Ben Todd: How can we best work together as a community?

By EA Global @ 2017-06-02T08:48 (+7)

This is a linkpost to https://www.youtube.com/watch?v=S7SFq4v3tRo&list=PLwp9xeoX5p8Pi7rm-vJnaJ4AQdkYJOfYL&index=3

In this 2017 talk, 80,000 Hours' Ben Todd argues that it is easy for effective altruists to overlook the importance of coordinating with other members of the effective altruism community. This perspective has a number of implications — e.g., it increases the importance of gaining knowledge and sharing it with the community, as well as the importance of specialization.

The below transcript is lightly edited for readability.

The Talk

Hi, everyone. Great to be here. I'm Ben, the CO and co-founder of 80,000 Hours. Here's one of the most powerful ways to have more impact with your career: working with a great community. One reason for that is that it's really motivating. If you're around people who want to help others, then that changes the social norms and makes you more keen to contribute. Another thing is it's kind of like networking on steroids. If you meet one person in the community, who can then vouch for you to everyone else, then you can make hundreds of connections at once.

The third thing I wanted to talk about more today is that you can trade and coordinate with the rest of the community. Let's suppose I want to, say, build and sell a piece of software. One approach would be for me to go learn all the things I need myself - engineering, management, marketing. I probably wouldn't get that far. A much better approach would be to instead form a team of people who are specialists in each area, and then build it, and sell it together. Although I'll then have to share the profits with other people, the total size of the gains will be much larger. Overall we'll all win, and we'll all be better off.

What's going on under this is, firstly, we've got division of labor. So by specializing, we can each become much better at the individual skills and therefore more productive as a group. Another thing is that we can share fixed costs. We can all share the same company registration. We can share the same operational procedures. It's not three times harder to fundraise three times as much money from an angel investor. This allows us to achieve economies of scale. In total, we've got what's called the gains from trade.

An interesting thing about trade is that you can do it with people even if you don't particularly share their values, even if you don't have a common aim. So here's a purely hypothetical example. Suppose I run a global poverty charity, and I meet someone who runs an animal welfare charity. Say I don't think animal welfare is a very pressing problem, so I don't think their project has much impact. Then it turns out they feel exactly the same way about me. So neither of us thinks each other's project has much impact.

But then, consider, maybe I know a donor who isn't going to donate to my charity but might be interested in donating to them. So I can make an introduction to that donor to the other charity. This is just a small cost to me, but it might be a really huge deal to them. And then suppose they could do the same. Maybe they know another donor. Then we could trade. This means we end up with a situation where we both get a large benefit for a small cost to ourselves. So we've both been made better off, relative to our own aims. This is why a hundred people working together can have much more impact than a hundred people doing individually what seems best.

Now what I've said so far applies to any community that you might want to get involved with. There are lots of great communities out there, but I know many people who feel that getting involved with the Effective Altruism community gives a particularly big boost to their impact and their career and so on. Why might that be? Well, I think although you can trade with people who don't particularly share your values, if you're in a community and you do share their values, then you don't even need to trade. What do I mean by that? Well, if I help someone else in the community have a greater impact, then I've also had a greater impact as well. So we've both achieved our goals. That means I don't even need to try and make sure that if I do someone a favor, I get a favor back from them later, which would be trading. Instead I can just help the other person and that's already having a social impact.

This unleashes all kinds of opportunities to work together that just wouldn't be efficient in a community where we didn't share our values as much. Technically, we're reducing the transaction costs and principal agent problems involved with working together. So here's an example. We don't normally think about it like this, but actually earning to give can be an example of coordination. Back in the early days of 80,000 Hours, we needed someone to run the organization and we needed funding. There were two of us, me and someone called Matt, who were considering working at 80,000 Hours. But we realized that Matt had much higher earning potential than me, while I would be better suited to running 80,000 Hours, at least hopefully.

What happened is that Matt went to earn to give and became one of our largest early donors, while I went to run 80,000 Hours. Matt also ended up seed funding several other new charities. So alternatively, we could have both earned to give, in which case 80,000 Hours would have never existed, or we could have both gone and worked for 80,000 Hours, in which case it would have been much harder to fundraise early on, so we would have grown more slowly. And those other charities wouldn't have benefited.

So then thinking about the community as a whole, there are going to be some people who are like Matt, and they're relatively best suited to earning money. We can have a greater impact as a group if those people earn to give and then fund everyone else to do direct work. In these ways, by working together, we all have the potential to achieve far more if we work together as a group then we could alone.

