What is meta-rationality?

Benjamin P. Taylor
4 min readMay 11, 2022

To work from facts and use theories, methods, and logic, we need to deal with the undefinability of the world. So if we want to get something done, we need to think about meta-rationality.

Roughly following David Chapman’s scheme (see www.metarationality.com), we can think about the world as infinite nebulosity and limitless pattern.

Diagram from David Chapman on Twitter — @meaningness

To ‘make sense’ of this in a rational system, we need ‘circumrational support’.

Think of a self-driving car. We can’t, in fact, identify traffic lights: there is no rule set that will always identify a traffic light and never misidentify something which is not a traffic light.

Therefore we need a whole set of support activity. We need to take the input to the car, and the context, and create a set of rules for what will be considered a traffic light, and what won’t

Much of driving is routine and predictable and that can be handled very well.

Then, we need to take account of context, in various ways — for example, to expect traffic lights more often to be at junctions, at the side of the road, and in the middle above the road. Not way up in the sky or on billboards. This means that we need another set of rules to decide what we’re going to identify as roads, what we identify as buildings — and how we define the sky.

And we will sometimes get this wrong. So we need to build in short-term improvisation and review of our way of seeing and understanding — and of course longer-term learning.

Catching sight of traffic lights on the bed of a highway maintenance truck in front of us on the highway, or on a billboard will fool the algorithm, so we might want to build in a check: ‘wait, traffic lights don’t usually travel at 70mph’…

We might also want to change the way we see — add or remove laser, radar, infrared as well as light sensing cameras, change angles, even change the infrastructure so road and traffic lights send us signals.

We’re not cars. Our thinking and reviewing is much more complex than what happens in a computer, and much more apt to adjust in real-time. We can also negotiate with other people as we engage with them, and share our perspectives and interpretations.

Remember, all of this is before we get to be able to do work with fact, theories, and approaches, before we’ve classified what we’re taking in.

And we’re not talking about emotions or habit or embodied thinking — or the physical constraints on what we can take in. All of these are relevant.

The magic is that although rational systems don’t contain their own foundations — in theory terms or in terms of engagement with the world — we can use our rational thinking to think about how we’re creating the circumrational support to make sense of the world.

It’s easy to take that for granted — because it usually works, for most purposes.

But if we want to get stuff done, we need to pay attention to that as well as the systems we use once that framework is in place and working.

How does this explanation land?

Can you see examples of it in practice?

Dr Jared Dempsey commented:

  • All I know is there are cat eyes on the last slide and any post that has a cat or dog in it is a winner. And I really liked your egg and nest picture to illustrate just near difference. From this you can guess I like pictures…

My response:

  • Ah yes, I’ve decided I need a cat or dog *every* post now… seems to work…
  • But are they really cat eyes, or is it a whole cat? That is the question…
  • And it’s an interesting connection to the story of the invention of ‘cats eyes’ in roads — the depressable reflective lights marking lanes. IIRC from school, the inventor nearly went off a cliff in the fog when he saw his headlights reflected in the eyes of a cat sitting on the barrier at the side of the road — a deeply meta-rational story (since I’m on the ‘driving’ theme).