Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Notes on utility function experiment

13 Post author: taw 05 September 2009 07:10PM

I just finished a two-week experiment of trying to live by a point system. I attached a point value to various actions and events, and made some effort to maximize the score. I cannot say it was successful in making me achieve more than normally during the same period of time, but it made more clear some of the problems with my behaviour.

Here's some notes from my experiment:

  • Points are marginal utilities, they are for things you want to do more of, but don't due to akrasia. If you want to exercise more you'll assign very high value to half an hour of exercise, but that doesn't mean you want to spend 8 hours a day cross-training. As expected, I got most points for thing that weren't that important but I finally got myself to do more.
  • Values can be put on terminal values (results), or instrumental values (effort, and partial results). Valuing only the former tends to be highly demoralizing, valuing the latter tends to be highly encouraging.
  • It's a good idea to assign some points to cleanup of your system (decide against doing something that was previously on your list, get trivial thing off your list). It cleans your mind, even if it doesn't progress your big goals.
  • Another good idea are points for sitting down and thinking, making mindmaps and so on.
  • One big problem that the system didn't cover at all were distractions, like spending too much time on Wikipedia or TvTropes.
  • Another big problem were times when I didn't have enough energy to do anything big, but wasn't sleepy enough to sleep. It's usually pure waste of time.
  • During the first week I was more successful (in terms of points) almost every day, then it went far downhill, perhaps due to my enthusiasm running out. This made me decide against extending the experiment past the original two week schedule.
  • In other words - there was some value to it, but akrasia mostly won again.

Anyone else wants to share their anti-akrasia experiments?

Comments (8)

Comment author: John_Maxwell_IV 06 September 2009 02:57:57AM *  8 points [-]

Anyone else wants to share their anti-akrasia experiments?

Making a list of things to accomplish next day each evening worked moderately well for me. Logging all my activities quickly became a pain in the ass. Following a routine in the mornings and evenings worked fairly well until I realized my routine needed refactoring, abandoned it, and never got around to refactoring it. (As a general principle of self-improvement, I think it's important to stick with solutions that you suspect are suboptimal but are still better than whatever your default pattern would be. Let's say you're thinking up some scheme for yourself and then say "I'll never get this right". Well, if the scheme you've got so far is better than what you'd normally do, you should still stick with it.)

Comment author: Emily 06 September 2009 03:12:58PM 1 point [-]

Do you use any system to make the points worth something real apart from just satisfaction at having maximised them? Something like... reaching a certain number of points in a week = favourite chocolate bar?

Comment author: taw 06 September 2009 05:05:06PM 0 points [-]

No. The points have no external value attached to them. If I wanted chocolate bars, I'd just get them anyway.

Comment author: Yvain 06 September 2009 01:49:41PM *  1 point [-]

I'm currently doing something similar to this as the natural progression of the method I described here. The main difference is that instead of maximizing points, I'm holding myself to a budget of points.

It started by me telling myself I was going to spend two hundred fifty minutes / day doing useful self-improvement type activities. Then I decided that certain activities were only marginally useful, and started counting them at a rate of one minute per two minutes, and others were extremely useful, and counting them at a rate of two minutes per minute, and now it's more or less the same as your point system. I also subtract points for certain things I want to do less of. I've been doing it for about a month now pretty successfully.

At some point I want to make a post on it, but not until I've got more to say about it. I'm trying to think of it with an economic metaphor, as the personal equivalent of subsidizing useful activities and taxing useless activities to change the resources devoted to each, but I can't really present the analogy coherently until I understand the mind better.

Comment author: taw 06 September 2009 05:06:55PM 0 points [-]

Is overhead of measuring time spent on a particular activity a big problem, especially if you're multitasking like most people?

Comment author: Yvain 07 September 2009 09:09:55AM 0 points [-]

Time budgeting hasn't been a problem; it's pretty easy to keep track of.

Comment author: Psychohistorian 05 September 2009 11:29:31PM 0 points [-]

An interesting idea. It sounds like it was structured more like a video game (get the highest score) than a budget. If you used the same system and assigned positive and negative point values to activities you want to do more or less of, perhaps with an appropriate sliding scale (e.g. the first half hour a week on tvtropes is free, but it gets pricey after that), I think you'd see less akrasia. Enforcement is obviously the big problem, though.

As it is, it sounds like you're just pretending you want to do things that you don't actually want to do and hope that getting a high score will be enough to motivate you to actually do them.

Comment author: taw 06 September 2009 02:45:35AM 2 points [-]

As it is, it sounds like you're just pretending you want to do things that you don't actually want to do and hope that getting a high score will be enough to motivate you to actually do them.

I often don't care much for particular activities, but I definitely care about results, which require me to do the activities. Add hyperbolic discounting (effort now, results later) and risk aversion (efforts certain, results uncertain, probability for risk-aversion is irrationally sublinear) and you'll see the problem.