DRMacIver's Notebook

Action generators

Action generators

Quick, do something.

Now do something without leaving your current position. e.g. if you're sitting down stand up.

Now, without leaving your current position, do something with your left hand.

Now, without leaving your current position, do something with your left hand without touching anything with it.

Now, without leaving your current position, do something with your left hand that feels good, without touching anything with it.

Each of these instructions is an action generator: Something that prompts you to act in the world. These can be instructions, but they can also be observations, thoughts, etc. If a ball comes flying towards your head, you hopefully duck. Action generator!

Any action generator has the default response to it: Do nothing. When I told you to do something, you probably ignored it and kept reading (this is of course doing something). Possibly by the end the instructions were interesting enough that you went "Oh, sure, I'll give that a go.", possibly not.

Doing nothing isn't always a good idea. If you have a do nothing response to the ball flying towards your head, you get hit. Hope it's not a cricket ball. Those hurt. A lot.

The thing about the list of action generators I gave above is that each of them is more specific than the one preceding it, and yet it is very unlikely that the actions generated by the last are ones you would have thought of in response to the first action generator.

This is because the broader the action generator is, the more possibilities there are, the harder it is to navigate to any particular possibility in particular.

It's also because the more specific action generator gives you permission. The ball coming at your head generates an action of duck, so you duck. You could have ducked even if the ball wasn't coming at your head, but why would you?

Action generators are often useful for doing things that you could have done with a broader action generator but, for one reason or another, didn't.

If someone is coming around, that's an action generator to clean your house. You could have done this at any time before, but you didn't. That isn't necessarily bad, but sometimes you wanted to do it before and still didn't.

This points to the fact that an action generator really has at least two properties:

  1. What actions it can generate.
  2. What actions it encourages you to take.

Someone coming over doesn't constrain your actions much compared to the default action generator, but it does create a very strong incentive to tidy up, in much the same way that you're free to do whatever you like in the lead up to the ball arriving, but after that if your actions have not taken the ball into account you're going to deal with the consequences.

One thing I'd like to highlight from the hand thing is that to a large degree, making the action generator more specific automatically creates an encouragement to action, because it lowers the activation energy to acting by removing a large class of decisions. "What should I do with my life?" - vast and impossible set of decisions. "What should I do today?" can still be pretty daunting. "What should I eat for breakfast?" - now pretty accessible. Onward! Towards the abyss.

Another thing worth noting is that any theory about the world is an action generator. You create a model of the world, and this gives you a set of actions that makes sense under that model. For example, when we invent symbolic games to improve our ability at something, we are doing this.

Importantly, a theory works as an action generator even if it's wrong. You can still adopt the theory and act based on it, and often those actions will be perfectly fine, because they're interesting in their own right even if the reasons that lead to them aren't correct. This is akin to the saying that all models are wrong, but some are useful. All models are useful, in that having action generators is useful because it may lead you to things that you wouldn't otherwise have thought of.

I think many "bad" theories persist because they're useful action generators. More, though, I think often bad theories are better than good theories, because they generate actions that you would never have thought of if you were constrained to adopting more reasonable theories.

For example, I think a lot of therapy modalities work like this. My reliably experience reading therapy books is that they explain their theory and I go "lol, sure" and they provide their evidence and I go "an easier explanation of this is that you're lying", and then I do their exercises and... well, most of the time it doesn't work, but it does something interesting often enough that it was worth reading the book.

I think this is often the case: It's worth having a stable of bad theories lying around that generate actions that you wouldn't otherwise have thought of and that are basically harmless in the worst case, and possibly interesting in the best case.