Is Sufficiently Compressed Thinking Indistinguishable from Magical Thinking?

Andrew Brown is in The Guardian claiming any sufficiently interesting theory of the world is indistinguishable from religion.  There’s a lot to debate there, but do you mind if I put the broader issue aside for a second to come to the defense of science!  Brown writes:

And atheism can be just as theologically incorrect: today’s paper told me that: “our bodies are built and controlled by far fewer genes than scientists had expected“. The metaphors of “building” and “controlling” have here taken a concrete form that makes them palpably untrue. Genes don’t do either thing. It seems to me that a belief in tiny invisible all-controlling entities is precisely a belief in the supernatural, yet that is the form in which entirely naturalistic genetics is widely understood in our culture. Religion can’t really be about doctrine and heresy either, because these concepts don’t make sense in pre-literate cultures. You can even ask whether the concept of “supernaturalism” makes any sense in most of the world without a developed idea of scientific naturalism, and scientific laws, that would stand for its opposite.

Whoa, whoa, whoa.  First of all, the conceptual errors and simplifications of science journalism (that linked quote comes from a reporter) are not a guide to the blind spots of science as an epistemology.  But I also take some issue with the idea that using agent-language is necessarily an error mode in science.

Science is a model-making endeavor.  It’s never possible to prove that your model is in one-to-one correspondence with the World-as-it-is, so you evaluate different theories by seeing how good they are at anticipating data collected from the world.

(Notice that I used a colloquialism there and let ‘theories’ take an active verb instead of saying something wordier and more precise like “there is a high degree of correlation between the outputs of your theory’s algorithm and previously unobserved observations, and it didn’t mean I literally believed that theories were immaterial actors or otherwise supernatural).

Your model may look different at different scales, because the most accurate model may be more granular than needed.  We can use a lot of Newtonian physics for human scale predictions, even though we know it’s not true.  It’s true enough for the problem at hand.  The trouble is, that if you’re making a habit of using an easier way but sloppier way of answering questions, you want to make sure you’re also setting aside time to practice recognizing the moments when precision matters enough that you should dust off your more accurate model and get out the instruction manual for all the little levers and bits that go boink.

Most of us, scientists or not, don’t spend enough time auditing our cached thoughts and heuristics, and the one that Brown flagged is a pretty persistent problem.  I think a better word than “supernaturalism” would be “agent-seeing.”  (Ok, better in the sense that it’s more accurate; I’m open to catchier substitutions).  Ascribing purpose or agency to inert matter or abstract concepts doesn’t necessarily lead to belief in the supernatural, though it may produce weird theories of consciousness or an increased expectation that other intelligent life exists.

The main problem when people talk about genes acting or wanting isn’t the use of shorthand.  It’s that once we use the word want, we assume genes must want what we want since we want the best things.  The map is supposed to reflect the territory, but you have to be careful about which way you’re letting the data flow.  The existence of fold lines on a map should not cause you to expect very regular divots in the landscape (Bryan Caplan terms this the metaphorical fallacy).

Don’t lose hope!  Shorthand makes life go a lot more smoothly, and if you take the time to check in on yourself, you can spot and correct for error modes.  So, if you notice that everything in the world has the same purpose and desires as you, that’s a good way to notice you’re slipping into error.  Pause, take a breath, and repeat to yourself: “Camelot! (It’s only a model).”