I’ve noticed two points in recent life where I’ve fallen prey to Goodhart’s law, and I’m working to improve them.
1) On my birthday a few days ago I finally dropped a goal I had for this year, to read as many books as possible. (My main strategy for this was to tell people how many books I’d read and maintain a list on Goodreads whenever I finished one.) For the last three years I’ve tried to measure books per year as a way to gauge how much I was learning, but as I improved and the number increased the differences between “reading lots of books” and “learning” began to magnify.
2) I’ve tried many different methods for organizing tasks, and my current method is to fill cells in a spreadsheet with goals and then color them in as I complete them. Despite that I enjoy it, this has led to many unnecessary problems, like running out of tasks to fill the space presented and feeling a gnawing incompleteness. Worse, if there’s a large task then either it can generate akrasia sitting alone in its cell or split up into a column of steps that are mostly unactionable.
I’ve also had to accept the repeated evidence that majestic-looking lists of tasks will only lead to eighty or so items being completed in a day, and I’ve moved more toward making one column of immediately-actionable tasks, finishing 75% of it or more, and then creating a second and a third and so on. (This still produces the flaw of all task systems I’ve found, that a kernel of akratic items will accrete and show up over and over from list to list, but I haven’t worked out a solution to that. Even typical “eat the frog” style advice only changes the composition of that kernel, hopefully to things you don’t care so much about.)
So in both of these cases I’ve noticed that I’ve gotten sidetracked by optimizing things I don’t (want to) care about. Strategizing my day and making sure I don’t forget anything doesn’t require thinking of a number of tasks divisible into columns of 20 or even keeping track of how many things are done. Likewise, reading lots of books is orthogonal to learning lots of things, and the right volume on its own can be worth fifty or a hundred random paperbacks.
The problem I continue to run into, and the root of why Goodharting is so easy to do, is that it’s difficult to operationalize these goals any other way. The desire to learn can be broken into “true” goals GTD-style and then measured by understanding certain texts, carrying out certain actions correctly or passing certain tests, but this requires a pretty solid understanding of the subject matter. I’d like to learn more about polyominoes, for instance, but knowing what that “more” consists of (or whether, having learned about them, I’d rather have learned something else) requires its own baseline of knowledge. My next solution to this going forward is probably going to be to have smaller, “scouting” goals for understanding fields enough to develop more concrete ones.
Operationalizing productivity itself seems more difficult, but I think part of it might be resolved with a better integration of habits or schedules. One very big flaw of all my lists has been that they create a sense of accomplishment whenever they’re filled out instead of encouraging a general sort of well-being throughout the doing of the tasks—writing only feels good once I’ve written a certain amount. The moments where I’m most thankful for lists like these are ironically the times that they drop out of my awareness and I start wanting to do more in one of the streams of work I’ve entered into.
I’ve noticed two points in recent life where I’ve fallen prey to Goodhart’s law, and I’m working to improve them.
1) On my birthday a few days ago I finally dropped a goal I had for this year, to read as many books as possible. (My main strategy for this was to tell people how many books I’d read and maintain a list on Goodreads whenever I finished one.) For the last three years I’ve tried to measure books per year as a way to gauge how much I was learning, but as I improved and the number increased the differences between “reading lots of books” and “learning” began to magnify.
2) I’ve tried many different methods for organizing tasks, and my current method is to fill cells in a spreadsheet with goals and then color them in as I complete them. Despite that I enjoy it, this has led to many unnecessary problems, like running out of tasks to fill the space presented and feeling a gnawing incompleteness. Worse, if there’s a large task then either it can generate akrasia sitting alone in its cell or split up into a column of steps that are mostly unactionable.
I’ve also had to accept the repeated evidence that majestic-looking lists of tasks will only lead to eighty or so items being completed in a day, and I’ve moved more toward making one column of immediately-actionable tasks, finishing 75% of it or more, and then creating a second and a third and so on. (This still produces the flaw of all task systems I’ve found, that a kernel of akratic items will accrete and show up over and over from list to list, but I haven’t worked out a solution to that. Even typical “eat the frog” style advice only changes the composition of that kernel, hopefully to things you don’t care so much about.)
So in both of these cases I’ve noticed that I’ve gotten sidetracked by optimizing things I don’t (want to) care about. Strategizing my day and making sure I don’t forget anything doesn’t require thinking of a number of tasks divisible into columns of 20 or even keeping track of how many things are done. Likewise, reading lots of books is orthogonal to learning lots of things, and the right volume on its own can be worth fifty or a hundred random paperbacks.
The problem I continue to run into, and the root of why Goodharting is so easy to do, is that it’s difficult to operationalize these goals any other way. The desire to learn can be broken into “true” goals GTD-style and then measured by understanding certain texts, carrying out certain actions correctly or passing certain tests, but this requires a pretty solid understanding of the subject matter. I’d like to learn more about polyominoes, for instance, but knowing what that “more” consists of (or whether, having learned about them, I’d rather have learned something else) requires its own baseline of knowledge. My next solution to this going forward is probably going to be to have smaller, “scouting” goals for understanding fields enough to develop more concrete ones.
Operationalizing productivity itself seems more difficult, but I think part of it might be resolved with a better integration of habits or schedules. One very big flaw of all my lists has been that they create a sense of accomplishment whenever they’re filled out instead of encouraging a general sort of well-being throughout the doing of the tasks—writing only feels good once I’ve written a certain amount. The moments where I’m most thankful for lists like these are ironically the times that they drop out of my awareness and I start wanting to do more in one of the streams of work I’ve entered into.