Beemind batchiness, not urgency

Akrasia
1

We’ve been talking in the Discord about meta goals for beeminding less edge-skating of one’s Beeminder goals. This has long been a popular topic here in the forum (see, for example, a thread from a year and a half ago and ongoing discussion now on refinements to beeminding urgency load, by @scarabaea and @alephnull: A new meta goal concept)

I want to give a “proof” that beeminding eepiness / edge-skatey-ness suffers from a TANSTAAFL problem, and then relay an insight from @poisson about what this argument misses.

Consider two worlds:

  1. Status quo: I have a dozen beemergencies every day at 5pm
  2. Meta goal beeminding urgency load: I have one beemergency every day at 5pm

World 2 sounds nice so far. But what does that one beemergency consist of? Um, bumping a dozen goals from n safe days to n+1 safe days. I.e., getting back to the urgency load I had yesterday.

But that’s exactly the same work as getting a dozen goals out of the red.

If I do any less than that, my urgency load goes up. Which my meta goal – if I’m edge-skating it, which I definitely am – doesn’t allow.

Maybe world 2 is still nicer because if life happens or whatever I derail one thing instead of a dozen things? That’s not really much help because derailing the meta goal means letting the dozen goals end up in the red. (This may depend on how much I derail the meta goal and how much urgency load I allow myself.)

Poisson’s Epiphany

The real value to beeminding urgency load (or whatever proxy metric for edge-skatiness you like) is that it minimizes fracturing of your attention. It does this by encouraging you to build up more safety buffer on goals you’re on a roll with, rather than nudge every single goal forward by epsilon.

Imagine I have just 2 goals that are in always in the red: every day I have to do 1 unit of work on each of them. Now I add a meta goal to maintain 3 days of safety buffer across both goals. The next day I need to dispatch both beemergencies (that’s 2 days of safety buffer) and also pick a goal to get ahead on. From the next day onward, I still have 2 units of work to do but only one goal in the red. I can devote both those units of work to one goal. The next day the other goal will be the beemergency and I can devote the 2 units to that one. Rinse, repeat.

If I beemind more than 3 days of total buffer I can have multiple days in a row of focusing on one goal. Plausibly this is more Deep-Work-compatible.

7 Likes
2

My main argument for having a meta goal rather then every single goal in the red is that I can choose what I can work more on, every single day. E.g. I have several time-based language goals contributing to the meta, and I can alternate between languages, not pull each of them out of the red every day.
@poisson 's conceptualization brilliantly wraps the same principle for N=2.

3 Likes
3

Yeah, I suspect this has been obvious to many of you for a long time and I finally had the forehead-smacking moment when @poisson explained it.

The N=2 example is my own stylization just to help make it more intuitive.

Poisson also emphasized that it adds valuable flexibility. Some days you physically can’t do certain goals. If instead of 12 beemergencies you have 1 beemergency to add 12 days of buffer across your 12 goals, you can concentrate on getting further ahead on the goals you physically can do.

(And now that I say it that way I realize it’s exactly the same argument.)

4 Likes
4

For me, watching my urgency load wasn’t just about attention fracturing.

My commitments were relatively variable, and I’d often have the opportunity to take another short job on. It was pretty difficult to actually get a good feel for how busy things were going to be a week or two into the future, vs how busy they were that day. Urgency load didn’t solve it (weirdly enough I had some things in my life that weren’t tracked by Beeminder) but it was a great help.

7 Likes
5

Reading @adamwolf 's old post on this is exactly why I started tracking urgency load this year (and also added a lot more regular things to beeminder). I’m still worried about what happens when my life inevitably goes south (beeminder can look forward to that day with excitement), but for now it’s given me a lot of interesting information about how much slack I have on any given day.

It’s also given me some motivation to not immediately use up any buffer I accumulate, because I can immediately see the impact on my slack.

3 Likes
6

I agree with this line in particular

It does this by encouraging you to build up more safety buffer on goals you’re on a roll with, rather than nudge every single goal forward by epsilon.

I do think that a lot of these type of discussions have less to do with urgency and more to do with kind of “superminding” if you will, for those with just a ridiculous # of goals

It kind of becomes a different type of product when you’re beeminding 50 - 60 goals instead of 5 - 10

Some things matter a lot less than others, but would still be nice to beemind to generally improve on / stay consistent at, as I find the graph & process of optimizing the pace to be intrinsically rewarding

But if I’m going to get charged for them constantly I would beemind far fewer things – indeed I found this weakened how much I cared about actually important goals (being charged is more of a mental signal than the amount)

I felt like buffer-managing a bunch of low priority goals kind of distorted how much I cared about my higher priority goals for some reason, I think basically this reason

But that’s exactly the same work as getting a dozen goals out of the red.

Personally my solution to this (for now) was to upgrade to the highest plan so I could do $0 goals, so then I have kind of 2 tiers of goals, the core ones I care a lot about and the more aspirational ones

I know this is viewed as non-ideal, the only alternative I could think of was to create an entirely new app / channel to manage those second tier goals, I started that but then was like “this is stupid, I should just do $0 goals”

So far $0 goals for those has worked fairly well. Sometimes I find “oh I care about this goal a lot more than I thought I would” and then bump it up, but I start at 0 now.

I’m not completely sure if that’s the right direction, will need to give it another 4-5 months or so to properly feel it out, maybe I come back to buffer management

Even if the tool were capped to say 10 goals it would still be immensely useful

But yeah going far beyond that is a complex problem that feels like it differs from the core (perhaps intended) workflow

5 Likes
7

So, one anecdotal data point on this.

For the first couple weeks of the month, mostly it was just easy keeping up with new urgency_load goal. I do think this has had small benefits in general—I still sometimes have goals in the red, but a couple goals that in the past have been edge-skaty have been pretty consistently in the blue or green for the past week or so.

Last night, I still had a deficit of ~4 days of urgency load… after a late-night social event, so it was only like 15 minutes until the midnight due date! I dispatched the beemergency by making a bunch of predictions on fatebook.

In a way, this ended up being an example of both advantages I mentioned, I think. Firstly, some of my goals simply take way more than 15 minutes to even produce a single day of progress—so if one of those had been in the red, instead of urgency goal, it simply would have been impossible to do it. So the buffer helped.

Secondly, in its own way, focusing on trying to push out several days of progress on fatebook at once was of some value, I think. I’ve kind of been doing the minimum every day on that one… but I think somehow just being forced to make one prediction a day hasn’t really been enough to force me to think systematically about how I can do something interesting with fatebook… I kind of just sit there staring at the wall until I think of a random thing to predict. But being forced to come up with 5 predictions in a row kind of pushed me to think a bit more about how I’m going to make use of fatebook in general. Of course, maybe I could have achieved the same thing by making the slope 5x as big but… well, obviously in the long term that would have problems.

4 Likes