Why It’s OK To Ignore Politics is a short philosophy book about the merits of political activities, such as voting and donating to political campaigns. It considers and rejects arguments that we are obligated to engage in politics, then goes further and argues that political engagement is positively unethical in some ways.
I think the book does show that it is okay to ignore politics, as long as it means that you are spending extra time and money on other altruistic activities instead. But I am not convinced that it is unethical to engage in politics, as long as it is done with good sense and in moderation; politics should still be considered one of many viable options for EA activity.
Introduction
The book includes three chapters which reject nonconsequentialist arguments in favor of political engagement: the idea that it’s unfair to ‘free ride’ on the benefits of other people’s political activity (chapter 4), the idea that people who do not act in politics are complicit in injustice (chapter 5), and the idea that it’s virtuous to express certain political messages (chapter 6). I did not take the time to read these chapters because I do not subscribe to those ideas in the first place. Christopher Freiman is a professional academic philosopher so I presume his arguments are good. They may be helpful in refuting philosophical criticisms of EA’s lack of inherent political engagement, especially as Freiman himself is basically supportive of the EA project.
He makes three other arguments which are straightforwardly applicable to consequentialist thought (but can apply to nonconsequentialist thought as well). First, that humans are deeply flawed at the task of political reasoning, so that our political judgments are no better or little better than chance; second, that the chance of making a difference in politics is so low that it fails to meet the Effective Altruist criterion of being a productive use of time and money; and third, that political engagement is personally and socially unhealthy.
It’s mainly written from an American point of view but its arguments can broadly apply to other democracies.
Are we all partisan hacks?
In chapters 1 and 2 Freiman lays out the problem of obtaining adequate information and making wise judgments on political issues. He points out that political issues are very complex and most people are terribly biased on political issues. I have quibbles with a few of his specific points and haven’t looked into many of the psychology studies he cites, and it would be sensible to also mention the fact that many voters are not just misinformed but also downright irrational and chaotic, for instance by voting for candidates who go against voters’ stated policy preferences or picking candidates for baseless emotional reasons. Regardless, I don’t think this line of thinking is directly relevant because the whole point of politics is simply to be better than other voters. I may not have a serious and rational approach to making a political impact, but if other voters are even worse than me, then my voting will make things better on average. If it turned out that typical Americans are very rational and well-informed about politics, that might actually dissuade me from caring about it. The fact that most people get politics badly wrong is equally interpretable as a reason to be optimistic about making a positive impact in politics.
What we really need to know is whether we can identify certain people (ourselves, or other people who we might trust) as having substantially better political judgment than most other people. Some of what Freiman writes helps address this.
Information and intelligence
One possibility is to just think to ourselves about whether we seem to know more than others. Freiman cites Anson (2018) about the Dunning-Kruger effect in a political knowledge quiz, finding that most people were overconfident in their political knowledge. But the study still finds that the highest scorers were more confident in their knowledge—in fact, they were underconfident. As I understand it, this is a common pattern with the Dunning-Kruger effect. The upshot is that we should be more dismissive of our own knowledge if we find ourselves feeling moderately better-informed than average, but if we feel much better than average then we are probably at least somewhat better than average, and likely even better than we think.
We can also look at things like IQ or SAT tests, but those also have dubious relevance. Freiman cites West et al (2012) who found that higher-SAT-scoring individuals were more prone to overestimate the degree to which they were less biased than other people, but they still seem to have found that more cognitively sophisticated individuals were slightly less biased on the actual tests that they were given than other people (although not by nearly enough to explain their higher opinions of themselves). In any case, they note that this finding is at odds with most literature—they point out one previous paper which supports the idea that smarter people are more biased, while also listing eleven that disagree. Meanwhile, Stanovich et al (2013), not cited by Freiman, finds that myside bias is unrelated to intelligence.
It would be nice to have a proper metanalysis tying such empirical findings together. But in practice, when it comes to basic political knowledge and susceptibility to bias, we can just take these tests ourselves rather than think about what studies say. That way we are experimenting on ourselves so that we don’t have to worry as much about the generalizability of results. And we can either get a high score, or find other people who get a high score and just listen to them.
Motivated cognition
The more difficult problem raised by Freiman is that more informed and intelligent people don’t appear to use their powers to actually make better judgments. Or do they? Freiman cites Kahan et al (2017) who shows that more numerate people display more politically biased interpretations of data, but as is apparent from Figure 7, they do still find that more numerate people have more accurate views even when the data goes against their side—they just get even more accurate when the data supports their side. So, there is a rather difficult question here about who is more trustworthy on politics: the person who is 35% accurate when it hurts their cause and 40% accurate when it helps their cause, or the person who is 50% accurate when it hurts their cause and 85% accurate when it helps their cause? (My followers are divided on this, but tend to prefer the latter. ) Then at the highest levels of numeracy, it seems that greater numeracy objectively increases accuracy all around without significantly affecting their partisan bias (see Figure 6).
Freiman also writes “training subjects on logical reasoning didn’t enable them to overcome their partisan bias in evaluating arguments” but the paper in question only found that people who went through logic training displayed a partisan bias, it did not make a valid comparison of trained and untrained subjects from the same sample.
Freiman also cites Kahan (2015) who finds that science literacy is correlated with greater polarization on climate change. The gap between Democratic and Republican recognition of the mechanism of man-made climate change is nonexistent among people who are scientifically illiterate, but grows monotonically as we look at people who are more scientifically literate. And in particular, Republicans become monotonically less likely to recognize man-made climate change when they become more scientifically literate. So Republicans and Democrats still get more polarized even as they go from the 90th percentile to the 99th percentile of scientific literacy. But Kahan goes to some lengths to argue that this doesn’t mean what it sounds like it means. It seems like the main issue here is that scientific literacy is presumably correlated with political literacy, and people who are more politically literate do a better job of recognizing what their political party believes and deferring to that. Low-info conservatives don’t know what the Republican Party thinks about climate change, so they express ordinary views. High-info conservatives know that every good Republican denies man-made climate change, so they deny it too.
Freiman cites Bolsen et al (2015) as showing that better education increases viewpoint polarization. While this study does show this dynamic for the claim that global warming is happening, it turns out (rather bizarrely) that low-info conservatives are more likely than low-info liberals to believe that global warming is anthropogenic, and as voters become more knowledgeable, viewpoint polarization decreases because conservatives become less likely to believe in anthropogenic global warming.
