Thank you, Lila, for your openness on explaining your reasons for leaving EA. It's good to hear legitimate reasons why someone might leave the community. It's certainly better than the outsider anti-EA arguments that do tend to misrepresent EA too often. I hope that other insiders who leave the movement will also be kind enough to share their reasoning, as you have here.

While I recognize that Lila does not want to participate in a debate, I nevertheless would like to contribute an alternate perspective for the benefit of other readers.

Like Lila, I am a moral anti-realist. Yet while she has left the movement largely for this reason, I still identify strongly with the EA movement.

This is because I do not feel that utilitarianism is required to prop up as many of EA's ideas as Lila does. For example, non-consequentialist moral realists can still use expected value to try and maximize good done without thinking that the maximization itself is the ultimate source of that good. Presumably if you think lying is bad, then refraining from lying twice may be better than refraining from lying just once.

I agree with Lila that many EAs act too glib about deaths from violence being no worse than deaths from non-violence. But to the extent that this is true, we can just weight these differently. For example, Lila rightly points out that "violence causes psychological trauma and other harms, which must be accounted for in a utilitarian framework". EAs should definitely take into account these extra considerations about violence.

But the main difference between myself and Lila here is that when she sees EAs not taking things like this into consideration, she takes that as an argument against EA; against utilitarianism; against expected value. Whereas I take it as an improper expected value estimate that doesn't take into account all of the facts. For me, this is not an argument against EA, nor even an argument against expected value -- it's an argument for why we need to be careful about taking into account as many considerations as possible when constructing expected value estimates.

As a moral anti-realist, I have to figure out how to act not by discovering rules of morality, but by deciding on what should be valued. If I wanted, I suppose I could just choose to go with whatever felt intuitively correct, but evolution is messy, and I trust a system of logic and consistency more than any intuitions that evolution has forced upon me. While I still use my intuitions because they make me feel good, when my intuitions clash with expected value estimates, I feel much more comfortable going with the EV estimates. I do not agree with everything individual EAs say, but I largely agree with the basic ideas behind EA arguments.

There are all sorts of moral anti-realists. Almost by definition, it's difficult to predict what any given moral anti-realist would value. I endorse moral anti-realism, and I just want to emphasize that EAs can become moral anti-realist without leaving the EA movement.

Hi Lila,

Since you've indicated you're not interested in a debate, please don't feel that this is directed at you. But as a general point:

I think extreme ideas get disproportionate attention compared to the amount of action people actually take on them. EAs are a lot more likely than other people to consider whether invertebrates matter, and to have thought about Pascal's-muggging type situations, but I think mainstream EAs remain pretty unsure about these. In general I think EAs appear a lot weirder because they're willing to carefully think through and discuss weird ideas and thought experiments, even if in the end they're not persuaded.

Ideas like "being willing to cause harm to one person to benefit others" sound bad and weird until you consider that people do this all the time. We have fire departments even though we know some fire fighters will die in the line of duty. Emergency room staff triage patients, leaving some to die in order to save others. I wash my toddler's hair even though she dislikes it, because the smell will bother people if I don't. It's hard to imagine how societies would work if they weren't willing to do things like these.

I think EA has work to do on making clear that we're not a philosophical monolith, and emphasizing our commonalities with other value systems.

Like many EAs, I'm a moral anti-realist. This is why I find it frustrating that EAs act as if utilitarianism is self-evident and would be the natural conclusion of any rational person. (I used to be guilty of this.)

You can be a moral realist and be very skeptical of anyone who is confident in a moral system, and you can be an anti-realist and be really confident in a moral system. The metaethical question of realism can affect the normative question over moral theories, but it doesn't directly tell us whether to be confident or not.

Anti realists reject the claim that any moral propositions are true. So they don't think there is a fact of the matter about what we morally ought to do. But this doesn't mean they believe that anyone's moral opinion is equally valid. The anti realist can believe that our moral talk is not grounded in facts while also believing that we should follow a particular moral system.

Finally, it seems to me that utilitarians in EA seem to have arguments for it which are at least as well grounded as people with any other moral view in any other community, with the exception of actual moral philosophers. I, for instance, don't think utilitarianism is self-evident. But I think that debunking arguments against moral intuitions are very good and that subjective normativity is the closest pointer to objective normativity that we have, that this implies that we ought to give equal respect to subjective normativity experienced by others, that the von Neumann-Morgenstein axioms of decisionmaking are valuable for a moral theory and point us towards maximizing expected value, and that there is no reason to be risk averse. I think a lot of utilitarians in EA would say something vaguely like this, and the fact that they don't do so explicitly is no proof that they have no justification whatsoever or respect for opposing views.

My view is that morality is largely the product of the whims of history, culture, and psychology.

Empirically, yes, this happens to be the case, but realists don't disagree with that. (Science is also the product of the whims of history, culture, and psychology.) They disagree over whether these products of history, culture, and psychology can be justified as true or not.

