Is the Wave non-disparagement thingy okay?

A few weeks ago, Linch, Auckland (pseudonym), and I (Ruby) decided to chat about something and settled on this topic. What follows is a mix of distillation and edited chat (Slack) transcript.

Auckland and I (Ruby) start from the position that the Wave non-disparagement agreement is probably bad. Linch expresses uncertainty about whether, if broader society (or at least your local reference class) makes use of non-disparagement agreements, that makes it okay to use them yourself.

We discuss various possible examples of not-good behavior (non-consent, lying, bribing, saying “awesome” when you mean “mildly positive”) that people empirically engage in and/​or seem more okay if everyone is doing them. We think about what “reference class” Wave is in, e.g. more general Silicon Valley vs EA.

A big is crux is whether Wave had been prompted to reflect on their non-disparagement agreement, and if so, their reaction. It might be that non-disparagements are bad, but you have to be paying attention to notice when you should be morally better than your surrounding society. Some employees wanting to negotiate out of the non-disparagement agreement seems like it was a prompt to reflect on it. Also how much the founders of Wave say “oops” when it’s pointed it out, vs defending the practice.

Beyond that, there’s a question of even if someone is not particularly blameworthy for engaging in a standard widespread not-good behavior, do you still want them in a leadership position? E.g. board of Effective Ventures/​CEA? As opposed to wanting leaders who’d have the discernment to catch this kind of thing and not do it. We have some more discussion about who you want for EA leadership related to how good you think EA is.

Linch raises that non-disparagement agreements are in fact not legally enforceable. We all agree that it’s plausible EAs should push Wave to make an official declaration to void their non-disparagement agreements.

Background Context

In the discussion of Nonlinear’s behavior, the topic of NDAs/​other agreements that restrict employees speaking negatively of former employer’s came up.

jefftk (permalink):

an NDA to prevent people from publicly criticizing their former workplace seems line-crossing to me.

I don’t like these, but they are (were) depressingly common. I know at least one org that’s generally well regarded by EAs that used them.

habryka (permalink):

Oh, wow, please tell me the name of that organization. That seems very important to model, and I would definitely relate very differently to any organization that routinely does this (as well as likely advocate for that organization to no longer be well-regarded).

licolnquirk (permalink):

Jeff is talking about Wave. We use a standard form of non-disclosure and non-disparagement clauses in our severance agreements: when we fire or lay someone off, getting severance money is gated on not saying bad things about the company. We tend to be fairly generous with our severance, so people in this situation usually prefer to sign and agree. I think this has successfully prevented (unfair) bad things from being said about us in a few cases, but I am reading this thread and it does make me think about whether some changes should be made.

I also would re-emphasize something Jeff said—that these things are quite common—if you just google for severance package standard terms, you’ll find non-disparagement clauses in them. As far as I am aware, we don’t ask current employees or employees who are quitting without severance to not talk about their experience at Wave.

Further conversation followed in that thread.

Transcript

This transcript is cleaned up, flattened (from Slack threads), and has less-interesting or well-fitting parts of the conversation removed, had a few bits edited after the fact, and finally ported into the LessWrong Dialogue format.

My quick take:

  • Lincoln is probably right that this is fairly standard in SV

  • It is in fact a pretty terrible practice that filters info and distorts people’s judgment

  • I think it’s a case where you need to be more moral than those around you, and that can be hard to notice, so I’m not sure how judge-y to be

  • Does make a difference how much people defend it once pointed out vs go “oh, obviously we inherited a terrible practice from broader society”

Ruby

At a high level, I view this as a clash between rationalist-y norms and the norms of a different subculture/​subsection of society.

Linch

It makes a difference to me how much people defend it once it’s pointed out vs go “oh, obviously we inherited a terrible practice from broader society”.

This is the thing I care most about.

I am more than happy to tolerate a lot of minor flubs like this in the name of moving fast and not thinking too hard about small details (though I think it would have been a positive sign if the founders had a more visceral “I will never let this happen in my company” reaction).

