometimes I was aware of what I should do seemingly without being primed - like when I'm away from home, or immediately when I wake up. This is testable though; perhaps I'll do a run without having the resolution in sight and check if the effects change.
Now take all of this with a grain of salt. Your mind is not my mind, so nothing I said above need apply to you. For example, not all akrasia has to occur because you simply aren't concsiously thinking "I should do X." Some people may be constantly thinking that and still do Y, but that doesn't seem to be the issue with me. I could also be very deluded about what actually happened. These are, after all, largely my subjective impressions. However, I intend to renew my resolution and monitor the results more closely. It did seem to work even if I'm completely wrong about why. Perhaps you have a better explanation? I'm listening.
1. It's my own solitare. The company that hosts the servers uses a different name for the game to avoid a lawsuit by the manufacturer. I'm not telling you what they call it because it really is a productivity buster. Google at your own risk.
2. Shortly after I posted my resolution, I ammeded it to add a few more specific duties which I'd rather not mention in such a public medium. Suffice to say that results were similar to that of the math duties.
3. I don't think this explanation and the reasons that the resolution should work that Yvain gave (and linked to) in his original post are mutually exclusive. Hyperbolic discounting might be the reason I was less conscious of what I should be doing, or the fact that I was less conscious might be the reason hyperbolic discountings works as a good model. Or neither, I'm not sure. But I see no reason why they are fundamentally incompatible.