It seems to me like the lesson to learn from that is having regulations that prevent building a facade with material that burns is reasonable.
Well this nerd sniped me.
So one thing worth saying is, I think no one here was suggesting that those regulations are unreasonable. And I can give at least two reasons they seem like better regulations than bedrooms-must-have-windows:
As a prospective resident of an apartment, I can’t easily evaluate whether the building’s cladding is flammable. I can easily evaluate whether its bedrooms have windows.
Preventing “fire” is different from preventing “dying in a fire, given that there’s a fire”. I think the costs of the second are more internalized, so I’d expect market solutions for the second are more efficient. (And I guess an external fire has more externalities, being more likely to spread from one building to another.)
But also, my immediate reaction is: “you observe X and Y together, and you think this is evidence that X makes Y less likely?”
(ETA: A reason not to trust that intuition: by conservation of expected evidence, every day previously where we observed X and not-Y, we should take this as evidence that X makes Y less likely. i.e. that regulations prevent fire, even if they aren’t followed. This is, uh, suspicious.)
The slightly less naive version of this reaction is something like, regulations in this case did not cause fireproof cladding. So this is evidence that regulations don’t in general cause fireproof cladding, and thus don’t prevent fires. But it’s still costly to have the regulations (someone has to write them, someone probably has to pretend to enforce them).
But on the other hand, suppose we have reason to think that regulations usually are followed. Then we observe the combination of “fire” plus “no fireproof cladding”, which is evidence that fireproof cladding prevents fire. (Assuming no one was doing a rain dance at the time, it is also evidence that rain dances prevent fire.)
So I currently think the question is roughly: “does this tell us more about whether regulations cause fireproof cladding, or about whether fireproof cladding prevents fire?”
My guess is that in most cases like this, we would start with lower uncertainty about the second thing, and so get a bigger update on the first? And in this this specific case I think that would be correct; my understanding is that following the fire, it turned out there were a lot of buildings in London with cladding that wasn’t up to code.
But if we have low uncertainty about the number of buildings with fireproof cladding; and high uncertainty about its effectiveness; then maybe it would go the other way. E.g. we could have a model like: “I don’t know how likely fires are in general, and I don’t know whether fireproof cladding makes them less likely. But I’m confident that around 90% of buildings have fireproof cladding. Then, if fireproof cladding doesn’t help, a fire that occurs is 10% likely to occur in a non-clad building. If it does help, it’s more likely than that. Since it occured in a non-clad building, that’s evidence that cladding helps”. (If I’m confident that only 10% of buildings are clad, instead of higher-than-10% versus 10% we get higher-than-90% versus 90%, and the update on does-cladding-help is smaller.)
Writing a lot of things into the code because you expect people not to follow the code in all cases likely leads to people thinking that the code overregulates everything and thus isn’t really worth following when you can avoid it.
There’s also the key question of how much you want to pay for one human life via regulations. Many regulatory agencies have standards that require them to prove cost-effectiveness for regulations. They have to justify that they can protect human lives for a given price. Often the price of human life is in the order of 10 million $.
The International Residential Code is made by an NGO and not a government agency and thus I don’t think they have to provide any justifications for cost-effectiveness.
The right step might be to move the authority for a code to a federal agency that has to justify its safety decisions on the basis of cost-effectiveness and that can be sued if they overreach.
Am I right in thinking that this comment is like, inspired by what I wrote but not particularly actually a reply to it?
Like I don’t think I suggested that one should write a lot of things into regulatory codes expecting people not to follow them; or that anyone is doing that; or even really brought up the idea.
So if you think I did bring up that idea then I’m bemused. But if you just read my comment and were like “oh that makes me think about the idea of writing a lot of things etc.”… well, that’s fine, but fwiw I find it more pleasant when such things are noted explicitly. As it is I find myself wondering if you’re trying to disagree with something I wrote, or agree, or tell me something you think is useful and relevant to me, or what.
(This also applies to your previous comment, where e.g. “this suggests cladding regulations are good” does not appear to be a reply to anything in thread. And fwiw I feel like I have this reaction to a lot of your comments.)