Freiman cites Drummond and Fischhoff (2017) as showing that science literacy and education are correlated with greater viewpoint polarization about politically polarized scientific issues, through an analysis of the General Social Survey. This includes greater polarization about support for stem cell research, but that could plausibly be a legitimate case of more informed conservatives learning about how something violates their moral values. They also find greater polarization about belief in the Big Bang and human evolution, but only education and science education are correlated with conservative disbelief in these cases; science literacy increases polarization but still correlates with conservatives being a bit more likely to believe in these things. Finally, they find greater polarization about climate change, but with caveats. First, as the authors point out, “the item on climate change asks participants to assume that climate change is happening and then to express their level of concern, despite the well-documented finding that a significant minority of Americans dispute its existence”. Second, while education increases polarization, it still makes conservatives a bit more concerned. Third, science literacy only has a modest effect on polarization because low-literacy conservatives are actually more concerned than low-literacy liberals. As people get more literate, things flip, with an end result of only slightly more polarization. So on this issue, going from low science literacy to average science literacy reduces polarization, but going from average literacy to high literacy sharply increases it.
It’s also probably true (though Freiman doesn’t cite this) that broadband internet access increases partisan hostility, apparently because the new information allows voters to learn more about what their party and the enemy party are really like. (It isn’t because of ‘fake news’.) And for the most surreal example of polarization-via-information, I recall David Shor theorizing—though I can’t find the source—that the election of Barack Obama increased racial polarization simply because it taught low-info Americans which party was ‘the party for white people’ and which was ‘the party for black people’.
Finally, Freiman doesn’t mention Achen and Bartels (2006) who looked at perceptions of the budget deficit and found that more political knowledge is correlated with greater viewpoint polarization except among the “best-informed 10 or 20% of the public”. Once we focus on that group of highly informed citizens, Republicans and Democrats finally converge a bit towards each other’s factual views as they become more informed.
OK, so different findings contradict each other. And most of these studies are just observational and I have no idea if they will generalize or replicate.
Tentatively, we can say that two things are probably true. First, that more information is associated with more polarized claims until we reach the point of people becoming very highly informed, at which point the impacts of additional information are unclear.
And second, that the fact that voters get more informed about the characteristics of the political parties is of comparable or greater importance to the fact that they get more informed about issues themselves. I predict that if someone does a similar study which controls for political literacy, they will find that scientific literacy is more likely to decrease viewpoint polarization. Of course, that still leaves open the question of why we should become politically literate if it leads us to distort more issues to fit our point of view. But I expect that it also leads to voters becoming more rational at voting for candidates who match their goals and values. If you have right-wing values, then being wrong about climate change is a small price to pay for knowing which politicians will protect guns, restrict abortion, and anger liberals.
Again, maybe someone could do a proper literature review or metanalysis. Michael Hannon’s more recent paper Are Smarter Voters Better Voters? includes more arguments to back up the idea that additional knowledge gets weaponized to entrench or exaggerate polarized viewpoints, but he seems to badly misrepresent Kahan et al (2017), and I just don’t feel up to going through the rest of it.
Edit, June 8 2021: I’ve since come across further studies testing the idea of motivated cognition. Van der Linden et al (2018), Tappin et al (2020) and Persson et al (2021) have tested the theory of motivated cognition and all found problems with the theory. So ultimately there really is not good evidence that more intelligence or information among partisans causes increases in polarization.
Context!
For the actual context of the EA community we should move beyond the basic framework used by Freiman and Hannon.
First, we should be more cautious about assuming that nonpolitical judgments will be unbiased. The studies linked previously compared political judgments to things like the efficacy of skin cream and the probability of having cancer. What we do in EA is not so anodyne. Donating to AMF or The Humane League or MIRI is typically not as tribal as politics but certainly never as clean as textbook statistics. There is still room for tribalism and selfish biases to affect our judgments on these issues. I haven’t seen any studies comparing people’s rationality about politics with their rationality about other forms of do-goodery. It still seems right to suppose that we’re more vulnerable to bias in politics than we are to bias in other cause areas, but this does weaken the argument.
Second, we should probably worry less about our individual cognition and more about our general environment. As I noted previously, the increased polarization from information probably has more to do with the fact that people learn more about political tribes and pick a side accordingly. If we want to do better, we need to ask ourselves: does the EA community do more to reward accuracy and wise judgment, and do less to enforce political conformity, compared to how most other politically relevant communities work? It’s not straightforward to empirically test this but it’s important to consider. And I think it gives more reason for us to be optimistic about our political impact.
Solutions
Freiman later suggests that it would be much easier to talk about politics if we weren’t attached to political sides. In a call for “apolitical politics” at the end of chapter 7, he says that if you really do want to engage in politics, you should at least “quit your political party and join a local effective altruism group in your newly spared time”. Michael Hannon—in “Are Smarter Voters Better Voters?”—similarly argues that we should reduce our attachments to partisan identity. Note that this doesn’t mean rejecting claims that one political party is much better than the other; sometimes that really is the case, and Freiman does a good job of distancing himself from bothsidesist assumptions. This also doesn’t mean that we should reform the political system to weaken or abolish the political parties, nor that we should vote for independent or third-party candidates; Freiman doesn’t mention these (bad) ideas. Rather, the point is that we shouldn’t identify ourselves with a political party or any other political group.
I would say that even if we don’t go so far as to avoid all political identity, we can give primacy to our EA identities over our political ones, and that should help us be more detached. I think I have seen a few EAs who unfortunately care more about political identity than about EA identity but there are many others of us who are more inclined to take a healthier approach.
So while Freiman says that political judgment is so hard that you need to be highly politically informed and comprehensive in order to a good job (see this page for an example of what that can look like in practice—it’s pretty ridiculous), an alternative takeaway is that setting yourself up with the right identity, environment and attitudes is itself sufficient to make you better than most citizens even if you possess relatively little political knowledge and spend relatively little time analyzing. Although of course it can be hard to tell whether you have “the right identity, environment and attitudes”.
We can also simplify electoral debates to a few issues which we know well. From our ordinary involvement in EA, we already know about foreign aid and animal welfare, and while we typically think about these things through the lens of private charity, they also have relatively straightforward implications for government policy. To a lesser extent, many EAs are also well acquainted with issues in housing, criminal justice, and pandemic response. If a politician or party is better on the issues which we do know about, then presumably they’ll tend to be better on the issues which we don’t know about.
