I don't want this norm to be adopted. Sometimes someone has personal information that someone did something, but doesn't have sufficient evidence to convince others of this. This is very likely when sexual assault happens, for example. It's also common in cases of e.g. abuse by employers toward employees. Also, some judgments of "this thing is bad" are based on intuitive senses (e.g. aesthetics) that, while often truth-tracking, are difficult to explain to those who don't have the same intuitive sense.
In cases like this, it's important for people to be able to state "I have information that leads me to believe X, and my saying this (and giving the details I can) might or might not be sufficient to convince you". Perhaps others will, upon hearing this, have more relevant information to add, eventually creating common knowledge; and they will also likely have more correct beliefs (and be able to make better decisions) in the meantime, before common knowledge is created.
Ben Hoffman has written on problems with holding criticism to a higher standard than praise:
The problem comes when this standard is applied to critics but not to supporters of EA organizations. This is effectively a tax on internal criticism of EA. If you ask that we impose a higher burden on criticism than on praise for you or your organization, you are proposing that we forgo the benefits of an adversarial system, in order to avoid potentially damaging criticism. If we forgo the benefits of an adversarial system, we can only expect to come to the right answers if the parties that are presenting us with information exhibit an exceptionally honest intent to inform.
If you ask people to hold criticism of you to a higher standard than praise, you are either asserting a right to misinform, or implicitly promising to be honest enough that a balanced adversarial system is not necessary. You are promising to be a reliable, objective source of information, not just a clever arguer.
If you're asserting a right to misinform, then it is clear enough why people might not want to trust you.
So, the norm as stated seems more likely to serve the interest of "create a positive impression of what's going on, regardless of what's actually going on" (i.e. wirehead everyone; this is as scary as it sounds!), than the interest of "share information about what is going on in a way that can at some point lead to common knowledge being created, and the problems being solved".
This norm could be workable if you can distinguish sharing the information "I believe this person did this bad thing" from "I accuse this person of doing this bad thing" (with the first being a denotative statement, and the second being a speech act).
I think there should be a "tax on criticism", especially if it is such a cheap one as this - having to explain yourself if questioned is in a sense really just politeness! Default behavior rewards criticism too much, especially in a contrarian community like this one.
Note also that I have this norm in mind for public discourse, which I think some of the more extreme cases you brought up may not be.
So, this depends on what is meant by "explain yourself if questioned". If I'm allowed to say "my research aesthetics say this project is useless, and I can point to a couple details, but not enough to convince that many others", or, "I had some negative experiences in relation to this organization that lead me to believe that it's causing harm, and I can share a few details, but others are confidential" then, fine. But, such justification norms could (and probably should) naturally apply to praise as well.
I think both of those are basically fine. The thing that I'm more worried about is stuff like:
Alice: "This organization is unethical, you shouldn't support it"
Bob: "Why?"
Alice: *no response*
It's very understandable that one might not be able to share all their evidence or fully explicate everything, but I think that providing at least some information is important.
Sure, this seems reasonable. I am also worried about content-free praise, but, independent of that, content-free criticism seems good to discourage.
If this norm were to be followed, I think it should apply to endorsements / positive statements as well. Unjustified positivity and endorsement is more harmful when criticism is discouraged.
I think that being unreasonably negative is generally a lot worse than being unreasonably positive, at least in terms of its effect on other participants -- while both might be epistemically unsound, the consequences of a space being too negative are much more stifling to a community than the consequences of a space being too positive.
That said, I do think it would be good form to justify or retract unreasonably positive statements as well if challenged!
the consequences of a space being too negative are much more stifling to a community than the consequences of a space being too positive.
I don't agree with this. I've felt pretty silenced by people having high opinions of e.g. certain orgs and seeming actively uninterested in information indicating that such opinions are false. Which means it's harder for me to talk about what I actually think.
I anticipate much more negative social feedback for criticizing things people like, versus for praising things people don't like.
As far as I can tell, "well-calibrated" is actually optimal, and deviations from that are stifling, because they contribute to a sense that everyone is lying all the time, and you have to either join in the lies, stay quiet, or be a rebel.
I quite strongly disagree. This will inevitably lead the most competent and busy people to not share their assessments of anything, since they will be met with the expectation of having to justify every assessment in detail, which is simply not workable in terms of time. It also means there is no way for someone to register that they have a bad feeling about something without being able to make it fully explicit. This also runs into problems with secret information, embarrassing information and situations where someone does not feel safe with the current norms of public discourse.
I recognize and agree with the failure modes of default discourse that this is trying to fix, but I don't think this norm as described is a good idea.
This will inevitably lead the most competent and busy people to not share their assessments of anything, since they will be met with the expectation of having to justify every assessment in detail, which is simply not workable in terms of time.
See some of the nuance Davis gives in other comment threads on this post that I think address your concerns, but I think this is on net beneficial if we also cause them to not share their assessment either way without explanation. Just because someone is competent or busy doesn't make their assessments that much more useful unless they actually take the time to make a thoughtful assessment, and I expect the amount their gut reactions to things will be not much better than anyone else's, caveat being there is probably some appropriate discount function on which their gut assessments are marginally more useful but the discount rate is so high we might as well just ignore it. Put another way, being competent, busy, or an expert doesn't make you less likely to do the things humans tend to do that lead to low-value judgements (cf. literally everything written about decision under uncertainty in humans) outside of limited contexts with training and rapid feedback, so in most places at most times having this norm is useful.
I'll also point out this already is the norm in in-person conversation where the only way to express pleasure/displeasure with someone or what they are saying is to do so in a way that opens you up to being asked for justification and failure to provide justification can result in dramatic discounting of your position. Or so the norm seems to be to me.