I also feel very hesitant to critique it, because as far as I know, Wave has been really successful in many ways and I haven’t, so I’m inclined to defer to them a lot.

Auckland

(important consideration to me is did anyone in the past say to them “uh, hey, this has bad effects on community epistemics”, if Elizabeth got to negotiate out of it, maybe counts as a prompt for reflection on that, a bit cruxy for me since it was a clear prompt to reflect on whether this was good or not)

Ruby

“Everybody else does it”

The thing I’m confused about is that conditional upon this being the thing that broader society does is it a bad thing to do it yourself

An example of something that I consider bad to do even if other people all do it is questionable sexual consent practices in the past, and some cultures today.

An example of something that I consider a bad general practice but fine for people to participate in if everybody else does so is bribing police officers at traffic stops.

I think some norms are better than others in the sense they lead to better long-run equilibria; ask culture is a good example;

Lying on apps [applications] is also bad (even though it’s common in broader society) and I would negatively judge someone who lies on EA apps, but I would negatively judge someone who lies on EA funding applications much more if they’re familiar with the culture, than if they aren’t.

But I’m not sure how much I’d judge someone who lies on non-EA funding applications. I feel like the answer is, like, moderately?

Linch

Some applications expect you to at least exaggerate if not lie, and they adjust for it. It seems like the standard practice for writing a letter of recommendation for colleges is to stretch the truth—if you mention any small flaw in the letter or don’t praise the applicant profusely enough, then the university will jump to the conclusion that the applicant is terrible.

Perhaps there’s an analog for past employees talking about their employer, but it feels much weaker.

Auckland

Sorry, why? Suppose we live in a world where 99 startups have non-disparagement agreements with severance packages and it’s well-understood the only reasons someone doesn’t sign a non-disparagement agreement are because either a) they hate the company so much they aren’t willing to do so or b) the company offers shitty severance.

As the 100th company, it sure feels like you’re playing “cooperate” when everybody else plays “defect”.

Linch

Agree with that. But EA is ostensibly a little cooperate bubble, where we built it out of people who’d play cooperate, and it’s important than we punish defection to maintain that.

Ruby

Importantly, I don’t know if Wave views their natural reference class as EA vs other startups, until recently I would’ve thought “other startups” is correct. “Natural reference class” might not be exactly the right word, but who do their employees talk to, what is the non-disparagement agreement meant to protect them from.

Linch

Relevant to that is where they hire from and who they get funding from. My sense is they hired quite a few EAs?

Ruby

They also hire hundreds of people who are very far outside the community, like folks on the ground in Sierra Leone and Senegal to do things like sales and customer support and exchanging money, yes?

Auckland

<we discuss the hiring there, how Sierra Leone norms might differ and affect the non-disparagement agreement, and whether the non-disparagement agreement was targeted at our would apply to those employees, before we return the question of just assuming Wave is a US company.>


jumping back


Cruxes

After some further discussion, we stop to check in on cruxes

Please take 3-minutes to think about where you’re at with the top-level question of ok or not, and your current cruxes. Share at 22:23

I think I want to separate out:

  1. How bad is it for Wave to have that agreement, and

  2. How blameworthy/​betrayal is their in Wave doing it?

Regard Point 1, I think really quite bad for the reasons of distorting recommendations people give (part of Oli’s complaint). I think this point is invariant to what broader society does. Regarding Point 2, unclear. I think depends how many times they got push back on it and they stuff to it. Like did something prompt them to reconsider societal default?

Ruby

Yes, I am interested in the above two questions and also maybe in “should he be on the board” in addition. Though that one feels perhaps drama-laden.

Auckland

Drama-y but gets at important stuff about judgment. Where “you didn’t act especially poorly given circumstances” might be true, but you also didn’t display enough good judgment for me to want to give you power (especially in something heavily political).

Ruby

I can see how one might want people who are good at not letting things like this happen on the EV board.

Auckland

My cruxes:

  1. How much Wave wants (and in practice, is) interfacing with the EA ecosystem?

    1. Eg, what percentage of their talent-adjusted US employees are EAs.

  2. How common are non-disparagement clauses in Silicon Valley, really?

  3. How much thought did they have about this specific question

    1. especially wrt to this ecosystem, plausibly thinking about it in general will give different answers.

