Cognitive bias
·
A long time ago, as a naive cognitive science student, I struggled a lot with the problem of cognitive bias in humans and how to overcome it, given the obvious harm it causes.
Eventually I came to terms with what I have since articulated as "human minds are not rational truth-preserving inference engines unfortunately corrupted by bias; we are bias-implementing engines that somehow manage not to routinely walk off cliffs."
In other words, there is no "eliminating the bias"; it is bias all the way down.
To be clear: that's ABSOLUTELY NOT to say we can't get rid of biases that differentially harm people we care about.
We absolutely can, and we ought to.
But we won't do it by pretending to be rational truth-preserving inference engines, any more than we will do it by pretending to be invulnerable superheroes or medieval knights or incorruptible paladins of infinitely benevolent deities.
We do it, when we do it, by exchanging patterns of thought and behavior that differentially harm people we care about for patterns that don't... or at least, that don't as much. The new patterns aren't somehow more rational than the old ones, or less biased, or anything of the sort, any more than they are medieval or divine.
They simply don't harm people we care about as much.
And that's REALLY IMPORTANT, but it's a different thing.
That said, it's common in my circles for self-described rationalists to talk about "eliminating bias" and "being rational" and so forth. They award themselves and each other rationality points for completing various rational quests along these lines, and imagine themselves and each other as truth-preserving inference engines having truth-preserving rational adventures. And all of this is perfectly fine. It's basically truth-preserving inference-engine LARPing, and it can be fun; I enjoy that sort of role-play myself when I'm in the right mood.
And sometimes it inspires actual real-world work to improve our truth-preserving inference abilities, in much the same way that a SCAdian LARPing as a medieval knight can get a healthy cardio workout and greatly improve their health and reduce their risk of heart disease. As aspirations go, there are far worse ones than rational thought.
And, sure, there are people who take it too far. That's true for role-playing in general. And here again, there are worse choices -- someone who thinks he _literally is_ a rational truth-preserving inference engine can get along in modern society way more comfortably than someone who thinks he literally is a medieval knight, for example -- but it's still delusional.
Anyway... I started down this road thinking about machine-learning algorithms. Which are ALSO bias-implementing engines, and if we want to minimize the differential harm they do we need to select the biases they implement accordingly, just like with human minds, and its best not to mistake them for rational truth-preserving inference engines.
But I think I've distracted myself into thinking about rational inference-engine LARP.
Beep-boop.
Posts
No comments:
Post a Comment