Any attempt to systematize such complex belief systems will necessarily lead to unwanted conclusions.

Plenty of realists have had this view. And you could be an anti-realist who believes in systematizing complex belief systems as well - it's not clear to me why you can't be both.

So I'm just not sure that your reasoning for your normative ideas is valid, because you're talking as if they follow from your metaethical assumptions when they really should not (without some more details/assumptions).

Note that many people in EA take moral uncertainty seriously, something which is rarely done anywhere else.

Absurd expected value calculations/Pascal's mugging -> Valuing existential risk and high-risk, high-reward careers rely on expected value calculations

Pascal's Mugging is a thought experiment with exceptionally low probability, arbitrarily high stakes events. It poses a noteworthy counterargument to the standard framework of universally maximizing expected value, and therefore is philosophically interesting. That is after all the reason why Nick Bostrom and Eliezer Yudkowsky, two famous effective altruists, developed the idea. But to say that it poses an argument against other expected value calculations is a bit of a non sequitur - there is no clear reason not to say that in Pascal's Mugging, we shouldn't maximize expected value, but in existential risk where the situation is not so improbable and counterintuitive, we should maximize expected value. The whole point of Pascal's Mugging is to try to show that some cases of maximizing expected value are obviously problematic, but I don't see what this says about all the cases where maximizing expected value is not obviously problematic. If there were a single parsimonious decision theory that was intuitive and worked well in all cases including Pascal's Mugging, then you might abandon maximizing expected value in favor of it, but there is no such theory.

There's actual reasons that people like the framework of maximizing expected value; such as how it's invulnerable to Dutch Book Theorems and doesn't lead to intransitivity. In Pascal's Mugging, maybe we can accept losing these properties, because it's such a problematic case. But in other scenarios we will want to preserve them.

It's also worth noting that many of those working on existential risk don't rely on formal mathematical calculations at all, or believe that their cause is very high in probability anyway, as people at MIRI for instance have made clear.

Prioritizing animals (particularly invertebrates) relied on total-view utilitarianism (for me). I value animals (particularly non-mammals) very little compared to humans

But you are wrong about that. Valuing animal interests comparably to humans' is not a uniquely utilitarian principle. Numerous non-utilitarian arguments for this have been advanced by philosophers such as Regan, Norcross, and Korsgaard, and they have been received very well in philosophy. In fact, they are received so well that there is hardly a credible position which involves rejecting the set of them.

You might think that lots of animals just don't experience suffering, but so many EAs agree with this that I'm a little puzzled as to what the problem is. Sure, there's far more people who take invertebrate suffering seriously in a group of EAs than in a group of other people. But there's so many who don't think that invertebrates are sentient that, to be quite honest, this looks less like "I'm surrounded by people I disagree with" and more like "I'm not comfortable in the presence of people I disagree with."

Also, just to be clear although you never stated it explicitly: the idea that we should make serious sacrifices to others according to a framework of maximizing expected value does not imply utilitarianism. Choosing to maximize expected value is a question of decision theory where many moral theories often don't take a clear side, while the obligation to make significant sacrifices to the developing world has been advanced by non-utilitarian arguments from Cohen, Singer, Pogge, and others. These arguments, also, are considered compelling enough that there is hardly a credible position which involves rejecting the set of them.

Lila, what will you do now? What questions or problems do you see in the path ahead? What good things will you miss by leaving the EA community?

For some reason, I've always felt a deep sense of empathy for people who do what you have done. It is very honest and generous of you to do it this way. I wish you only the very best in all you do.

(This is my first post on this forum. I am new to EA.)

I'd be interested in an elaboration on why you reject expected value calculations.

My personal feeling is that expected-value calculations with very small probabilities are unlikely to be helpful, because my calibration for these probabilities is very poor: a one in ten million chance feels identical to a one in ten billion chance for me, even though their expected-value implications are very different. But I expect to be better-calibrated on the difference between a one in ten chance and a one in a hundred chance, particularly if-- as is true much of the time in career choice-- I can look at data on the average person's chance of success in a particular career. So I think that high-risk high-reward careers are quite different from Pascal's muggings.

Can you explain why (and whether) you disagree?

Yea as a two-level consequentialist moral anti-realist I actually am pretty tired of EA's insistence of "how many lives we can save" instead of emphasizing how much "life fulfillment and happiness" you can spread. I always thought this was not only a PR mistake but also a utilitarian mistake. We're trying to prevent suffering, so obviously preventing instances where a single person goes through more suffering on the road to death is more morally relevant utils-wise than preventing a death with less suffering.

Nonetheless, this is the first I've heard that violence and exploitation are under-valued by EA's. It always seemed the case to me that EAs generally weep and feel angsty feelings in their gut when they read about the violence and exploitation of their fellow man. But, what can we do? Regions of violence are notoriously difficult for setting up interventions that are tractable. As such it always seeemed to me that we should focus on what we know works since lifting people out of disease and poverty empowers them to address issues of violence and exploitation themselves. And giving someone their own agency back in this way is, in my view, something worth putting a lot of moral weight on due to its long-term (albeit hard-to measure) consequences.