Additionally, there is a better alternative to measuring rote political knowledge or philosophical rationality—we can look at forecasting track records. Freiman cites Tetlock’s finding that many experts are terrible at forecasting, but conveniently leaves out the finding that quite a few people have good forecasting track records and/or can be taught to do even better (at least for short-run impacts). In 2020 we even got explicit forecasts for the consequences of the United States presidential election. Of course this can only cover a limited subset of policy impacts, but it still helps a lot.
Finally, sometimes political disputes come down simply to moral values. Freiman points this out as a source of complexity—look, now we have to know about empirical questions and moral ones. But clear moral convictions could make things easier by sidestepping our problems with empirical uncertainty. For instance, if we care equally about people around the world, then we don’t have to worry as much about overspending on preventing climate change. Alternatively, if you’re a pluralist or very uncertain among moral theories, then many political issues will be more complicated for you but you also might have an easier time dismissing political ideas which are very morally brittle.
Directions for further inquiry
Freiman or others may have responses to these points but I feel that we would only end up in circles trying to solve the problem of philosophical skepticism. We can escape that by trying to actually make political judgments in the real world and seeing how well they fare against (a) collective critique and (b) hindsight. Ultimately, the reason I believe that certain political interventions are effective is not based on philosophical notions of my abilities, it is quite simply that I read arguments for and against them and sometimes determine that the former are more convincing. I’ve been making serious political judgments for a while, and have had a few chances to check them against collective critique and hindsight. At this point, I do think that we can have more confidence in AMF than in almost any political cause, and there is greater risk that political efforts cause harm. But I don’t think it’s a big difference. Electoral politics seems like something that’s more speculative than GiveWell charities but still not as speculative as direct x-risk prevention efforts. And there are a few specific policy views which are justified just as robustly as GiveWell charities. And if political inquiry really is very unreliable, then someone should be able to point out serious flaws in my specific political rationales. Or we could look at the investigation which Open Philanthropy Project has already done on certain policy areas.
Edit, June 8 2021: As described above, the evidence for the theory of motivated cognition is really not very strong, so it is fair to presume that gathering information and smart thinkers will indeed produce better political judgments.
Costs and Benefits
In chapters 3 and the conclusion, Freiman disputes the idea that voting is an effective use of time which could be better spend learning and helping with other causes, such as GiveWell charities. He points out that it’s incorrect to say that voting only takes an hour or so because that ignores the cost of obtaining knowledge about who to vote for. If we want to become properly informed about politics so that we can be confident that our votes are doing good rather than harm, that will a much longer period of time. A lifetime of serious political engagement can involve lots of time spent doing things like reading, studying, chatting, and volunteering. And even if you intend to spend minimal time on politics, it might lead to time-consuming habits anyway.
What are the benefits? Freiman correctly points out that a vote in a swing state has a tiny chance of swinging a presidential election: about one in ten million (and in 2020, for most swing states it was actually closer to 1 in 20 million). He then incorrectly argues (citing Dubner and Levitt) that even if you are in a perfectly close election, your vote won’t matter because the election will be decided by recounts and judges anyway. Veterans of debates over meat consumption will recognize the problem here; see Hilary Greaves’ point about vegetarianism. Elections are analogous.
Freiman never presents a true cost-effectiveness calculation, but to his credit he estimates the opportunity cost of political engagement. One hour of political engagement per month starting from the age of 20 (heaven knows many of us are guilty of more than this), sacrificing the average American wage of $22.50 per hour, with an investment return rate of 8%, implies a loss of $70,000 by the time you turn 60. If you donated that much money to a GiveWell-recommended charity, which saves lives for $4,000 each, then you could save 17 lives. But he neglects the possibility that in 2060 it may not be possible to save lives for $4,000 each. Of course maybe it will actually get cheaper (though I really doubt that). But clearly this comparison depends partly on the virtues of patience and giving now versus later. For a clearer comparison we should look at immediate impacts from both efforts.
I have a calculation based on the 2020 presidential election which suggests that the average swing state vote in a near future presidential election will have a direct impact which is just as good as 0.02 lives saved in expectation. This counts all direct human and animal social impacts which can be decently estimated. It excludes some speculative benefits including long-run foreign policy issues and the probability of nuclear war. Considering that the cost per life saved through Givewell is around $4,000, a swing state vote is worth $80 from a typical Effective Altruist. That’s equivalent to an unimpressive 3.5 hours of time at $22.50/hour. Keep in mind that these numbers are not easily comparable—the GiveWell figures are more robust, but my figures exclude numerous long-run benefits from things like sound foreign policy and the integrity of American democracy. Of course we should also discount these for deep uncertainty about political consequences—maybe assume a 5 percentage point greater chance that our political actions are net harmful compared to our charity efforts, so we should discount these estimates by 10%.
Still, this seems good enough to justify the practice of voters in swing states taking a rudimentary look at what the EA community or other experts believe and voting accordingly, while failing to justify any practice of being seriously politically informed. Even a project of becoming seriously informed in order to tell other EAs how to vote might not be justified at this level.
Freiman also points out that even those such as William MacAskill who support swing-state voting still acknowledge that voting for the president in safe states (which includes most Americans) is pointless. And this is fair. The EA community seems particularly underrepresented in swing states.
But things look better if you add other elections. Freiman dismisses congressional elections on the basis that they are frequently safe incumbencies, which seems true, and Congress doesn’t get much done anyway, so Congress is probably a lower priority than the presidency. A similar mindset to presidential elections seems appropriate: if you live in a close district for an upcoming election, then take a bit of time to choose and vote; otherwise don’t worry about it.
Then with state and local elections, we can have still more impact. These are obviously less important but also more tractable. In fact there are reasons to believe that these should be a higher priority than federal elections (I intend to share analysis of this soon).
In deciding whether to vote, we only need to find one election on the ballot that really matters. If you live in a presidential swing state or a state with a close Senate race or a state with a close gubernatorial race or a city where you can find a good person running for city council or … then you can justify a basic level of familiarity with political issues and time spent voting in at least some of the elections. And while you’re at the polls or filling out a mail-in ballot, you may as well cast a vote in all the other simultaneous elections as long as you have any real inclinations for certain candidates.
When you add different elections up, maybe the average direct impact of a ballot even in a safe state is equivalent to 0.08 lives saved, which is like donating $290 to a GiveWell charity, or 13 hours of time. Now it is looking more reasonable to be mildly politically informed.
Note that in some circumstances like recent U.S. presidential elections, the case for voting for Democrats is highly robust. There really seems to have been a stronger consensus among economists and political scientists about the threat of Donald Trump than about the merits of RCT-driven development. But the majority of elections aren’t so stark.