But I don't think we actually do work together maybe as effectively as we could a lot of the time. Effective Altruism encourages us to ask what individual actions are highest impact. Some critiques of the community has suggested this could make us biased. This is perhaps the most well known critique of this kind from the London Review of Books: "There is a small paradox in the growth of effective altruism as a movement when it is so profoundly individualistic. The tacit assumption is that the individual, not the community, class or state, is the proper object of moral theorising."

Now Professor McMahan at Oxford had a response to this article which I agree with. He said "I am neither a community nor a state. I can determine only what I will do, not what my community or state will do, though I can of course, decide to concentrate my individual efforts on changing my state's institutions." I think that's absolutely right, in the sense I'm an individual, so ultimately all I have control over is what my individual actions are and what their impact is going to be. But I also think there is truth in the criticism in the London Review of Books, that although maybe individual actions are ultimately what matters, we still have to be wary of the biases in that perspective. If we ask the question of which individual actions are highest impact, then it might cause us to overlook individual actions that would actually have a greater impact.

So in particular, I often see people in the community taking what I call a narrow single player analysis and making a narrow single player analysis of their options, not fully factoring in the relevant counterfactuals, not thinking about how the community will adjust to their actions. While this might have worked when we didn't really have a community, nowadays it doesn't work so well. Instead we need to adopt what I call a multi-player perspective. That breaks down into these two things, which I'm going to cover over the rest of the talk. Firstly, we need new rules of thumb for choosing between our actions. And secondly, there are new options that become worth considering when there's a community involved. These will just be some rough ideas from our latest research.

So firstly, how to choose between our options? I'm going to consider just a single question: should I work at a charity in the community, like Against Malaria Foundation, Givewell, CEA, or so on. Say Amy is considering taking a job in the community. What will her impact be? Now the naïve view of that might be, well the job is high impact so if Amy takes the job, then Amy will have a big impact. But then you hear about effective altruism and someone says "Ah, but if you don't take the job then Bob will take it instead." Therefore Amy won't have much impact. In fact, it would only be worth Amy taking the job if she was going to be much better at it than Bob.

Okay. So I call that the simple analysis of replaceability, and it's an example of single player style of thinking. It leads to a lot of people thinking they shouldn't do direct work but instead should earn to give. But I think this is wrong, and I apologize because it's partly our fault for talking loosely about the simple analysis of replaceability back in the early days of 80,000 Hours. Today I want to try and stamp it out. So the first problem is you might not actually be replaced. There's a chance the charity just wouldn't hire anyone otherwise. In fact, this often seems to be true. When we talk to charities in the community, they often have positions they've been trying to fill for a while but they haven't been able to fill.

One reason for this, one cause that's driving this is that there are donors who want to support the community with money on the sidelines. That means if you're a charity that finds someone who's worth hiring, then you can just fundraise extra money to hire that person. This creates a situation where you have threshold hiring, where anyone who's above a bar can get hired because you just raise more money to cover that person. So then all those people are not replaceable. In fact, I think there's a bunch of other ways you can end up not being replaceable, such as through supply/demand effects, which we cover elsewhere.

This actually means that you can end up being pretty valuable to the organization that you're working at. One way to estimate how much impact you might have at the organization is to ask them to make this trade off. You say "I could either work for you or I could donate X dollars per year to you. At what value of X would you be indifferent between these, roughly?" That's just a way of gauging the relative impact of you working there. We actually did this with 12 organizations in the community last summer and these were some of the figures that came out.

This is all pretty rough. There's also reason to think that the organizations might be biased upwards. I think it at least suggests that because these figures are much more than most people would be able to donate, these people are having a greater impact than they would through earning to give.

We also just asked the organizations how talent versus funding constrained they think they are. You can see there's a clear tilt towards being talent constrained. The interesting exception is the animal charities, which report being more funding constrained relative to the others.

So the first problem with the simple view of replaceability is that you might not actually be replaced. The second problem is where the community really comes in and that's what I call spillovers. Suppose Amy takes job one and she would have been replaced, so someone else would have taken the job anyway. But by taking job one, that means that Amy actually then frees up someone else to go and do job two. Now if you were just considering a job that would be filled by someone who doesn't really care about social impact anyway, and would just go and do maybe some random job in the corporate sector, which you don't think has much impact, then you can safely ignore the spillover.

But in the current community, it's not so obvious that you can ignore it. Probably this person, Bob, would also be concerned by social impact and would go and do some other high impact thing. They might earn to give. They might work at another charity in the community. This is actually then a significant component of the impact. In fact, I think this kind of case is not even a hypothetical example. I've actually seen cases where someone didn't take job one because they thought they'd be replaceable. That meant that someone actually had to be pulled out of job two to go and take job one instead, where job two was a similarly important role as job one.

