Home Archive Wiki About
Home Archive Wiki About

It might be interesting to have a realistic planning system

Posted on 2018-12-09

There’s many different kinds of things to do

There’s a bunch of things I want or need to do, and they have different shapes.

Productivity methodologies I know about are way too narrow

There’s a bunch of methodologies for productivity, but I feel they often only model a small part of planning. For example, (my interpretation of) GTD puts everything in a framing of “you have Projects and projects have Actions”. But I don’t think that’s a good framing for, say, learning a language.

The model of “cost of time” is also wrong. It is a useful heuristic, but the value you assign to any activity is going to assume perfect elasticity. For example, you cannot just walk up to your employer and say “I want to work 120 hours per week”. No (reasonable) employer will say “yes” to that. Also, the many things a person wants cannot be converted into a one-dimensional value. (Footnote: Yeah I know about von Neumann-Morgenstern theorem. But people are not rational agents, and von Neumann-Morgenstern does not say anything about how practical will the resulting utility function be to evaluate.)

What I tend to do is some sort of intuitive “higher-level planning” which is sometimes a bit reflective, but not very often. When I’m in “work mode” (i.e., in my actual job), I have a few ways I try to figure out what to do on any particular day.

But the process by which I decide, say, “enough work today, let’s go get some sleep”, or maybe “I’m a bit tired but let’s go walk on the treadmill for a while”, is very intuitive.

I don’t know about any formal methodology which would basically take an input like:

and which would output a plan of what I should be doing at any given time.

And I would like the plan to be reasonable. The methodology should not, for example, assume I can do without sleep, or without a regular sleep schedule.

And that’s what I mean by a “realistic planning system”.

Maybe good old fashioned AI-style planning could be adapted?

In uni, I studied a bunch of planning and scheduling, which is mostly used for cases like “this is how you construct a submarine; you can’t screw in screw B217 until you screw in screws B210-B216; make a schedule which takes as little time as possible”.

These algorithms can be extended to work in more general environments, like:

I wonder if you could make a realistic and formal planning system from some extension of that setup. Say something like:

Meh, too hard. I got other stuff to do.

I guess getting any model halfway realistic would be too complicated, and I probably will keep using my bunch of ad-hoc heuristics to make decisions. I have way too many things that I want to do to spend the first couple years developing a planning algorithm rooted in psychological theory.

A thing I used to do that might be useful to start doing again is having some time in which I try to optimize what I’m doing. It could bootstrap into more conscious/mindful action.