Freiman doesn’t say anything about direct activism, advertising, lobbying and campaign donations. He dismisses protests by saying that adding an individual to the size of the group has barely any effect, but that can be challenged; again I intend to do a cost-effectiveness analysis of this soon. But they offer further opportunities to make a significant difference, assuming that you actually do them in prudent ways that don’t cause political backlash (a nontrivial problem). And the more opportunities one has to leverage political knowledge into taking a variety of productive actions besides voting, the more one can amortize the upfront costs of becoming generally politically wise.
Is ignorance bliss?
Beyond time and money, political engagement has indirect costs. Freiman points out that political engagement is linked to stress, degradation of families and friendships, employment discrimination, relationship discrimination, systematic misunderstanding of other people’s views (in the direction of believing them to be more extreme), and vicious attitudes including openness to political violence. And after his book was published, a more recent study has shown that consuming news on social media leads to uncivility and unfriending. Again, I must point out that what’s true about the population in general may not be true for EAs. Maybe when we get more politically engaged, we become more understanding and less stressed. So there is some uncertainty here. And as we hold our EA identities more firmly than our political identities (I hope), we should be able to escape some of the madness. Freiman says that political debates would go more smoothly if we had “apolitical politics”, and suggests that you “quit your political party and join a local effective altruism group in your newly spared time”.
(Edit June 2021: as a counterexample to the general trend of political engagement leading to polarization, a more recent study by Kalla and Broockman has found that when activists become less affectively polarized when they participate in voter outreach campaigns featuring deep two-way conversations trying to persuade opposing voters.)
As I said previously, regardless of whether we go so far as to quit political parties, clearly we can at least give primacy to our EA identities over our political ones, and that should help us be more detached. That said, we cannot handwave these issues by insisting that we’re going to do everything perfectly. It’s hard enough to try to become meaningfully better informed and more rational at politics than the rest of the electorate; can we also develop a strong resistance against all the petty problems of political culture? Even if we could, how much additional effort would that require? It’s just a lot to take on. I trust EAs to be above average at politics just like I trust us to be above average at most things that we try to accomplish, but ceteris paribus, politics undeniably imposes more psychological and social costs than most cause areas. Of course, the displeasures of political engagement are not themselves the problem—they are trivial compared to the positive impacts that we can achieve through politics on rest of the world. The problem is that they can actively inhibit our altruism, both by degrading our personal characters and by troubling the EA community.
Freiman doesn’t ignore the idea that sometimes we have genuine reasons to discriminate against people with different political views. Sometimes people have genuinely vicious political beliefs and we can be morally justified in cutting them off. Freiman avoids extremes—don’t assume that it is evil to have the policy opinions of the other side, and “don’t be friends with Stalin”—but leaves out the big thorny question that is pertinent to America today. What are we to make of the nontrivial minority of Americans who do not endorse atrocities, but still disrespect democracy, tolerate political violence as a legitimate tactic, judge people for their skin color, and/or enjoy the suffering of voters on the other side? Freiman tells us not to be one of them, but doesn’t tell us whether we should tolerate them.
If we really believe that ignorance is bliss, then maybe we should be okay with naively treating such people as moral equals, as long as we can’t effectively do anything about the problem. I’m not sure if I’m comfortable with that sort of utilitarianism. But either way, at the same time that political literacy exposes us to the viciousness of some of our opponents, it can blind us to the viciousness of some of our allies. And while the majority of vicious citizens in America are probably on the right, there are some on the left too. So the politically disengaged may really have the best handle on this. There are plenty of theories arguing why majorities of Republicans or Democrats are vicious people, but if you can’t explain such a theory simply and convincingly to someone who is politically disengaged, then it probably isn’t true.
Conclusion: is this book correct?
Freiman shows a number of obstacles to effective political engagement, but overstates the case in some ways (edit, June 8, 2021: and the theory of motivated cognition which forms a significant part of his argument has been undermined by two new studies published since he wrote the book). He does not succeed with his ultimate argument that politics is so ineffective that it is generally morally wrong. A clearer cost-effectiveness estimate suggests that politics could be a viable way for many EAs to use a small minority of their time and could be a viable thing for a small minority of EAs to focus much of their time on, and EAs can and do engage in politics without becoming overly vicious and divisive. But Freiman does make an adequate case for the idea that it’s okay to ignore politics. Even if one estimates that politics is slightly more cost-effective than other EA areas, it’s close enough and there are enough indirect downsides that it’s within the realm of things that must be socially accepted as optional for different EAs who have different beliefs and different levels of personal fit.
Postscript: is this book responsible?
It is sort of taboo to discourage people from voting, at least in America, and Freiman’s book is opposed by a strong conventional wisdom. Freiman seems to attribute this phenomenon to bland civic ideology. But I think there’s something more substantive at play. The project to get everyone to vote is fundamentally liberal and progressive, whereas political stasis inherently serves the ends of the political right. Restrictions on American voting rights have typically been reinforced by conservatives, from 19th-century racial prohibitions to modern measures regarding mail-in ballots. So people who believe that right-wing politics are harmful will be understandably suspicious of rhetorical attempts to discourage people from voting.
Still, a book like Freiman’s is more likely to be read by younger, higher-educated people who are disproportionately likely to vote for Democrats. So there is cause for concern if you think that the Democratic Party really is superior.
And aside from partisan impacts, lower turnout can affect the political system by concentrating more political power in the hands of highly engaged political junkies. This does not lead to epistocracy, it leads to dysfunction. As we have seen from Freiman and Hannon’s arguments, the people who are very enthusiastic, informed, and knowledgeable about politics generally aren’t particularly responsible voters. If we cede more power to them then we will get a more extreme and polarized political system.
But before we get very worried about these things, we should appreciate the upsides of the book making people more aware of the dangers of motivated cognition, more willing to detach themselves from political identity, and more interested in nonpolitical forms of beneficence. And above all we should appreciate the book’s refutation of certain criticisms of Effective Altruism and its effort to persuade more people to adopt our recommended charitable practices. (Although it does make the argument that we are only morally required to donate to effective charities as a substitute for donating to ineffective charities, not required to donate to effective charities as a substitute for personal spending; I and some other EAs disagree with that notion.)
