content warning: Zack Davis. so of course this is merely the intro to Zack’s unquenchable outrage at Yudkowsky using the pronouns that someone wants to be called by
content warning: Zack Davis. so of course this is merely the intro to Zack’s unquenchable outrage at Yudkowsky using the pronouns that someone wants to be called by
Zack thought the Times had all the justification they needed (for a Gettier case) since he thought they 1) didn’t have a good justification but 2) also didn’t need a good justification. He was wrong about his second assumption (they did need a good justification), but also wrong about the first assumption (they did have a good justification), so they cancelled each other out, and his conclusion ‘they have all the justification they need’ is correct through epistemic luck.
The strongest possible argument supports the right conclusion. Yud thought he could just dream up the strongest arguments and didn’t need to consult the literature to reach the right conclusion. Dreaming up arguments is not going to give you the strongest arguments, while consulting the literature will. However, one of the weaker arguments he dreamt up just so happened to also support the right conclusion, so he got the right answer through epistemic luck.
Ooooh I get it for Yudkowsky now, I thought you were targeting something else in his comment
No no, not the term (my comment is about how he got his own term wrong), just his reasoning. If you make a lot of reasoning errors, but two faulty premises cancel each other out, and you write, say, 17000 words or sequences of hundreds of blog posts, then you’re going to stumble into the right conclusion from time to time. (It might be fun to model this mathematically, can you err your way into being unerring?, but unfortunately in reality-land the amount of premises an argument needs varies wildly)
If I had to pick a mathematical model, it’d be a drunken walk.
I suppose your use of this term “cancel each other out” confuses me, since as far as I can tell it isn’t the premises which cancel each other. I get that in your schematism Davis has got to the right conclusion by luck, through faulty reasoning, but I can’t see any way to make sense of this by pointing at the interaction between his faulty premises.
The way I see your working:
1). Davis thought that the NYT didn’t HAVE any justification for calling Alexander racist (F)
2). Davis thought that the NYT didn’t NEED any justification for calling Alexander racist (F)
C). Davis thought that the NYT had all the justification they needed for calling Alexander racist (T)
Now he would be “epistemically lucky” to have reached that conclusion regardless of how he got there (as long as it was the wrong way). In fact writing it out like that it’s clear that the the argument, as you’ve presented it anyway, is poorly constructed: one of the premises is completely irrelevant, because whether or not the NYT has justification is irrelevant if they don’t need any. The valid premise is (2), which is what would make his conclusion correct by his own lights, except that it’s an untrue premise.
(The role that premise (1) plays in Davis’s actual argument is different: it disputes the claim that the NYT had justification. But this is only good for what it’s intended to do: refutation of a counter-argument. But the refutation of a counter-argument is a separate component in the broader conclusion “The NYT is right AND for the wrong reasons AND the NYT’s supporters are wrong, insofar as they think the NYT is justified”)
Where premise (1) comes back in is that the NYT did have justification, or better, they “didn’t not” have justification. That they didn’t have justification is false. But the falsehood of this doesn’t GET Davis anywhere, what gets him to the conclusion is premise (2) (which is still false). Rather, had he believed the correct opposite of premise (1) (“the NYT did not fail to have justification”) he would have got to the true conclusion. But the “cancel[ing]” here doesn’t enhance Davis’s good luck, it enhances his epistemic misfortune: his premise would have been relevant had he believed its opposite.
Notably, premise (2) doesn’t play any functional role in the truth or falsity of premise (1), and it doesn’t play any functional role in the narrative of how we get to premise (1) (or the opposite of premise 1). So it cannot be due to any cancelling effect that the opposite of premise (1) turns out to be true. Rather, what it does do is deepen the intuitive sense that Davis got especially lucky to get to the right conclusion, given that his premises pointed so far in the opposite direction when the truth was so close by.
But with things laid out like that it seems that Davis didn’t really have much epistemic luck at all, and the intuitive sense is a misleading artefact of your reconstruction of his argument as involving some procedure by which his faulty premises cancelled one another out. The deepened intuition I just mentioned is false, because in fact the odds of Davis getting it right or wrong are not especially interesting - he just thinks, for completely independent reasons, that Scott Alexander is in fact a racist, and indeed, contra Davis, Davis is using much of the same evidentiary/rational base as the rest of us, including (his version of) the NYT. Moreover, since Davis is just getting to that conclusion by premises which turn out to be faulty (AND independent premises which are true), and there is no “cancelling” procedure, he’s epistemically lucky in an extremely ordinary way, and at the very least not in the way you’re describing (whereby some mathematically interesting procedure gets him to that true conclusion).
I had a long reply which i think made some errors of interpretation as to what you’re saying. I find this “cancels” language confusing, but I don’t have the energy to do any more in-depth clarification on this thing!