I don't think you have to explain everything in excruciating detail, but I think if you're willing to attack someone you need to be willing to justify that attack if called on it. That does mean that someone might not get to share some criticisms, but in general I think the costs of losing some casual criticism is worth the benefits of improving the discourse overall.
I think registering a bad feeling about something is fine and doesn't need to have a bunch of time spent on it, as the level of justification needed for something like that is pretty low. I don't really even consider that an "accusation" per se, but I suppose you could conceptualize it as on a spectrum, where on one side you have "I don't really like this" and on the other you have extreme stuff like "allowing this is a moral atrocity and the creator is a war criminal". The latter obviously needs quite a lot of justification, while the former doesn't much.
Why the asymmetric burden on criticism? Seems like positive claims are more likely to imply demands on others' resources or attention.
Can you say more about why positive claims would imply demands on others' resources or attention?
Example: EA is built on strong positive claims implying a duty to contribute to a few specific projects.
That hasn't much been my experience of EA, but I think applying this standard to claims that one is obligated to contribute to something seems fine too.
The "drowning child" thought experiment is very explicitly about this, and one of the most common ways people enter the recruitment funnel. Slate Star Codex's Infinite Debt is about this as well.
Agreed, but also noting that drowned child doesn't seem like an instance of "praise that doesn't get backed up" (or praise, period)
"Givewell is great!" (or any random org is great!) felt more like the thing... but also feels less like an explicit demand on my resources. (I realize there's a thing where the X is Great! statements work in conjunction with the Child Pond arguments, but by then the model is getting a bit more complicated)
[hmm, re-reading, you're saying "positive claim" which is different from praise. But also, positive claim isn't the obvious counterpart to "criticism". Doesn't feel like part of the same schema to me)
I do want to thank for Davis for making this discourse-norm bid explicitly. I think there's a failure mode where people don't notice that they are trying to enforce mutually exclusive norms, and getting frustrated when people seem to be defecting.
(this is neither an endorsement or disendorsement of the norm in question)
in general norms online often trend too much in the direction of low-content mockery rather than reasoned debate.
Do you think this is currently a problem on LessWrong.com? I have a hard time bringing up examples of cases where someone made public online accusations against other people that they weren't willing to justify.
I think LW 1.0 absolutely had a big problem with low-quality criticism, and LW 2.0 doesn't feel "fully realized" yet so it's hard to say. But I think low-quality criticism and an IMO bad response to it basically drove the best active user in the early days of LW 2.0 off the site (Duncan_Sabien), so it absolutely strikes me as still a relevant concern.
While there was low quality criticism on LW 1.0 I don't think it was a place where anybody who was asked to justify their criticism wouldn't do so.
It doesn't feel to me like failure to justify criticism when there was a demand was an issue, posters were generally willing to double-down on low quality criticism.
For in-person conversations (I know this was meant as a norm for public discourse): Personally I tend to have a hard time digging into my memories for "data points" when I have a negative or positive impression of some person. It's kind of the same thing with people asking you "What have you been working on the past week?" – I basically never remember anything immediately (even though I do work on stuff). This creates asymmetric incentives where it's easier to make negative judgments seem unjustified or at least costly to bring up, which can contribute to a culture where justified critical opinions almost never reach enough of a consensus to change something. I definitely think there should be norms similar to the one described in the post, but I also think that there are situations (e.g., if a person has a reliable track record or if they promise to write a paragraph with some bullet points later on once they had time to introspect) were the norm should be less strict than "back the judgment up immediately or retract it." And okay, probably one can manage to say a few words even on the spot because introspection is not that slow and opaque, but my point is simply that "This sounds unconvincing" is just as cheap a thing to say as cheap criticism, and the balance should be somewhere in between. So maybe instead of "justify" the norm should say something like "gesture at the type of reasons," and that should be the bare minimum and more transparency is often preferable. (Another point is that introspecting on intuitive judgments helps refine them, so that's something that people should do occasionally even if they aren't being put on the spot to back something up.)
Needless to say, lax norms around this can be terrible in social environments where some people tend to talk too negatively about others and where the charitable voices are less frequent, so I think it's one of those things where the same type of advice can sometimes be really good, and other times can be absolutely terrible.
I would think of this in terms of rights. Who has the right to post a new theory? Who has the right to challenge an existing concept? Who has the right to reply? Who has the right to defence?Who has the right to demand?
Everyone can choose which ones you want to and which ones you don't want to, but it's not possible to bind other people to your preferences against their will.
One discourse norm that I think is really important is that of having to either support or retract accusations if challenged. If you say something negative about another person, their work, etc. and they ask you to explain yourself, I believe you are compelled to either justify or retract your statement. This creates a strong barrier against unjustified attacks and gossip, while still allowing justified criticism.
Here are some examples of what this norm might look like in action:
1. Alice posts about her thoughts on an issue; Bob, who dislikes Alice, responds with snarky insults about Alice's motivations. Alice asks Bob to explain his accusations, and he doesn't do so or replies with more insults. Moderation intervenes against Bob.
2. Carol posts a brief comment saying a project is incompetent. Darryl replies asking her to provide more detail or retract. Carol links to a post that explains her critique in more detail.
3. Efren posts a statement criticizing an event that will soon be held. Faye asks Efren to back up his criticisms. Efren decides that his claim was actually more emotional and less grounded than he first thought, so he decides to retract his original statement.
Now, someone might ask "why try to make it more difficult to be critical of something?" The answer is that making fun of things is easy [1], and in general norms online often trend too much in the direction of low-content mockery rather than reasoned debate. Holding norms that require people to back up or retract controversial statements can be a good step away from that failure mode.
[1] Full disclosure: I wrote the linked post under my old username.