[…] Those 16,000 lives are unlikely to all be the same ones lost in an alternate world without robot cars. When we say autonomous cars can slash fatality rates by half, we really mean that they can save a net total of 16,000 lives a year: for example, saving 20,000 people but still being implicated in 4,000 new deaths. There’s something troubling about that, as is usually the case when there’s a sacrifice or “trading” of lives.”
—
- “The Ethics of Saving Lives With Autonomous Cars Is Far Murkier Than You Think”, Wired
sometimes I forget how objectively terrible the field of moral philosophy is
[clipped]
(via yxoque)
That article strikes me as a really poorly written version of this article. The argument as I understand it is that, while less people overall would die, it’s not just different people who would be killed, but a specific class of people would become more likely to be killed.
(via nextworldover)
Wow! It seems…. a little too generous to ascribe that stance to the Wired article, but that’s a much more interesting position and more worth discussing. […]
On one of his recent posts, Scott Alexander backed away a bit from steelmanning1, in response to the comment
James A. Donald (external link)
Please, no more steelmanning.
Person A says something that Person B finds strange and unreasonable, something that makes no sense given the rest of Person B’s worldview. So, Person B, “charitably” invents a somewhat similar position that makes sense given person B’s worldview Which worldview person A massively disagrees with, and is indeed violently insulted by.
Whenever I see the word, I assume, usually correctly, that you are about to demonize your opponents with some version of their argument that only makes sense, only is steel, given that everything that they believe is wrong, everything you believe is correct, and they are evil and stupid.
And even if you have the best of intentions, and I doubt your intentions are as pure as you think, you are going to wind up inadvertently doing that.
This comment seemed a bit off to me, and with this example I see why. If you work under a moral system where the correct answer to the trolley problem is “don’t kill some small number of people to save a larger number of people”2, then this does look like perverting an argument to some version that only makes sense if everything you believe3 is wrong.
On the other hand, for people who think the answer to the trolley problem is “do kill some small number of people to save a larger number of people” this steelmanned argument is important, interesting, novel, and potentially true where the previous argument seemed pretty clearly false4.
And I don’t think the point of steelmanning is to present someone else’s points the way that person would most like them presented. I think the point is to make the argument as strong as you can. And that’s very likely going to be by making it match your own belief framework5.
James A Donald is probably right that this doesn’t quite maximize charity. The person whose argument you’re rewriting will likely think it worse6. If you are valuing charity above all else7, this will likely fail. In most cases just republishing the other person’s argument will be more charitable.
But if you value being correct above all else8, I think steelmanning is a pretty good idea
He only mentioned it as a tangent to his main point, so I’m not sure how much this is a general shift in his approach and how much it was just about that one post. ↩
Which is presumably the Wired author’s moral system, judging by the article. ↩
or at least one thing you believe. ↩
and not in a particularly interesting way, either. ↩
Both because that’s probably where you’ll end up if you try to make the argument as strong as you can (since you believe what you believe), and also because each person has the most comparative advantage for strong argument writing under the “using own belief system” condition. ↩
This is not guaranteed; they might be persuaded by you or they might think you were better able than them to make the point they wanted to make. If one of these happens, everyone wins, and the rewrite will almost definitely be considered charitable by the involved parties. ↩
Valuing one thing above all else is usually a bad idea. ↩
See previous footnote. ↩
I really like these points on steelmanning. I guess generally I feel like there are two goals you might have in reading articles like this:
1) decide if there are good reasons to agree with the author’s point (i.e. self-driving cars are ethically problematic)
2) understand the author’s worldview (i.e. killing people in the trolley problem is bad)
These are both good goals and you have to read the article in different ways for each. Steelmanning is aimed at the first, not the second.
And the first is more practical to do in the short term. For instance, like 90% of what we read in my philosophy class is super deontological. If I had to read all of it as deontological, I guess there’s some small chance I’d come to understand their value system better and convert to deontology at some point. But in the meantime, I wouldn’t be able to believe anything they said, whereas by steelmanning them as consequentialist I can find some parts that make sense in my belief system.