Well this nerd sniped me.
So one thing worth saying is, I think no one here was suggesting that those regulations are unreasonable. And I can give at least two reasons they seem like better regulations than bedrooms-must-have-windows:
As a prospective resident of an apartment, I can’t easily evaluate whether the building’s cladding is flammable. I can easily evaluate whether its bedrooms have windows.
Preventing “fire” is different from preventing “dying in a fire, given that there’s a fire”. I think the costs of the second are more internalized, so I’d expect market solutions for the second are more efficient. (And I guess an external fire has more externalities, being more likely to spread from one building to another.)
But also, my immediate reaction is: “you observe X and Y together, and you think this is evidence that X makes Y less likely?”
(ETA: A reason not to trust that intuition: by conservation of expected evidence, every day previously where we observed X and not-Y, we should take this as evidence that X makes Y less likely. i.e. that regulations prevent fire, even if they aren’t followed. This is, uh, suspicious.)
The slightly less naive version of this reaction is something like, regulations in this case did not cause fireproof cladding. So this is evidence that regulations don’t in general cause fireproof cladding, and thus don’t prevent fires. But it’s still costly to have the regulations (someone has to write them, someone probably has to pretend to enforce them).
But on the other hand, suppose we have reason to think that regulations usually are followed. Then we observe the combination of “fire” plus “no fireproof cladding”, which is evidence that fireproof cladding prevents fire. (Assuming no one was doing a rain dance at the time, it is also evidence that rain dances prevent fire.)
So I currently think the question is roughly: “does this tell us more about whether regulations cause fireproof cladding, or about whether fireproof cladding prevents fire?”
My guess is that in most cases like this, we would start with lower uncertainty about the second thing, and so get a bigger update on the first? And in this this specific case I think that would be correct; my understanding is that following the fire, it turned out there were a lot of buildings in London with cladding that wasn’t up to code.
But if we have low uncertainty about the number of buildings with fireproof cladding; and high uncertainty about its effectiveness; then maybe it would go the other way. E.g. we could have a model like: “I don’t know how likely fires are in general, and I don’t know whether fireproof cladding makes them less likely. But I’m confident that around 90% of buildings have fireproof cladding. Then, if fireproof cladding doesn’t help, a fire that occurs is 10% likely to occur in a non-clad building. If it does help, it’s more likely than that. Since it occured in a non-clad building, that’s evidence that cladding helps”. (If I’m confident that only 10% of buildings are clad, instead of higher-than-10% versus 10% we get higher-than-90% versus 90%, and the update on does-cladding-help is smaller.)
Writing a lot of things into the code because you expect people not to follow the code in all cases likely leads to people thinking that the code overregulates everything and thus isn’t really worth following when you can avoid it.
There’s also the key question of how much you want to pay for one human life via regulations. Many regulatory agencies have standards that require them to prove cost-effectiveness for regulations. They have to justify that they can protect human lives for a given price. Often the price of human life is in the order of 10 million $.
The International Residential Code is made by an NGO and not a government agency and thus I don’t think they have to provide any justifications for cost-effectiveness.
The right step might be to move the authority for a code to a federal agency that has to justify its safety decisions on the basis of cost-effectiveness and that can be sued if they overreach.
Am I right in thinking that this comment is like, inspired by what I wrote but not particularly actually a reply to it?
Like I don’t think I suggested that one should write a lot of things into regulatory codes expecting people not to follow them; or that anyone is doing that; or even really brought up the idea.
So if you think I did bring up that idea then I’m bemused. But if you just read my comment and were like “oh that makes me think about the idea of writing a lot of things etc.”… well, that’s fine, but fwiw I find it more pleasant when such things are noted explicitly. As it is I find myself wondering if you’re trying to disagree with something I wrote, or agree, or tell me something you think is useful and relevant to me, or what.
(This also applies to your previous comment, where e.g. “this suggests cladding regulations are good” does not appear to be a reply to anything in thread. And fwiw I feel like I have this reaction to a lot of your comments.)