  4. How naive am I being about powerful people?

    1. Like how in distribution is the badness of non-disparagement agreements for people with Lincoln’s level of success, even among say the top 20% most ethical people?

    2. This also came up a few times when I’m thinking about other cases that I looked at when I think someone is suss.

    3. if powerful people usually do at least a few suss things, we should be careful not to punish them when they’re actually being unusually honest.

Linch

I’m not sure what claim these are cruxes for. But I think only 3 is a crux for me, out of those. (though I have my own long list of cruxes). I am interested in 4 though, although I don’t know if I get what you’re saying.

Auckland

Let me reframe. If 99% of successful startup founders are more suss than Lincoln, not wanting to interface with Lincoln means that you either (a) aren’t working with startup founders or (b) are creating strong selection effects for deceptive alignment

Linch

Not wanting to work with them feels strong. Again I am not clear on exactly what claim your points are a crux for. Is the claim “this is sufficient evidence to never work for or with Wave under any but the most dire circumstances”? I was imagining it isn’t? (I would clearly not refuse to work with Wave over something like this.)

Auckland

Who do you want for EA Leadership?

Ruby

From my perspective, it’s like, EAs have some leaders. Some of them sure seem mediocre. Now we have Lincoln, who is plausibly a new leader. Whether we want to give him more or less powers depends a lot not only on how ethical he is but on how ethical his competition is. (both existing leaders and other candidates for new ones)

Linch

Kind of. At some point you decide everyone is so corrupt you give up and do something else (ofc you can’t always just go do your own thing, as Habryka complains)

Ruby

I agree.

Linch

(perhaps we should talk about whether EA is bad for the world now?) [one of the topics we brainstormed talking about)

Ruby

hmm. I feel more into the narrowly-scoped topic, especially if we want content that’s half-publishable. I have more to say on Wave personally.

Auckland

I suspect if we talk for a couple of min, we can get some more scoped claims about EA.

Linch, can you give a couple of sentences for why “is EA bad for the world” was a thing you listed as possible topic of discussion tonight?

Ruby
  • Sure seems like longtermist EA, or precursors to it, accelerated capabilities

  • Sure seems like there are other negative effects (FTX, cults)

  • Sure seems like EA is taking up a bunch of bright/​agentic/​moral (?) etc people whose counterfactuals are nontrivial

  • It’d suck for me to work most of my adult life on something that’s bad for the world

Linch

You say accelerated like it happened in the past. (I have to tow the Lightcone party line here, you see ;) )

Ruby

yes, a point of contention for me is how much the die is cast. so to speak, re: capabilities

Linch

I’d be interested in talking about that, though perhaps it is a tangent. But I feel super interested in the question. Also if it was cast, who cast it?

Auckland

https://​​en.wikipedia.org/​​wiki/​​Die_(integrated_circuit)

Ruby

I think it might be good to create more clarity about who is responsible for what here, especially because people seem to point fingers a lot (in a way that I worry is a bit tribal).

Auckland

Does seem like a good topic, though maybe for another round

Ruby

one interesting take I feel like people outside this cluster seem to have but seem to have relatively little representation in LC-adjacent cluster is something like:

the more you think EA/​LT EA sucks, the more you should want other leaders to take over

Linch

I don’t think I buy that. For one thing, you might not think most of why it sucks is that the leadership is bad. For another thing, you might be worried about adding variance to a thing like EA.

Auckland

sure there are extenuating factors

Linch

Man, something just feels very icky by the time you’re getting adversarial like that with the system/​EA

Linch, Auckland: agree react

Ruby

Also the discussion gets accusatory and political quickly if you start advocating for specific new leadership.

Auckland

Something just feels very icky by the time you’re getting adversarial like that with the system/​EA

A more positive framing is that in areas where we know our shit, external expertise is less helpful. In areas where we don’t (plausibly leadership is one of them), you want more external expertise.

