Thursday, October 15, 2009

Causation as Storytelling

I've pointed out before (and it's not my original observation) that no branch of modern science contains a notion of "cause" more than vaguely similar to the folk psychology notion -- causation, as we commonsensically understand it, is something that we humans introduce to help us understand the world; and most directly, to help us figure out what to do....

[David Orban, in a comment on an earlier version of this post, noted that some formulations of relativity theory contain the term and concept of "causation." But causation as used in that context is really just "influence" -- the restrictions on light cones and so forth tell you which events can influence which other ones, but don't tell you how to distinguish which events are causal of which other ones in the stronger, commonsense usage of the "causation" concept.]

Cause, in our everyday intuitive world-view, is tied to will: "A causes B" means "I analogically intuit that if I were able to choose to make A happen, then B would happen as a result of my choice."

And, I think cause is also tied to storytelling. Causal ascription is basically a form of storytelling.

Think about the archetypal story structure:

If we envision a typical causal ascription as fitting into this structure, we have:

The Setup is the situation in which the causation occurs, the set of "enabling conditions." For instance, we rarely would say that oxygen is the cause of us being alive -- oxygen is considered an "enabling condition" rather than a cause of our life ... it's part of the set-up....

The Confrontation is the introduction of something unusual into the Setup. This must be something that is not always there in the Setup, otherwise one wouldn't be able to isolate it as the cause of some particular events. It's not necessarily a violent confrontation, but it's a violation of the norm. Could be someone shooting a gun, could be a couple having sex, could be a finger pushing down on a computer keyboard. The less expectable and frequent it is, the better -- i.e. the more convincing it will be as a potential cause of some event.

The End is the event being caused.

My suggestion is that, if one digs into the matter deeply, one will find that many of the same patterns distinguishing compelling stories from bad ones, also distinguish convincing causal ascriptions from unconvincing ones.

What Would Aristotle Say in a Situation Like This?

The Aristotelian distinction between efficient and final cause is also relevant here.

"Efficient cause" is what we usually think about as causation these days: roughly, A causes B if

P(B|A,Setup) > P(B|Setup)

and there is some "plausible causal mechanism" (i.e. some convincing story) connecting A and B in the context of the Setup.

"Final cause" is telos, teleology -- A causes B if B, as a goal, somehow reaches back in time and makes A happen as part of the inevitable movement toward B's occurrence.

Modern physics theories have no place for final causes in the Aristotelian sense. But, human psychology does! Very often, when a human seeks a cause for something, what they're doing starting with some event they've observed and trying to find a "good reason" for it.

Why did I fall in love with her? It must have been because she was beautiful ... or smart ... or rich ... or whatever.



Why did my business succeed? It must have been because I was smart ... or because it was the right time ... etc.

Storytelling generally mixes up efficient and final causation in complex ways. Many stories give a feeling of inevitability -- final causation -- by the end. And when postmodernist stories avoid giving this feeling, it's generally done intentionally, with a view toward violating the known psychological norm and doing something disconcerting or shocking.

Convincing causal ascriptions, like compelling stories, tend to mix up efficient and final causation.


Cause and Will

Nietzsche wrote that (paraphrasing) "free will is like the army commander who takes responsibility, after the fact, for the actions of his troops."

Experiments by Gazzaniga, Libet and other neuroscientists have validated that in many cases the reflective, willing portion of the human brain-mind "decides" to do something only well after some other part of the brain has actually already started to do it.

This fits in fine with the notion of causal ascription as storytelling. Willing is a matter of making up a story about how one came to do something. It had better be a compelling story or the illusion of free will will fall apart, which is bad for the maintenance of the self-model!


Causation and Storytelling in Neuroscience, AGI and Early Cognitive Development


One concrete hypothesis that comes out of this train of thought is that, when the neural foundations of causal ascription and storytelling are unravelled, it will turn out that the two share a large number of structural and dynamical mechanisms.

Another hypothesis is that, if we want our AGI systems to be able to ascribe causes in humanlike ways, we should teach our AGI systems to tell and understand stories in a humanlike way.

I strongly suspect that one of the major roles that storytelling plays in human childhood, is to teach children patterns of narrative structure that they will use throughout their lives in constructing causal ascriptions (along with many other kinds of stories).

...

Ahh ... I would love to improve this blog post with a bunch of concrete examples but that will need to wait for later ... I'm tired and need to wake up early in the AM ... at least "I" think that is the cause of me not wanting to improve it right now ;-D ...