There are a number of key questions in Effective Altruism that depend on an understanding of political science, policy, international relations, and related fields. Unfortunately, these areas are hard to understand, and in my view, there is a general lack of understanding of them in the community. It's particularly important to have background because you can't figure out what you need to know unless you know a little bit about what is out there. Given that, I will be writing a series of posts on the topic that discusses ways to think about policy and international relations, one that hopefully gives enough background to ask intelligent questions and know where to look.

In this introductory post, I won't actually discuss any of the substantive topic. Instead, I will explain why I think this is important, why it is hard, and a bit about why most people don't need to read the rest of the posts. If you're already convinced you need to know this, and why, feel free to skip to the next post (which is now posted.)

Why I'm Writing This

One part of the reason I am writing about this is that effective altruism has been attacked in the past, occasionally justifiably, for being naive or dismissive about many of the issues considered by experts in these areas. Most of these criticisms are leveled at a weak-man versions of effective altruism, or based on years-old claims - but these views are sometimes still espoused by people in the community. Even if EA Organizations are now better informed, getting the community more familiar with these areas will help them understand the issues with simplistic approaches.

There are also a large and growing number of people involved in effective altruism who are more acquainted with these academic disciplines than I am, some with significant expertise, and many more who have learned or are learning about the topic. In fact, some portion of them would be better than I am at writing these posts, and I will welcome feedback from them either here or privately, and from everyone else. These experts are also critical sources of information within the movement. As mentioned above, however, their expertise is much more useful if the questioner knows enough to understand what the experts will and will not know, and what their discipline involves.

Why does this matter to Effective Altruism?

A number of cause areas are deeply affected by policy and both domestic and international politics, including at least global health, poverty, and existential risk. In some cases, there are debates about whether to prioritize systemic change or direct action, or how to prioritize policy interventions. In other cases, the interventions themselves involve international relations or are attempts to improve how organizations in these realms work. Some people are considering work in these areas in different countries, and need to understand the area. And perhaps most worrying for effective altruism, uncertainties from politics can completely change the value of even typical, non-political interventions. For example, some studies seem to show that private funding for public health is displacing government investments, making estimates of impact overstate the value of those charities, or even have perverse net effects.

If this is an issue, perhaps we just need to ask the experts? If we can assess expertise (a critical skill worth learning), we can ask the right questions and get answers. But even asking the right questions requires a basic level of knowledge that most people lack - and you might not even know what you don't know. And evaluating the answers is often hard as well - once upon a time, political science PhDs weren't centered around statistics, game theory, and intervention design. That's no longer true, and it's critical to understand the limits of what the experts know, and what tools they use.

And not only does understanding international relations and public policy require expertise, it's also very easy to screw up when you only partly know what you are doing. In statistics, even experts are vulnerable: a review by a pair of statisticians of the book "Freakonomics", written by a PhD economist, found numerous basic mistakes. I would argue that much of the Effective Altruism community is similarly naive about the policy and international relations impacts of our work - and a basic level of understanding is critical to prevent errors.

This is Hard

There's an old joke about the difference between hard sciences and soft sciences - the reason hard sciences are harder is because someone can check your answers and tell you you're wrong. In this view, "soft sciences" like political science don't have clear answers, just debates and opinions. The joke isn't entirely fair, but it does contain more than a grain of truth. John D. Cook put it a bit differently: "the soft sciences are hard in the sense of being difficult, but not hard in the sense of studying indisputably measurable effects and making sharp quantitative predictions." But it's simply not true that political sciences doesn't study indisputably measurable effects - the number of votes cast, national GDP and economic impacts, and the fighting of wars are all obviously measurable${}^1$.

The problem with soft sciences is the second half half of the above quote - soft science cannot make sharp predictions. We can predict the position of Mars in 50 years to many decimal places, more accurately and more easily than we can predict the number of people who will live there then. Phil Tetlock's first claim to fame${}^2$ was his book, "Expert Political Judgement" (EPJ). In the book, he found that "centrist, cognitively flexible foxes had significantly better predictive track records than wedded-to-orthodoxies hedgehogs. [But] this success was relative; foxes did only slightly better than extrapolation algorithms." It wasn't impossible to make well-calibrated predictions, it's just that all well-calibrated predictions were imprecise. Being more certain is usually a mark of overconfidence.

And this means that we cannot let Effective Altruism off the hook. If we could simply say that politics is unknowable, we can claim maximal cluelessness, and perhaps make decisions without considering politics. But we are not maximally clueless, at least in the medium term. That is, our lack of knowledge is bounded. That means our expected value considerations can account for the uncertainties to some extent. If we care about doing the most good we can do, that means we must deal with the uncertainties. That is, to fairly evaluate our altruistic endeavors, we must consider the political impacts of our decisions, and we must consider the possibility that political or policy interventions are highly effective.

But if the best that can be done in reducing uncertainty is superforecasting or similar, and the predictions from such attempts are still barely better than chance, spending time on political analysis is, perhaps, low-value. For instance, if we want to be confident that our work does good, political interventions with very uncertain but very effective outcomes in expectation may not be our best bet. But the uncertainty matters, and even though the uncertainty is not fully resolvable, our efforts should be non-zero. But finding what non-zero level of effort is needed is a hard question.

What and how much should you want to know?

What you want to know, and how much you need to know, are going to depend heavily on what you're trying to do. I hope most readers will find the rest of the post in this series informative and interesting, but I don't expect that it will change anything about what they are doing. However, for some percentage of readers, those working in designing interventions or considering their impacts, this will matter a significant amount.

If you're choosing between charities to donate to, you just want to know enough to check that they thought about relevant issues. If you're evaluating charities, you should read these posts, and then spend some time thinking about what the unintended and second order could be, and whether there is a good way to ask an expert about it, and what you want to know. If you're designing an intervention to be effective, you really want to consult heavily with an expert to make sure you've thought about everything, and (I hope) the posts contain useful background to have before doing that. And if you're working on a domain that involves international relations - like AI or biorisk policy, or climate change, or electoral reform, these posts aren't nearly enough${}^3$ - you should be reading textbooks on the topic, and then considering which graduate-level courses on these topics you should enroll in, and which experts you should hire or collaborate with.

Given all of that, I'm hoping to review some of these topics, and provide a number of links and similar resources for those who are interested in learning more - if they need to.

1) In EPJ, Tetlock points out that originally, many experts said that expert political judgement was not just hard, but impossible - that outcomes are subjective and forecasting isn't meaningful. This is nonsensical. The number of votes cast for each candidate in an election, or the number of people killed in a war, can be known in the same way we know the position of a planetary body. And (I would argue,) the fact that public opinion about a topic isn't directly measurable and uses statistical estimates is no more damning than the fact that in biology the energy of particles isn't directly measured, and we rely on aggregate measures of temperature.

2) In effective altruist circles, he's better known for his work on Superforecasting, which we'll get to shortly.

3) Even then, I think you can gain something from this series of posts.