Linch

Wait that sounds like a different claim, not a reframing?

Oh is it a reframing of your thing?

Auckland

Yeah, sorry

Linch

I think management, maybe, but I don’t think there’s external to EA leadership that can be brought in and do a better job

Pareto frontiers and shit

Ruby

You also might want to throw your hands up and declare “we need leadership who understands EA. Those people are sadly not excellent leaders, but the counterfactual of a person who doesn’t understand EA isn’t acceptable”.

Ruby, Linch: thumbs up

Auckland

I agree that the case isn’t ironclad. I’m also not sure I believe it on net. I just think this is the more obvious causal direction.

Linch

<some discussion of need to wrap up, briefly return previous topic

More on opting in to lying + are non-disparagement agreements legally enforceable?

Auckland, rephrase your opt-in question. What’s the question?

Ruby

Do you think the point about Harvard or grantmakers or whatever group being able to change the incentive landscape and expectations they set to prevent people from exaggerating while the general public can’t do anything about the non disparagement agreement situation holds water for you?

Are there scenarios that feel less like the Harvard or grant application thing which might be compelling?

Auckland

I think directionally it does, I don’t have a sense of the magnitude of the effect.

Linch

I’m curious about the model for why UK is less exaggeration-y than US. That’s interesting [note: some relevant context was edited out here]

Ruby

I don’t have great models of the situation in the UK or what causes the difference. (I’m not even that confident the difference is real.)

… (skip ahead)

Auckland

In some cases people very clearly opt in into lying to each other, like kinks
or social deception games, etc

Linch

Yes, but I am curious for more examples?

I think kinks, theater, etc are fine.
Also it feels like it’s easy-ish for Wave to make it legible they aren’t allowing their employees to speak out so people can recalibrate

It’s one thing to lie about your height on Tinder; it’s another to prevent your doctor from sharing independent measures of your height publicly.

Auckland

Note that the employees aren’t prevented from talking by the state directly
They willingly signed a contract not to talk

I think it’s plausible EAs should try to push Wave to make an official declaration to void their non-disparagement agreements

Auckland, Ruby: agree react

Linch

They aren’t legally enforceable anyway. So the main effect is allowing people who are conscientious about promises (like Jeff and Elizabeth presumably are) to be willing to speak up.

Linch

They aren’t legally enforceable anyway

This is interesting to me. Say more?

Auckland

https://​​www.axios.com/​​2023/​​03/​​27/​​labor-board-says-non-disparagement-clauses-are-unlawful

labor board ruled non-disparagement agreements aren’t lawful in the US, the ruling was clear that it applies retroactively.

Linch

Man, not even allowed to mention the agreement. Jesus.

Ruby

though iirc the Leverage thing was definitely not legally enforceable but people felt bound to it enough to not speak up anyway

Linch

That post you linked is recent, 2023. Was that true back in the years of relevance?

Ruby

no, but it’s relevant for getting people to speak up now.

Linch

So why wasn’t the Leverage case not legally enforceable?

Also I think people are just not that inclined to research what is or isn’t enforceable, they just assume

Ruby

iirc the document itself said it wasn’t. also the people signed under duress, which makes it easier to void

https://​​medium.com/​​@zoecurzi/​​my-experience-with-leverage-research-17e96a8e540b

Geoff had everyone sign an unofficial NDA upon leaving agreeing not to talk badly about Leverage, and agreeing that he would narrativize the whole place in a way that was favorable to all of us. I am the only person from Leverage who did not sign this, according to Geoff who asked me at least three times to do so, mentioning each time that everyone else had (which read to me like an attempt to pressure me into signing).

I believe Geoff knows the value of an oath, a name on paper, a psychologically binding agreement. Most of the people I talk to want distance from the whole thing, and some are downright paranoid...

I read that section as “clearly not legally enforceable or even trying to be)

Linch

going over time

I’ll say some wrap-up words...Thanks for doing this, has been really quite good (no exaggeration)


I’ll follow up with a processed chat log before publishing, and also have some questions about your experience of the process, if that’s alright

Ruby