Review: “Why It’s OK To Ignore Politics” by Christopher Freiman
Why It’s OK To Ignore Politics is a short philosophy book about the merits of political activities, such as voting and donating to political campaigns. It considers and rejects arguments that we are obligated to engage in politics, then goes further and argues that political engagement is positively unethical in some ways.
I think the book does show that it is okay to ignore politics, as long as it means that you are spending extra time and money on other altruistic activities instead. But I am not convinced that it is unethical to engage in politics, as long as it is done with good sense and in moderation; politics should still be considered one of many viable options for EA activity.
Introduction
The book includes three chapters which reject nonconsequentialist arguments in favor of political engagement: the idea that it’s unfair to ‘free ride’ on the benefits of other people’s political activity (chapter 4), the idea that people who do not act in politics are complicit in injustice (chapter 5), and the idea that it’s virtuous to express certain political messages (chapter 6). I did not take the time to read these chapters because I do not subscribe to those ideas in the first place. Christopher Freiman is a professional academic philosopher so I presume his arguments are good. They may be helpful in refuting philosophical criticisms of EA’s lack of inherent political engagement, especially as Freiman himself is basically supportive of the EA project.
He makes three other arguments which are straightforwardly applicable to consequentialist thought (but can apply to nonconsequentialist thought as well). First, that humans are deeply flawed at the task of political reasoning, so that our political judgments are no better or little better than chance; second, that the chance of making a difference in politics is so low that it fails to meet the Effective Altruist criterion of being a productive use of time and money; and third, that political engagement is personally and socially unhealthy.
It’s mainly written from an American point of view but its arguments can broadly apply to other democracies.
Are we all partisan hacks?
In chapters 1 and 2 Freiman lays out the problem of obtaining adequate information and making wise judgments on political issues. He points out that political issues are very complex and most people are terribly biased on political issues. I have quibbles with a few of his specific points and haven’t looked into many of the psychology studies he cites, and it would be sensible to also mention the fact that many voters are not just misinformed but also downright irrational and chaotic, for instance by voting for candidates who go against voters’ stated policy preferences or picking candidates for baseless emotional reasons. Regardless, I don’t think this line of thinking is directly relevant because the whole point of politics is simply to be better than other voters. I may not have a serious and rational approach to making a political impact, but if other voters are even worse than me, then my voting will make things better on average. If it turned out that typical Americans are very rational and well-informed about politics, that might actually dissuade me from caring about it. The fact that most people get politics badly wrong is equally interpretable as a reason to be optimistic about making a positive impact in politics.
What we really need to know is whether we can identify certain people (ourselves, or other people who we might trust) as having substantially better political judgment than most other people. Some of what Freiman writes helps address this.
Information and intelligence
One possibility is to just think to ourselves about whether we seem to know more than others. Freiman cites Anson (2018) about the Dunning-Kruger effect in a political knowledge quiz, finding that most people were overconfident in their political knowledge. But the study still finds that the highest scorers were more confident in their knowledge—in fact, they were underconfident. As I understand it, this is a common pattern with the Dunning-Kruger effect. The upshot is that we should be more dismissive of our own knowledge if we find ourselves feeling moderately better-informed than average, but if we feel much better than average then we are probably at least somewhat better than average, and likely even better than we think.
We can also look at things like IQ or SAT tests, but those also have dubious relevance. Freiman cites West et al (2012) who found that higher-SAT-scoring individuals were more prone to overestimate the degree to which they were less biased than other people, but they still seem to have found that more cognitively sophisticated individuals were slightly less biased on the actual tests that they were given than other people (although not by nearly enough to explain their higher opinions of themselves). In any case, they note that this finding is at odds with most literature—they point out one previous paper which supports the idea that smarter people are more biased, while also listing eleven that disagree. Meanwhile, Stanovich et al (2013), not cited by Freiman, finds that myside bias is unrelated to intelligence.
It would be nice to have a proper metanalysis tying such empirical findings together. But in practice, when it comes to basic political knowledge and susceptibility to bias, we can just take these tests ourselves rather than think about what studies say. That way we are experimenting on ourselves so that we don’t have to worry as much about the generalizability of results. And we can either get a high score, or find other people who get a high score and just listen to them.
Motivated cognition
The more difficult problem raised by Freiman is that more informed and intelligent people don’t appear to use their powers to actually make better judgments. Or do they? Freiman cites Kahan et al (2017) who shows that more numerate people display more politically biased interpretations of data, but as is apparent from Figure 7, they do still find that more numerate people have more accurate views even when the data goes against their side—they just get even more accurate when the data supports their side. So, there is a rather difficult question here about who is more trustworthy on politics: the person who is 35% accurate when it hurts their cause and 40% accurate when it helps their cause, or the person who is 50% accurate when it hurts their cause and 85% accurate when it helps their cause? (My followers are divided on this, but tend to prefer the latter. ) Then at the highest levels of numeracy, it seems that greater numeracy objectively increases accuracy all around without significantly affecting their partisan bias (see Figure 6).
Freiman also writes “training subjects on logical reasoning didn’t enable them to overcome their partisan bias in evaluating arguments” but the paper in question only found that people who went through logic training displayed a partisan bias, it did not make a valid comparison of trained and untrained subjects from the same sample.
Freiman also cites Kahan (2015) who finds that science literacy is correlated with greater polarization on climate change. The gap between Democratic and Republican recognition of the mechanism of man-made climate change is nonexistent among people who are scientifically illiterate, but grows monotonically as we look at people who are more scientifically literate. And in particular, Republicans become monotonically less likely to recognize man-made climate change when they become more scientifically literate. So Republicans and Democrats still get more polarized even as they go from the 90th percentile to the 99th percentile of scientific literacy. But Kahan goes to some lengths to argue that this doesn’t mean what it sounds like it means. It seems like the main issue here is that scientific literacy is presumably correlated with political literacy, and people who are more politically literate do a better job of recognizing what their political party believes and deferring to that. Low-info conservatives don’t know what the Republican Party thinks about climate change, so they express ordinary views. High-info conservatives know that every good Republican denies man-made climate change, so they deny it too.
Freiman cites Bolsen et al (2015) as showing that better education increases viewpoint polarization. While this study does show this dynamic for the claim that global warming is happening, it turns out (rather bizarrely) that low-info conservatives are more likely than low-info liberals to believe that global warming is anthropogenic, and as voters become more knowledgeable, viewpoint polarization decreases because conservatives become less likely to believe in anthropogenic global warming.