Taking into account these spillovers, then what is the impact of taking a job? How do we analyze it? I think this is actually still an unsolved problem, but here's a sketch of our current thinking. Basically, if Amy takes job one than she frees up someone else to do job two, who then frees up someone else to do job three. It causes a chain of replacements. Where does the chain end? There are two main endings it could have. One is that someone goes into a job that wouldn't have been filled otherwise, a threshold hiring situation and then it stops. Or it can keep going until you hit the marginal opportunity, so the best job that wouldn't have been taken otherwise. That means that at worst, even if Amy was fully replaceable, she'd be adding someone to the margin of the community, which would still be a significant impact. This shows that the simple analysis of replaceability is actually underestimating impact in both of the cases we've covered.

So the first point is knocking someone out into the margin. Secondly, hopefully you're enabling all these people to switch into jobs they're a slightly better fit for, because they'll be choosing jobs that are slightly better. Amy's taking job one because she's a better fit for that one, and then that's freeing up Bob to go and do something else that's a better fit for him. So potentially there's another benefit as well, that the community gets to a better allocation overall.

I think if we zoom out a bit, when you're working in a community the picture is more like this one, where we have a pool of thousands of roles and thousands of people. From the community's perspective, you want to match the people into the ideal allocation over those roles. From an individual point of view, the question then becomes what can I do to move the community towards that ideal allocation? That becomes the highest impact role for me. I think the key concept here is comparative advantage compared to the rest of the community. Comparative advantage can be a bit counterintuitive so I'm just going to explore that in a little bit more depth.

Suppose the community needs two more roles at the margin. It needs someone doing research and it needs someone doing outreach. If Charlie did research then he would have two units of impact, whereas if he did outreach, he'd have one unit of impact. Meanwhile Dora would have ten and one, respectively. So then the question is what is the best allocation? There are two possibilities and the best one is this one, where Dora does research and Charlie does outreach, because ten plus one is bigger than two plus two. But what's weird about this example is that Charlie is actually worse at outreach than he is at research. He's also worse at outreach than Dora is at outreach. So in no sense does he have an absolute advantage, nor does he have good personal fit with outreach. But nevertheless, it's the highest impact thing for him to do. That's because he has a comparative advantage in outreach. Basically, he is relatively less bad at outreach then Dora is.

As we saw, the simple analysis of replaceability encourages you to only do a job if you're better than the person who would have replaced you in that job. However here's an example where you should do something that you're actually worse at because then it enables the community to achieve the best overall allocation. I actually have a suspicion that an example like this might even be a real example. Lots of people in the community reason, well I'm good at analytical things, so I probably should do research. But that doesn't actually follow. What actually matters is how good you are at research compared to the other people who might do research and outreach.

If we have lots of good analytical people but few outreach people, then even though you might be good at research compared to people in general, you might have a comparative advantage in outreach. Something similar seems maybe true in operations roles as compared to research roles, and may be also with earning to give. People sometimes say "Well, I have high earning potential so therefore I should earn to give." But again that doesn't follow. What actually matters is your earning potential relative to others who might do direct work. If it's high relative to them, then you should earn to give. But if the other people doing direct work would also have high earning potential, then you might still have a comparative advantage in direct work. Unless of course, you're Chuck Norris, who has a comparative advantage in everything.

How can you find your comparative advantage in real cases? I think basically it boils down to asking people in charge of hiring at the organizations about your relative strengths and weaknesses.

So summing up. How do you analyze the impact of taking a job? Firstly, you probably cause some boost to the organization. You're not fully replaceable. Then you can try and ask them about it or estimate it yourself by asking them to trade donations against you working there. Secondly, you cause some spillover effect potentially to the rest of the community. And then the question is, is that role above the bar for the community as a whole and does it play to your comparative advantage? Are you getting the community to a better overall allocation?

And I think similar considerations apply to donating as well. I sometimes see people saying "Well, I shouldn't donate to charity one because someone else will donate there instead, and so I'll be replaceable. I won't actually have much impact." But then I think a whole kind of similar analysis applies. So firstly, you probably won't be fully replaceable. Not all the money would have been given to that charity otherwise, especially if you consider over time. Maybe the money would have been given otherwise, but with a delay, which would slow down the charity. And then secondly, even if it is fully replaced, you're still then just freeing up some other donor in the community to go and donate somewhere else, which would probably be somewhere reasonably high impact as well.

