I completely agree on the substantive cases. But I see an issue for the idea of “reasonableness”. You say:
> But there’s a reason why blameworthiness norms track the target’s epistemic situation rather than objective consequences: an agent can only ever be guided by the former, so any attempt to inculcate reliable virtue via social incentives to conform to inaccessible facts is a non-starter.
Now there’s an issue about what exactly someone’s “epistemic situation” is and what exactly it means for facts to be “inaccessible”. You don’t mean this to be entirely internal, about what the person knows (Williamson notes that knowledge is partly external, but I think he doesn’t get at substantial enough external things). Your caveat about whether the person is being reasonable is meant to deal with cases where someone could have easily known something, but either willfully or carelessly failed to find out the truth.
It certainly makes sense to hold people responsible for things they could easily have known - especially if they plausibly *should* have known them - even if they don’t actually know these things, and the facts are therefore in that sense inaccessible.
This reasonableness condition really needs to encompass requiring people to do some relevant investigation and study when they are saying or doing something sufficiently important, not just believing the most reasonable thing to believe given the investigation and study they have in fact done. And I suspect that some disagreements about whether particular hostility is warranted involves disagreeement about what someone should have known in these cases.
Re that post accusing people who talk about positive use cases for AI as being guilty of narcissism (morally condemning them) it’s worth noting that it started with a quote from X (Grok, anyone?) and was posted on Meta’s Facebook. If this isn’t precisely an example of a performative contradiction, it’s awfully close to one.
It's reasonable to expect good faith. Most people aren't really interested in that level of good faith. Far too many interactions aren't really a quest for understanding or getting closer to the truth, or pushing the borders of knowledge.
A lot of the time it's about our Bayesian priors. We can't always get what you're recommending everywhere we can just create it in communities via norms. Since communities are imagined to norms are also going to be imaginary with their reliability depending on the force of consensus and consequence.
You've identified a real problem that is larger than EA skeptics.
I think your constraint doesn’t actually work in practice for one thing, reasonableness is not a binary thing. Most people are in a grey zone instead of being clearly reasonable or unreasonable, but in any case, unless you define it very strictly your constraint will constraint too much. For example, a meat industry lobbyist who very sincerely believes on the basis of reasonable philosophical argumentation that animals are not conscious would be protected by your constraint for that matter. It might even protect someone who has a reasonable theory for why autocracy is good and happens to be a politician, trying to enact this vision in general, whether or not, someone is reasonable, doesn’t seem to have much to do with whether we should be hostile to them, whether you should be hostile to an autocrat doesn’t seem to turn on whether they are political scientist with a good defence of their position. It’s entirely about the objective consequences of their actions as you judge them. You can maybe get out of this by defining reasonable, most strictly, for example by including substantive beliefs of yours that you think must be reasonable, but then it’s so strict as to be actively misleading as most people that would be generally considered reasonable would not qualify and I question the value of such a constraint in that case. I also question the value of your defence you point to things like whether they deserve blame, but that assume is a lot of controversial things like that just deserts is a thing or at least that we can coherently speak of what people deserve independent of the consequences of their action, all of that is very controversial and also a little self defeating as it would be basically constraining yourself not to be hostile to people who reasonably hold the view that you should be hostile on the basis of objective consequences, as one can certainly mount a reasonable defence of the position.
I'm very puzzled by your comment. Why would you need to be *hostile* towards someone who is woefully confused about animal consciousness or the comparative strengths of liberal democracy vs autocracy?
We can *oppose* people's goals and actions if we think that are bad ones, without having to be *hostile* to any confused-but-well-meaning people who are carrying them out. (As a wise U.S. President once said, "We can disagree without being disagreeable.")
I think we just have different pictures of what works to me, the fact that for example, people fighting would be dictators frequently resort to things like demonising, narratives, pressure campaigns, and generally not talking about how the would be dictator is really a reasonable man with good intentions, who just happens to have a few mistaken ideas is strong evidence that that’s probably not the Effective way to stop a dictator now to be fair. I suspect you disagree with me regarding how often reasonable would be dictators can be found, but still it just seems intuitively obvious that in the modern world of movement cannot fight a politicians attempt to become a dictator with maximum effectiveness. If they start making sure that everyone knows the dictator is really reasonable, and we shouldn’t be mean to them or attack him with snark and bombast. To be fair, you yourself could maintain such a posture while still fighting the dictator, but that’s because you’re doing it alone and on the margin I can buy that such strategies don’t help, but if literally everyone followed, your example, I suspect stopping would be dictators to pick the most extreme example from my list would be much harder. The fact that opponents of dictators don’t do this. Obviously might have some connection to the fact that generally the dictators are not reasonable, but I really sincerely doubt that the dictator being reasonable actually affects any other relevant facts in a way where it would alter what works, and I also just suspect that if what work was being more chill and gentlemanly in politics, then selection effects and incentives facing politicians would ensure that our politics would be more gentlemanly as there are strong incentives to do what works and politicians that don’t follow. Such incentives get removed from the process. Like I think a major departure from standard practice should be all else equal expected to generate major cost to effectiveness and therefore given that your constraint appears to outlaw standard operating procedure in a lot of domains, I suspect you constraint is too constraining and would not work outside of a particular subculture on the internet.