Freiman cites Drummond and Fischhoff (2017) as showing that science literacy and education are correlated with greater viewpoint polarization about politically polarized scientific issues, through an analysis of the General Social Survey. This includes greater polarization about support for stem cell research, but that could plausibly be a legitimate case of more informed conservatives learning about how something violates their moral values. They also find greater polarization about belief in the Big Bang and human evolution, but only education and science education are correlated with conservative disbelief in these cases; science literacy increases polarization but still correlates with conservatives being a bit more likely to believe in these things. Finally, they find greater polarization about climate change, but with caveats. First, as the authors point out, “the item on climate change asks participants to assume that climate change is happening and then to express their level of concern, despite the well-documented finding that a significant minority of Americans dispute its existence”. Second, while education increases polarization, it still makes conservatives a bit more concerned. Third, science literacy only has a modest effect on polarization because low-literacy conservatives are actually more concerned than low-literacy liberals. As people get more literate, things flip, with an end result of only slightly more polarization. So on this issue, going from low science literacy to average science literacy reduces polarization, but going from average literacy to high literacy sharply increases it.
It’s also probably true (though Freiman doesn’t cite this) that broadband internet access increases partisan hostility, apparently because the new information allows voters to learn more about what their party and the enemy party are really like. (It isn’t because of ‘fake news’.) And for the most surreal example of polarization-via-information, I recall David Shor theorizing—though I can’t find the source—that the election of Barack Obama increased racial polarization simply because it taught low-info Americans which party was ‘the party for white people’ and which was ‘the party for black people’.
Finally, Freiman doesn’t mention Achen and Bartels (2006) who looked at perceptions of the budget deficit and found that more political knowledge is correlated with greater viewpoint polarization except among the “best-informed 10 or 20% of the public”. Once we focus on that group of highly informed citizens, Republicans and Democrats finally converge a bit towards each other’s factual views as they become more informed.
OK, so different findings contradict each other. And most of these studies are just observational and I have no idea if they will generalize or replicate.
Tentatively, we can say that two things are probably true. First, that more information is associated with more polarized claims until we reach the point of people becoming very highly informed, at which point the impacts of additional information are unclear.
And second, that the fact that voters get more informed about the characteristics of the political parties is of comparable or greater importance to the fact that they get more informed about issues themselves. I predict that if someone does a similar study which controls for political literacy, they will find that scientific literacy is more likely to decrease viewpoint polarization. Of course, that still leaves open the question of why we should become politically literate if it leads us to distort more issues to fit our point of view. But I expect that it also leads to voters becoming more rational at voting for candidates who match their goals and values. If you have right-wing values, then being wrong about climate change is a small price to pay for knowing which politicians will protect guns, restrict abortion, and anger liberals.
Again, maybe someone could do a proper literature review or metanalysis. Michael Hannon’s more recent paper Are Smarter Voters Better Voters? includes more arguments to back up the idea that additional knowledge gets weaponized to entrench or exaggerate polarized viewpoints, but he seems to badly misrepresent Kahan et al (2017), and I just don’t feel up to going through the rest of it.
Edit, June 8 2021: I’ve since come across further studies testing the idea of motivated cognition. Van der Linden et al (2018), Tappin et al (2020) and Persson et al (2021) have tested the theory of motivated cognition and all found problems with the theory. So ultimately there really is not good evidence that more intelligence or information among partisans causes increases in polarization.
Context!
For the actual context of the EA community we should move beyond the basic framework used by Freiman and Hannon.
First, we should be more cautious about assuming that nonpolitical judgments will be unbiased. The studies linked previously compared political judgments to things like the efficacy of skin cream and the probability of having cancer. What we do in EA is not so anodyne. Donating to AMF or The Humane League or MIRI is typically not as tribal as politics but certainly never as clean as textbook statistics. There is still room for tribalism and selfish biases to affect our judgments on these issues. I haven’t seen any studies comparing people’s rationality about politics with their rationality about other forms of do-goodery. It still seems right to suppose that we’re more vulnerable to bias in politics than we are to bias in other cause areas, but this does weaken the argument.
Second, we should probably worry less about our individual cognition and more about our general environment. As I noted previously, the increased polarization from information probably has more to do with the fact that people learn more about political tribes and pick a side accordingly. If we want to do better, we need to ask ourselves: does the EA community do more to reward accuracy and wise judgment, and do less to enforce political conformity, compared to how most other politically relevant communities work? It’s not straightforward to empirically test this but it’s important to consider. And I think it gives more reason for us to be optimistic about our political impact.
Solutions
Freiman later suggests that it would be much easier to talk about politics if we weren’t attached to political sides. In a call for “apolitical politics” at the end of chapter 7, he says that if you really do want to engage in politics, you should at least “quit your political party and join a local effective altruism group in your newly spared time”. Michael Hannon—in “Are Smarter Voters Better Voters?”—similarly argues that we should reduce our attachments to partisan identity. Note that this doesn’t mean rejecting claims that one political party is much better than the other; sometimes that really is the case, and Freiman does a good job of distancing himself from bothsidesist assumptions. This also doesn’t mean that we should reform the political system to weaken or abolish the political parties, nor that we should vote for independent or third-party candidates; Freiman doesn’t mention these (bad) ideas. Rather, the point is that we shouldn’t identify ourselves with a political party or any other political group.
I would say that even if we don’t go so far as to avoid all political identity, we can give primacy to our EA identities over our political ones, and that should help us be more detached. I think I have seen a few EAs who unfortunately care more about political identity than about EA identity but there are many others of us who are more inclined to take a healthier approach.
So while Freiman says that political judgment is so hard that you need to be highly politically informed and comprehensive in order to a good job (see this page for an example of what that can look like in practice—it’s pretty ridiculous), an alternative takeaway is that setting yourself up with the right identity, environment and attitudes is itself sufficient to make you better than most citizens even if you possess relatively little political knowledge and spend relatively little time analyzing. Although of course it can be hard to tell whether you have “the right identity, environment and attitudes”.
We can also simplify electoral debates to a few issues which we know well. From our ordinary involvement in EA, we already know about foreign aid and animal welfare, and while we typically think about these things through the lens of private charity, they also have relatively straightforward implications for government policy. To a lesser extent, many EAs are also well acquainted with issues in housing, criminal justice, and pandemic response. If a politician or party is better on the issues which we do know about, then presumably they’ll tend to be better on the issues which we don’t know about.