There's a lot more to say about that and there's a bit more detail in this article online called The Value of Coordination. That was all about the new rules of thumb for choosing between your options. But being part of a community also changes your career by opening up new options that might not be on the table if you were just thinking from a single player point of view. The EA mindset however, with a focus on individual actions, can lead us to neglect paths that allow us to have an impact through helping others do more, because the impact is less salient. But now that there are thousands of Effective Altruists at least, the highest impact option for you could well be one that just involves helping others have more impact, boosting others.

So I'm just going to give five examples. These aren't exhaustive or exclusive, but they're just some ideas. So first one is five minute favors. We all have different strengths and weaknesses, knowledge and resources, and now that there are thousands of people involved, there are probably lots of small ways that we can help others have much more impact at very little cost to ourselves. I call those five minute favors, which is just a term from Adam Grant. That means these kinds of things are really worth looking for. Do you know a job that needs filling? There's a good chance that there's someone in this room or at this conference who would be a good fit for that job. If you could make an introduction, that might only take you an hour, but it would help them for years. Or there's probably someone in this room who has a problem that you know how to solve. You know someone who's solved it before. You know a book that could help them solve it and so on.

Second example, Operations roles in general. Kyle moved to Oxford and ended up becoming Nick Bostrom's assistant. He thought if I can save Nick Bostrom some time, then that will enable more research and outreach to be done, which I think could actually be really high impact. Instead, people often feel like these operations roles are replaceable by someone from outside of the community. But actually because you have to make lots of little decisions in these roles, that require quite a good understanding of the aims of the organization, they're often very hard to outsource and hard to replace.

Third example. Community infrastructure becomes much more valuable the larger the community. So for instance, having a job board isn't really needed when there's just a hundred people, but when there's thousands of people, it can play quite a useful role, which is why we added one to the 80,000 Hours site a few months ago. By community infrastructure, I mean anything that helps make the community coordinate more efficiently, such as this event, or it could even mean stuff like setting up good norms of communication that make it easier to work together, like always stating the evidence you’ve used, or just being nice. If you can help a thousand people have one percent more impact, then that's like having the impact of ten people. On the other hand, it means if you do something destructive, then you ruin it for everyone else.

Fourth example is sharing knowledge with the community. The more people there are in the community, the more worthwhile it is to do research into what the community should do, and then share it with everyone else, because then there's just more people who can act on the findings, so that the value of new information is higher. An example of this is just writing up reports in areas where we have special knowledge of. These are some examples recently from the Effective Altruism forum.

For instance, Lee Sharkey is a public health consultant to the World Health Organization and he wrote up some ideas for a new cause area, which is increasing access to pain relief in the Balkan Countries. It can also mean that it's sometimes worth going and learning about areas that don't seem like the highest priority but just might turn out to be. In a smaller community this exploration wouldn't be worth the time, but as we become larger it becomes more and more worth it.

The fifth and final example is specialization. If the community were just a couple of people, then we'd all need to become generalists, but in a community of 10,000 people, then we can all become experts in our individual areas, and therefore be more than 10,000 times as productive as an individual. This is just the division of labor like we mentioned right at the start with the software example.

For instance, Dr. Greg Lewis did our research into how many lives a doctor saves. This convinced him that his greatest impact wouldn't come through his direct clinical practice. Instead he decided to go and do a Masters in public health. Part of the reason for that is because it's an important area for the community, especially around pandemics, but there's a lack of people with that skill set. Greg actually thinks that AI risk might be a higher priority in general, but as a doctor he has a comparative advantage in doing public health. Right now I and many others think that one of the greatest weaknesses of the community is a lack of specialist expertise and knowledge. We're all pretty young and inexperienced. These are just some areas that I'd like to highlight but I'm not going to go over right now.

Summing up everything. When choosing whether to take a job or donate somewhere, don't just assume that you're replaceable. Rather, ask the organization about how big the boost would be or try and estimate it yourself. Consider the trade off between donations and working there. Secondly, consider the spillover benefits and whether the role plays to your comparative advantage. Then look for new options that become available. Doing five minute favors. Operations roles in general. Setting up better community infrastructure. Gaining and sharing knowledge with the rest of the community. And specialization.

We still have a lot to learn about how best to work together. I think there's a lot more we could do, but I really believe that if we do work together effectively then in our lifetimes we can make a major impact on reducing catastrophic risks, eliminating factory farming, reducing global poverty, and many other issues.

Thanks for listening.