If people generally are stupid and unreasonable, then sure, there's no guarantee that reasonable people will win out. I'm not a deontologist, let alone an absolutist, so my constraint makes no claim to being exceptionless. If the stakes are high enough, and *if* breaking it on some occasion would actually help (I think people are motivated to exaggerate how often this is true, given how much they enjoy indulging in negative emoting / hating on out-groups), then sure. That doesn't stop it from being a pro tanto moral constraint that applies in *most* circumstances.
I think there are probably few things more valuable in the long run than increasing the share of the population who are actually intelligent and reasonable. Norms like the one I'm recommending seem pretty vital for increasing reasonableness in general. So that's another reason to think it's pretty important to promote. (But I won't hate on you for disagreeing!)
I think part of the difference here is that I don’t actually think the people are being stupid or unreasonable demonising narratives and like are useful coordination and signaling mechanisms. The reason abandoning them doesn’t work is because people given human psychology correctly. Guess that not being hostile to someone implies, you don’t think what they are doing is that bad and therefore they should not be worried so from my perspective it’s kind of like in a country where most of the people speak German, you are insisting we talk in English just because it’s a better language for the kind of communication you want like given the public correct views on human psychology. You recommended strategy with Signal a bunch of incorrect things for no particular reason. You and I may also just understand different things from hostility. I was imagining more attacking someone with ridicule or saying their evil and the like not, for example, blatantly making up ridiculous conspiracy theories about them. Although I grant you, I think sometimes that also works.
I like this idea. One concern: abuse of the civility embodied by this proposal. Isn’t this what’s going on with people “just asking questions”, who are a known time sink that in some cases have earned “hostility on sight” status? Of course, a rule that says “give people the benefit of the doubt except when they’re pissing you off” is not much of a rule so I don’t think I’ve added much here.
I completely agree on the substantive cases. But I see an issue for the idea of “reasonableness”. You say:
> But there’s a reason why blameworthiness norms track the target’s epistemic situation rather than objective consequences: an agent can only ever be guided by the former, so any attempt to inculcate reliable virtue via social incentives to conform to inaccessible facts is a non-starter.
Now there’s an issue about what exactly someone’s “epistemic situation” is and what exactly it means for facts to be “inaccessible”. You don’t mean this to be entirely internal, about what the person knows (Williamson notes that knowledge is partly external, but I think he doesn’t get at substantial enough external things). Your caveat about whether the person is being reasonable is meant to deal with cases where someone could have easily known something, but either willfully or carelessly failed to find out the truth.
It certainly makes sense to hold people responsible for things they could easily have known - especially if they plausibly *should* have known them - even if they don’t actually know these things, and the facts are therefore in that sense inaccessible.
This reasonableness condition really needs to encompass requiring people to do some relevant investigation and study when they are saying or doing something sufficiently important, not just believing the most reasonable thing to believe given the investigation and study they have in fact done. And I suspect that some disagreements about whether particular hostility is warranted involves disagreeement about what someone should have known in these cases.
Yeah, that sounds right.
Re that post accusing people who talk about positive use cases for AI as being guilty of narcissism (morally condemning them) it’s worth noting that it started with a quote from X (Grok, anyone?) and was posted on Meta’s Facebook. If this isn’t precisely an example of a performative contradiction, it’s awfully close to one.
"Of course, “Internet denizen angry, reasons poorly” is not exactly a newsworthy headline."
Perhaps, but this was still a very important topic to discuss. You challenge something deeply ingrained and calcified. Very important post. Thank you!
It's reasonable to expect good faith. Most people aren't really interested in that level of good faith. Far too many interactions aren't really a quest for understanding or getting closer to the truth, or pushing the borders of knowledge.
A lot of the time it's about our Bayesian priors. We can't always get what you're recommending everywhere we can just create it in communities via norms. Since communities are imagined to norms are also going to be imaginary with their reliability depending on the force of consensus and consequence.
You've identified a real problem that is larger than EA skeptics.
