Thread by @GeePawHill

I wrote about one of my big and comical mistakes, and it took off. It was cool being internet-famous for about 48 hours. Let's take up the general topic of mistake-making, considering it through the eyes of the change-harvesting worldview, and see what we get.

(I'm a little daunted. That mistake story has 2300 rt's and 9000 likes, which is truly staggering. How do you follow up a wretched excess like that? I've decided the key is just to keep on keepin' on, and do what I normally do, which is write a muse on Sunday morning.)

Mistakes are a kind of change, yeah? You make a change, and over some period of time, possibly quite long, possibly only a few seconds, you realize that the change has consequences you do not like.

Or, is that right? I mean, it's not really making a change, per se, it's more making a choice. Some choices are "not to change", and those choices can sometimes also be later seen as mistakes.

Even granting that some mistakes are "not-changing" mistakes, you'll notice that most of our fear about mistakes, and our prescriptive advice, isn't actually about the risk of not-changing, but about the risk of changing.

That's the first difference between change-harvesting and traditionalist models: we say the *only* non-changing thing about a dynamic unity is the changes it harvests. The day it stops changing is the day it dies.

For the moment, let's blow past that, I wanna get to mistakes.

In the software trade, when we see a mistake, we usually try to do two things: 1) Fix the mistake. 2) Adjust our behavioral strategy so that we won't make the same mistake again.

It's possible not to do that second part. A sure sign of a noob geek is actually how much skipping of that second part goes on in their minds. Those noobs are still developing their grasp.

If you see a given mistake as a one-off, a single of a unique error, you react to it differently than if you see it as one member of a family of related errors.

In my bug story, I had the same important value expressed as a literal in two different places. A noob might say, fix this error by changing both places. An olb might say, fix this family of errors by using a manifest constant and referencing it in both places.

This is where behavioral advice comes from, w.r.t errors: it comes from people connecting seemingly disparate unique events into a family of events, and then devising some new behavior that will eliminate or mitigate those kinds of events.

And these families of mistake aren't just parent-child pairs, they form larger and larger composites. You gradually see you can meta- it and meta- it a long ways up into your behavior. When you see two parent mistakes are themselves related, you posit another level, and so on.

In that way, "use manifest constants instead of repeating the same literal number" gradually becomes "don't repeat yourself".

Important: Don't get all reductio ad absurdum with these ideas. They're not really rules, they're advice, and they're vague and shmooshy, they're majority cases with exceptions, and they're best understood as preferences. Rule-izing everything is itself a family of mistake.

So, turn to change-harvesting. Way way way up the enormous family tree of mistakes, the change-harvester is offering very meta- advice about both what we change and how we change it.

A change-harvester says: the most successful change strategy, producing the greatest progress, with both the fewest and least-damaging mistakes, are those whose approach and substance can be described as human, local, oriented, taken, and iterative.

Human: the systems we're trying to change involve humans, and the humans are the most powerful determinant of the results of our change. We "lean in to" common human strengths and we "lean away from" common human weaknesses.

Humans are, for instance, *great* at 1) live active direct conversation, 2) holding <10 entities in their head at one time, 3) naming and chunking and stacking, 4) rapid visual feature detection, 5) using short-term gratification to spur energy.

Humans are, for instance, *lousy* at 1) math combining probability and non-linearity, 2) boring repetitive tasks, 3) knowing when & how their bodies need care, 4) remembering 371 simple rules, 5) taking action when they're scared.

Local: The systems we're trying to change are both large and subtle, with lots of what Hofstadter calls "strange loops". The most successful changes are small ones, occurring in some sense of nearness, be it mental, physical, temporal, or gratification.

There are a million drawings of the Satir change curve. They involve a triggering event, a performance trough, an integration, and a climb to a different level of performance, in that order. (Just do an image search, you'll find one.)

The key insight of locality is that the relationship between the size of the trigger event and the depth and width of the performance trough is not linear. The bigger the event, the bigger-*er* the trough.

Oriented: If we make all our changes local, how do we get to a change that is far away on the horizon? Orientation says, at each point where we make a change, our best strategy is to simply turn to face that horizon point, and choose any local step we think will bring us closer.

Orientation is about accepting the limits of our ability to plot a straight idealized line from where we are now to any horizon point. It notes that the land we'll travel is unknown and constantly changing, and that we need sustenance as we go.

By orientation, we sidestep detailed long-term planning, with all of its faux precision, simple addition, and incredibly hairy politics, and focus on the process of acting now to gain sustenance now. It is orientation and locality together that focus us on limiting scope.

Taken: It is no good starting at the city on the hill and working backwards. We have to start where we are and work forwards. A taken change is one that is a change to what is right here in front of us right now.

This notion embeds the idea of "change" itself into our worldview. So many approaches to change are based around "greenfield" thinking: the idea that the site of our change is virgin land, ready for any arbitrary alteration. The "taken" idea points out that this is nonsense.

All change is change to a thing. It's all brownfield. The taken approach changes what is there, locally and respecting orientation, but above all what is really there.

Iterative: Change-harvesters make changes over and over again, see no such thing as an end to change, and cheerfully accept that our model implies that we will be changing the same thing more than once.

Watch a skilled artist drawing, say, a cat. She draws the *whole* cat, vaguely, loosely, broadly, then refines her drawing many times. Each layer of refinement involves numerous changes to the same square inch on which she'd worked before. This is iterative change.

In the iterative approach, we want each change to be local and oriented and value-yielding. We care not one whit about whether we've changed something we've already changed once before. The effects of this style are global and gigantic.

We started with mistakes, then families of mistakes, then advice, then families of advice, and we wound up here, talking about the change-harvester's approach to all of this.

There are lots of little change-harvesting advice-elements, captured more-or-less pithily: "Easiest Nearest Owwie First", "Try Different, Not Harder", "Don't Avoid Change, Make It Cheaper", and so on. Many of them apply in lots of situations.

If I had to summarize our stance on mistakes: "Don't worry so much about whether you're going to make a mistake, because you are definitely going to make a mistake. Worry about how quickly you'll notice it, how easily you'll repair its damage, how calmly you move beyond it."

Hate long twitter threads, but like my content? We cool: These muses, on a short delay, turn into blogs and podcasts, and you can guarantee to get them in your favorite format. It's free, spam-free, and helps support my work.

geepawhill.org

Subscribe today!

(Phew. Okay, fine. We say you gotta get back on the horse when you fall off, but we rarely mention how hard it is to get back on the horse when you've just had a fabulous ride. I'm glad this first "post-viral" muse is done. Maybe now I can settle down and get back to work.)