Additionally, there is a better alternative to measuring rote political knowledge or philosophical rationality—we can look at forecasting track records. Freiman cites Tetlock’s finding that many experts are terrible at forecasting, but conveniently leaves out the finding that quite a few people have good forecasting track records and/or can be taught to do even better (at least for short-run impacts). In 2020 we even got explicit forecasts for the consequences of the United States presidential election. Of course this can only cover a limited subset of policy impacts, but it still helps a lot.
Finally, sometimes political disputes come down simply to moral values. Freiman points this out as a source of complexity—look, now we have to know about empirical questions and moral ones. But clear moral convictions could make things easier by sidestepping our problems with empirical uncertainty. For instance, if we care equally about people around the world, then we don’t have to worry as much about overspending on preventing climate change. Alternatively, if you’re a pluralist or very uncertain among moral theories, then many political issues will be more complicated for you but you also might have an easier time dismissing political ideas which are very morally brittle.
Directions for further inquiry
Freiman or others may have responses to these points but I feel that we would only end up in circles trying to solve the problem of philosophical skepticism. We can escape that by trying to actually make political judgments in the real world and seeing how well they fare against (a) collective critique and (b) hindsight. Ultimately, the reason I believe that certain political interventions are effective is not based on philosophical notions of my abilities, it is quite simply that I read arguments for and against them and sometimes determine that the former are more convincing. I’ve been making serious political judgments for a while, and have had a few chances to check them against collective critique and hindsight. At this point, I do think that we can have more confidence in AMF than in almost any political cause, and there is greater risk that political efforts cause harm. But I don’t think it’s a big difference. Electoral politics seems like something that’s more speculative than GiveWell charities but still not as speculative as direct x-risk prevention efforts. And there are a few specific policy views which are justified just as robustly as GiveWell charities. And if political inquiry really is very unreliable, then someone should be able to point out serious flaws in my specific political rationales. Or we could look at the investigation which Open Philanthropy Project has already done on certain policy areas.
Edit, June 8 2021: As described above, the evidence for the theory of motivated cognition is really not very strong, so it is fair to presume that gathering information and smart thinkers will indeed produce better political judgments.
Costs and Benefits
In chapters 3 and the conclusion, Freiman disputes the idea that voting is an effective use of time which could be better spend learning and helping with other causes, such as GiveWell charities. He points out that it’s incorrect to say that voting only takes an hour or so because that ignores the cost of obtaining knowledge about who to vote for. If we want to become properly informed about politics so that we can be confident that our votes are doing good rather than harm, that will a much longer period of time. A lifetime of serious political engagement can involve lots of time spent doing things like reading, studying, chatting, and volunteering. And even if you intend to spend minimal time on politics, it might lead to time-consuming habits anyway.
What are the benefits? Freiman correctly points out that a vote in a swing state has a tiny chance of swinging a presidential election: about one in ten million (and in 2020, for most swing states it was actually closer to 1 in 20 million). He then incorrectly argues (citing Dubner and Levitt) that even if you are in a perfectly close election, your vote won’t matter because the election will be decided by recounts and judges anyway. Veterans of debates over meat consumption will recognize the problem here; see Hilary Greaves’ point about vegetarianism. Elections are analogous.
Freiman never presents a true cost-effectiveness calculation, but to his credit he estimates the opportunity cost of political engagement. One hour of political engagement per month starting from the age of 20 (heaven knows many of us are guilty of more than this), sacrificing the average American wage of $22.50 per hour, with an investment return rate of 8%, implies a loss of $70,000 by the time you turn 60. If you donated that much money to a GiveWell-recommended charity, which saves lives for $4,000 each, then you could save 17 lives. But he neglects the possibility that in 2060 it may not be possible to save lives for $4,000 each. Of course maybe it will actually get cheaper (though I really doubt that). But clearly this comparison depends partly on the virtues of patience and giving now versus later. For a clearer comparison we should look at immediate impacts from both efforts.
I have a calculation based on the 2020 presidential election which suggests that the average swing state vote in a near future presidential election will have a direct impact which is just as good as 0.02 lives saved in expectation. This counts all direct human and animal social impacts which can be decently estimated. It excludes some speculative benefits including long-run foreign policy issues and the probability of nuclear war. Considering that the cost per life saved through Givewell is around $4,000, a swing state vote is worth $80 from a typical Effective Altruist. That’s equivalent to an unimpressive 3.5 hours of time at $22.50/hour. Keep in mind that these numbers are not easily comparable—the GiveWell figures are more robust, but my figures exclude numerous long-run benefits from things like sound foreign policy and the integrity of American democracy. Of course we should also discount these for deep uncertainty about political consequences—maybe assume a 5 percentage point greater chance that our political actions are net harmful compared to our charity efforts, so we should discount these estimates by 10%.
Still, this seems good enough to justify the practice of voters in swing states taking a rudimentary look at what the EA community or other experts believe and voting accordingly, while failing to justify any practice of being seriously politically informed. Even a project of becoming seriously informed in order to tell other EAs how to vote might not be justified at this level.
Freiman also points out that even those such as William MacAskill who support swing-state voting still acknowledge that voting for the president in safe states (which includes most Americans) is pointless. And this is fair. The EA community seems particularly underrepresented in swing states.
But things look better if you add other elections. Freiman dismisses congressional elections on the basis that they are frequently safe incumbencies, which seems true, and Congress doesn’t get much done anyway, so Congress is probably a lower priority than the presidency. A similar mindset to presidential elections seems appropriate: if you live in a close district for an upcoming election, then take a bit of time to choose and vote; otherwise don’t worry about it.
Then with state and local elections, we can have still more impact. These are obviously less important but also more tractable. In fact there are reasons to believe that these should be a higher priority than federal elections (I intend to share analysis of this soon).
In deciding whether to vote, we only need to find one election on the ballot that really matters. If you live in a presidential swing state or a state with a close Senate race or a state with a close gubernatorial race or a city where you can find a good person running for city council or … then you can justify a basic level of familiarity with political issues and time spent voting in at least some of the elections. And while you’re at the polls or filling out a mail-in ballot, you may as well cast a vote in all the other simultaneous elections as long as you have any real inclinations for certain candidates.
When you add different elections up, maybe the average direct impact of a ballot even in a safe state is equivalent to 0.08 lives saved, which is like donating $290 to a GiveWell charity, or 13 hours of time. Now it is looking more reasonable to be mildly politically informed.
Note that in some circumstances like recent U.S. presidential elections, the case for voting for Democrats is highly robust. There really seems to have been a stronger consensus among economists and political scientists about the threat of Donald Trump than about the merits of RCT-driven development. But the majority of elections aren’t so stark.
