On psychology and Newtonian mechanics
A note to Jasen Murray, then a researcher at Leverage Research — a Bay Area organization founded by Geoff Anders that was attempting to build “Connection Theory”, a putative comprehensive theory of human psychology and motivation. Murray had presented some of this work the night before; the response below is methodological advice from one rationalist to another about what it would take for a theory of psychology to be more than a sea of words.
In short, you’re almost entirely right about almost everything, which is ridiculously impressive.
To use the analogy with Newtonian mechanics, the most important thing about classical physics (and the reason it wasn’t discovered until the 17th century) was that it broke down intuitive notions into precise, rigidly defined concepts. Speed, velocity, and acceleration are very different things, and have to be spliced apart from the intuitive notion “fast”. Ditto weight and mass, energy and momentum, etc. These concepts were then tested in simplified, pure experiments, which eliminated as many confounding factors as possible. E.g. Galileo’s experiments with gravity eliminated friction, air resistance, etc. as much as possible, to show only the effects of gravity.
The book Thinking Physics (Epstein), which I just purchased a copy of, is a really awesome way to train your mind to do this sort of thing. Everyone at Leverage really really really ought to read it cover to cover. I will personally purchase it for anyone Leverage has hired or is considering hiring, as long as they promise to actually read it.
To make the comparison to psychology, concepts like “beliefs” and “intrinsic goods” really have to be rigidly defined, with many many many precise, black-and-white examples of what they are and what they are not. Otherwise, you’ll just wind up being Aristotle, who was really smart but got lost in a sea of words and vagueness (see e.g.
Paul Graham on philosophy ).
This doesn’t mean that you have to be able to take a real-life human, and exhaustively and perfectly describe their beliefs. Real life is messy and people are messy and doing things perfectly is impossible. But you should be able to construct simplified thought experiments, where everything is nice and neat and you can make exact predictions. We still can’t describe all the forces acting on the hundreds of pieces of a car when it smacks into a wall, but we can understand a perfectly smooth perfectly elastic two-ton steel sphere well enough, and its behavior is extremely useful for modeling the real car.
To take the example of a “belief”, what is the simplest possible computer program that has a belief? Suppose I build a robot that moves forward at 5 mph. Does it have a belief? Now, suppose I program it to only move forward if it is light out. Now does it have a belief? What if I program it, every time it gets bright, to speak the English sentence “It is light out” and then start moving forward? Does it have a belief about ambient light levels? Does a thermometer have a belief about the temperature? What about a thermostat, which reads the temperature, and uses that information to keep the house at 75 degrees? What about more sophisticated AIs? Does Deep Blue have beliefs about chess? Does Watson have beliefs about Jeopardy? What about all of Watson’s sub-agents? Does a dog have beliefs about the food he sees in front of him? What about a mouse? A lizard? An insect?
Humans are really messy and complicated, so before we talk about whether humans do or do not have property X, it’s important to define whether lots of simpler systems have property X, so we have a clear idea of what exactly X is and how we can test for it. To quote Feynman: So I began by asking, ‘Is a brick an essential object?’
Then the answers came out. One man stood up and said, ‘A brick is an individual, specific brick. That is what Whitehead means by an essential object.’
Another man said, ‘No, it isn’t the individual brick that is an essential object; it’s the general character that all bricks have in common — their “brickiness” — that is the essential object.’
Another guy got up and said, ‘No, it’s not in the bricks themselves. “Essential object” means the idea in the mind that you get when you think of bricks.’
Another guy got up, and another, and I tell you I have never heard such ingenious different ways of looking at a brick before. And, just like it should in all stories about philosophers, it ended up in complete chaos. In all their previous discussions they hadn’t even asked themselves whether such a simple object as a brick, much less an electron, is an ‘essential object’.