And now I'm going to say something that I feel some people probably wont like.

I consistently feel that a lot of the critique on EA has to do with how others perceive EAs rather than what they are really like. i.e prejudice. I mentioned above that I generally feel EAs are legit moved to tears (or whatever is a significant feeling for them) regarding issues of violence. But, I find that as soon as this person spends most of his/her time in the public space talking about math and weird utilitarian expected value calculations this person is suddenly viewed as no longer having a heart or "the right heart." The amount of compassion and empathy a person has is not tied to what weird mathematical arguments they push out but what they do and feel inside (this is how I operationalize "compassion" at any rate: an internal state leading to external consequences. Yes I know, that's a pretty virtue ethics way to look at it, so sue me.).

Anyway, maybe part of this is because I know what it feels like to be the highschool nerd that secretly cries when he sees someone getting bullied at break time but who then talks to people about and cevelops exstensivly resaeched weird ideas like transhumanism as a means of optimizing the human flourishing (instead of say caring to go to an anti-bullying event that everyone instead thinks I should be going to if I really cared about bullying). It makes sense to me that many people think I have my priorities wrong. But it certainly isn't due to a lack of compassion and concern for my fellow man. It's not too hard to go from this analogy and argue that

This is perhaps what I absolutely love about the EA community. I've finally found a community of nerds where I can be myself and go in depth with uber-weird (any and all) ideas without being looked at as any less compassionate <3.

When people talk about ending violence and exploitation by doing something that will change the system that keeps these problems in place I get upset. This "system" is often invisible and amorphous and a product of ideology rather than say cost-effectiveness calculations. Why this gets me upset is that I often find this means people are willing to sacrifice giving someone their agency back when it is clear you can do so through donating to proven disease and poverty alleviation interventions to instead donate/support a cause against violence and exploitation because it aligns with their ideology. This essentially seems to me a way of making donation about yourself - trying to make sure you feel content in your own ethical worldview because specifically not doing anything about that violence and exploitation makes you feel bad - rather than making it about the individuals on the receiving end of the donation.

Yea I know, my past virtue ethics predilections are showing again. Even if someone like what I've described above supports an anti-violence cause that though difficult to get a effectiveness measure from is still nontheless doing a lot of good in the world we cant measure I still don't like it. I'm caring what people think and arguing that certain self-serving thoughts appear morally problematic independent of the end-result they cause. So let me show I'm also strongly opposed to forms of anti-realist virtue ethics. It's not enough to merely be aligned with the right way of thinking/ideology etc and then good things come from that. The end result: the actual people on the receiving end - are what actually matter. And this is why I find a "mostly" utilitarian perspective so much more humanizing than people a lot of people who get uncomfortable with its extreme conclusions and then reject the whole thing. A more utilitarian perspective forces you to make it about the receiver.

Whatever the case, writing this has made me sad. I'm sad to see you go, you seem highly intelligent and a likely asset to the movement, and as someone who is on the front-line of EA and PR I take this as a personal failure but wish you the best. Does anyone know of any EA-vetted charities working on violence and exploitation prevention? Even ones that are a stretch tractability-wise would be good. I'd like to donate - always makes me feel better.

These may be silly questions, apologies if so, -Can one be a moral anti-realist and a moral particularist? (Do you mean non-cognitivist? It's just because I didn't think many in EA were moral anti-realists, but perhaps could be non-cognitivists) -What do you feel the consequences of moral uncertainty are? -Are you saying that moral particularism is closest to your beliefs, as a result of moral uncertainty? Or are closest to your beliefs, were you to be a moral realist? -In being an anti-realist, does that mean none of the claims made above are morally normative in nature?

Like others have below I'd like to thank you for an honest and interesting post.

Should we maybe take this as a sign that EA needs to become more like Aspirin, or many other types of medicine? I just checked an Aspirin leaflet, and it said clearly exactly what Aspirin is for. The common “doing the most good” slogan kind of falls short of that.

The definition from the FAQ is better, especially in combination with the additional clarifications below on the page:

Effective altruism is using evidence and analysis to take actions that help others as much as possible.

We’ve focused a lot on finding (with high recall) all the value aligned people who find EA to be exactly the thing they’ve been looking for all their lives, but just like with medicine, it’s also important to prevent the people who are not sufficiently aligned from taking it – for the sake of the movement and for their own sake.

Asprin may be a good example because it’s not known for any terrible side effects, but if someone takes it for some unrelated ailment, they’ll be disillusioned and angry about their investment.

Do we need to be more clear not only about who EA is for but also who EA is probably not for?

If morality isn't real, then perhaps we should just care about our selves.

But suppose we do decide to care about other people's interests - maybe not completely, but at least to some degree. To the extent that we decide to devote resources to helping other people, it makes sense that we should do this to the maximal extent possible and this is what utilitarianism does.