Freiman doesn’t say anything about direct activism, advertising, lobbying and campaign donations. He dismisses protests by saying that adding an individual to the size of the group has barely any effect, but that can be challenged; again I intend to do a cost-effectiveness analysis of this soon. But they offer further opportunities to make a significant difference, assuming that you actually do them in prudent ways that don’t cause political backlash (a nontrivial problem). And the more opportunities one has to leverage political knowledge into taking a variety of productive actions besides voting, the more one can amortize the upfront costs of becoming generally politically wise.
Is ignorance bliss?
Beyond time and money, political engagement has indirect costs. Freiman points out that political engagement is linked to stress, degradation of families and friendships, employment discrimination, relationship discrimination, systematic misunderstanding of other people’s views (in the direction of believing them to be more extreme), and vicious attitudes including openness to political violence. And after his book was published, a more recent study has shown that consuming news on social media leads to uncivility and unfriending. Again, I must point out that what’s true about the population in general may not be true for EAs. Maybe when we get more politically engaged, we become more understanding and less stressed. So there is some uncertainty here. And as we hold our EA identities more firmly than our political identities (I hope), we should be able to escape some of the madness. Freiman says that political debates would go more smoothly if we had “apolitical politics”, and suggests that you “quit your political party and join a local effective altruism group in your newly spared time”.
(Edit June 2021: as a counterexample to the general trend of political engagement leading to polarization, a more recent study by Kalla and Broockman has found that when activists become less affectively polarized when they participate in voter outreach campaigns featuring deep two-way conversations trying to persuade opposing voters.)
As I said previously, regardless of whether we go so far as to quit political parties, clearly we can at least give primacy to our EA identities over our political ones, and that should help us be more detached. That said, we cannot handwave these issues by insisting that we’re going to do everything perfectly. It’s hard enough to try to become meaningfully better informed and more rational at politics than the rest of the electorate; can we also develop a strong resistance against all the petty problems of political culture? Even if we could, how much additional effort would that require? It’s just a lot to take on. I trust EAs to be above average at politics just like I trust us to be above average at most things that we try to accomplish, but ceteris paribus, politics undeniably imposes more psychological and social costs than most cause areas. Of course, the displeasures of political engagement are not themselves the problem—they are trivial compared to the positive impacts that we can achieve through politics on rest of the world. The problem is that they can actively inhibit our altruism, both by degrading our personal characters and by troubling the EA community.
Freiman doesn’t ignore the idea that sometimes we have genuine reasons to discriminate against people with different political views. Sometimes people have genuinely vicious political beliefs and we can be morally justified in cutting them off. Freiman avoids extremes—don’t assume that it is evil to have the policy opinions of the other side, and “don’t be friends with Stalin”—but leaves out the big thorny question that is pertinent to America today. What are we to make of the nontrivial minority of Americans who do not endorse atrocities, but still disrespect democracy, tolerate political violence as a legitimate tactic, judge people for their skin color, and/or enjoy the suffering of voters on the other side? Freiman tells us not to be one of them, but doesn’t tell us whether we should tolerate them.
If we really believe that ignorance is bliss, then maybe we should be okay with naively treating such people as moral equals, as long as we can’t effectively do anything about the problem. I’m not sure if I’m comfortable with that sort of utilitarianism. But either way, at the same time that political literacy exposes us to the viciousness of some of our opponents, it can blind us to the viciousness of some of our allies. And while the majority of vicious citizens in America are probably on the right, there are some on the left too. So the politically disengaged may really have the best handle on this. There are plenty of theories arguing why majorities of Republicans or Democrats are vicious people, but if you can’t explain such a theory simply and convincingly to someone who is politically disengaged, then it probably isn’t true.
Conclusion: is this book correct?
Freiman shows a number of obstacles to effective political engagement, but overstates the case in some ways (edit, June 8, 2021: and the theory of motivated cognition which forms a significant part of his argument has been undermined by two new studies published since he wrote the book). He does not succeed with his ultimate argument that politics is so ineffective that it is generally morally wrong. A clearer cost-effectiveness estimate suggests that politics could be a viable way for many EAs to use a small minority of their time and could be a viable thing for a small minority of EAs to focus much of their time on, and EAs can and do engage in politics without becoming overly vicious and divisive. But Freiman does make an adequate case for the idea that it’s okay to ignore politics. Even if one estimates that politics is slightly more cost-effective than other EA areas, it’s close enough and there are enough indirect downsides that it’s within the realm of things that must be socially accepted as optional for different EAs who have different beliefs and different levels of personal fit.
Postscript: is this book responsible?
It is sort of taboo to discourage people from voting, at least in America, and Freiman’s book is opposed by a strong conventional wisdom. Freiman seems to attribute this phenomenon to bland civic ideology. But I think there’s something more substantive at play. The project to get everyone to vote is fundamentally liberal and progressive, whereas political stasis inherently serves the ends of the political right. Restrictions on American voting rights have typically been reinforced by conservatives, from 19th-century racial prohibitions to modern measures regarding mail-in ballots. So people who believe that right-wing politics are harmful will be understandably suspicious of rhetorical attempts to discourage people from voting.
But universal vote by mail doesn’t actually benefit either party, and either party could have benefited from high turnout in 2020. David Shor thinks that mobilization in general still tends to be good for the Democrats, but clearly it’s not a strict relationship.
Still, a book like Freiman’s is more likely to be read by younger, higher-educated people who are disproportionately likely to vote for Democrats. So there is cause for concern if you think that the Democratic Party really is superior.
And aside from partisan impacts, lower turnout can affect the political system by concentrating more political power in the hands of highly engaged political junkies. This does not lead to epistocracy, it leads to dysfunction. As we have seen from Freiman and Hannon’s arguments, the people who are very enthusiastic, informed, and knowledgeable about politics generally aren’t particularly responsible voters. If we cede more power to them then we will get a more extreme and polarized political system.
But before we get very worried about these things, we should appreciate the upsides of the book making people more aware of the dangers of motivated cognition, more willing to detach themselves from political identity, and more interested in nonpolitical forms of beneficence. And above all we should appreciate the book’s refutation of certain criticisms of Effective Altruism and its effort to persuade more people to adopt our recommended charitable practices. (Although it does make the argument that we are only morally required to donate to effective charities as a substitute for donating to ineffective charities, not required to donate to effective charities as a substitute for personal spending; I and some other EAs disagree with that notion.)