I think your constraint doesn’t actually work in practice for one thing, reasonableness is not a binary thing. Most people are in a grey zone instead of being clearly reasonable or unreasonable, but in any case, unless you define it very strictly your constraint will constraint too much. For example, a meat industry lobbyist who very sincerely believes on the basis of reasonable philosophical argumentation that animals are not conscious would be protected by your constraint for that matter. It might even protect someone who has a reasonable theory for why autocracy is good and happens to be a politician, trying to enact this vision in general, whether or not, someone is reasonable, doesn’t seem to have much to do with whether we should be hostile to them, whether you should be hostile to an autocrat doesn’t seem to turn on whether they are political scientist with a good defence of their position. It’s entirely about the objective consequences of their actions as you judge them. You can maybe get out of this by defining reasonable, most strictly, for example by including substantive beliefs of yours that you think must be reasonable, but then it’s so strict as to be actively misleading as most people that would be generally considered reasonable would not qualify and I question the value of such a constraint in that case. I also question the value of your defence you point to things like whether they deserve blame, but that assume is a lot of controversial things like that just deserts is a thing or at least that we can coherently speak of what people deserve independent of the consequences of their action, all of that is very controversial and also a little self defeating as it would be basically constraining yourself not to be hostile to people who reasonably hold the view that you should be hostile on the basis of objective consequences, as one can certainly mount a reasonable defence of the position.
I'm very puzzled by your comment. Why would you need to be *hostile* towards someone who is woefully confused about animal consciousness or the comparative strengths of liberal democracy vs autocracy?
We can *oppose* people's goals and actions if we think that are bad ones, without having to be *hostile* to any confused-but-well-meaning people who are carrying them out. (As a wise U.S. President once said, "We can disagree without being disagreeable.")
I think we just have different pictures of what works to me, the fact that for example, people fighting would be dictators frequently resort to things like demonising, narratives, pressure campaigns, and generally not talking about how the would be dictator is really a reasonable man with good intentions, who just happens to have a few mistaken ideas is strong evidence that that’s probably not the Effective way to stop a dictator now to be fair. I suspect you disagree with me regarding how often reasonable would be dictators can be found, but still it just seems intuitively obvious that in the modern world of movement cannot fight a politicians attempt to become a dictator with maximum effectiveness. If they start making sure that everyone knows the dictator is really reasonable, and we shouldn’t be mean to them or attack him with snark and bombast. To be fair, you yourself could maintain such a posture while still fighting the dictator, but that’s because you’re doing it alone and on the margin I can buy that such strategies don’t help, but if literally everyone followed, your example, I suspect stopping would be dictators to pick the most extreme example from my list would be much harder. The fact that opponents of dictators don’t do this. Obviously might have some connection to the fact that generally the dictators are not reasonable, but I really sincerely doubt that the dictator being reasonable actually affects any other relevant facts in a way where it would alter what works, and I also just suspect that if what work was being more chill and gentlemanly in politics, then selection effects and incentives facing politicians would ensure that our politics would be more gentlemanly as there are strong incentives to do what works and politicians that don’t follow. Such incentives get removed from the process. Like I think a major departure from standard practice should be all else equal expected to generate major cost to effectiveness and therefore given that your constraint appears to outlaw standard operating procedure in a lot of domains, I suspect you constraint is too constraining and would not work outside of a particular subculture on the internet.
If people generally are stupid and unreasonable, then sure, there's no guarantee that reasonable people will win out. I'm not a deontologist, let alone an absolutist, so my constraint makes no claim to being exceptionless. If the stakes are high enough, and *if* breaking it on some occasion would actually help (I think people are motivated to exaggerate how often this is true, given how much they enjoy indulging in negative emoting / hating on out-groups), then sure. That doesn't stop it from being a pro tanto moral constraint that applies in *most* circumstances.
I think there are probably few things more valuable in the long run than increasing the share of the population who are actually intelligent and reasonable. Norms like the one I'm recommending seem pretty vital for increasing reasonableness in general. So that's another reason to think it's pretty important to promote. (But I won't hate on you for disagreeing!)
LOL, I see what you did there at the end.
I think part of the difference here is that I don’t actually think the people are being stupid or unreasonable demonising narratives and like are useful coordination and signaling mechanisms. The reason abandoning them doesn’t work is because people given human psychology correctly. Guess that not being hostile to someone implies, you don’t think what they are doing is that bad and therefore they should not be worried so from my perspective it’s kind of like in a country where most of the people speak German, you are insisting we talk in English just because it’s a better language for the kind of communication you want like given the public correct views on human psychology. You recommended strategy with Signal a bunch of incorrect things for no particular reason. You and I may also just understand different things from hostility. I was imagining more attacking someone with ridicule or saying their evil and the like not, for example, blatantly making up ridiculous conspiracy theories about them. Although I grant you, I think sometimes that also works.
I like this idea. One concern: abuse of the civility embodied by this proposal. Isn’t this what’s going on with people “just asking questions”, who are a known time sink that in some cases have earned “hostility on sight” status? Of course, a rule that says “give people the benefit of the doubt except when they’re pissing you off” is not much of a rule so I don’t think I’ve added much here.