Epistemics (Part 8: Two types of decoupling)

Imagine that an African American man boards a public bus on which all the other passengers are white. Unhappy with the newcomer, an elderly white man, turns to the African American man and says, “Just so you know, because I realize that you’re not very bright, we don’t like your kind around here, … boy. So, go back to Africa!” What the elderly white man said is a lot of things. It is outrageous, insulting, scary, intimidating, hateful, rude, and so socially inappropriate to be unnerving on solely those grounds. It is also harmful.

Mary Kate McGowan, Just speech: On speech and hidden harm

1. Introduction

This is Part 8 in my series on epistemics: practices that shape knowledge, belief and opinion within a community. In this series, I focus on areas where community epistemics could be productively improved.

Part 1 introduced the series and briefly discussed the role of funding, publication practices, expertise and deference within the effective altruist ecosystem.

Part 2 discussed the role of examples within discourse by effective altruists, focusing on the cases of Aum Shinrikyo and the Biological Weapons Convention.

Part 3 looked at the role of peer review within the effective altruism movement.

Part 4 looked at the declining role of cost-effectiveness analysis within the effective altruism movement. Part 5 continued that discussion by explaining the value of cost-effectiveness analysis.

Part 6 looked at instances of extraordinary claims being made on the basis of less than extraordinary evidence.

Part 7 looked at the role of legitimate authority within the effective altruism movement.

Today’s post discusses decoupling. In many recent controversies around harmful speech within the effective altruism movement, harmful speech has been defended by invoking decoupling norms. Roughly, the idea is that the correctness of a claim should be decoupled from the social context in which it is asserted or the harms of asserting it in this context.

I think that this defense of decoupling conflates two types of decoupling. The first, truth decoupling, is a good practice but does not give a blank check to harmful speech. The second, harm decoupling, is a bad practice precisely because it ignores the harms of speech. The project of today’s post is to distinguish truth decoupling from harm decoupling, to say why truth decoupling is a better practice than harm decoupling, and to discuss the role of standpoint in debates about decoupling.

2. Truth decoupling

Truth decoupling involves decoupling the truth of a proposition from truth-irrelevant factors. On one popular philosophical view, the only factors directly relevant to the truth of a proposition are items of evidence which tell for and against that proposition. If this is right, then truth decoupling involves treating evidence for and against a proposition as the sole factor relevant to its truth.

Truth decoupling acknowledges the importance of many facts which do not bear on the truth or falsity of a proposition. However, truth decoupling urges that the importance of these facts should not be confused for truth-relevance.

For example, truth decoupling suggests that the truth of a proposition should be divorced from the person propounding it. The person propounding a claim may be a known liar, cheat, or opponent of bathing, but this does not mean that what they say is incorrect. For example, for all his faults, Donald Trump could well be right (though he could also be wrong) about the need to take a hard-line stance towards Iran. This is not to say that Trump is a good person, or even that his reasons for proposing a hard-line stance are good. It is simply to say that neither Trump’s personal characteristics nor his inability to produce a decent argument rules out the possibility that his beliefs could be true.

Truth decoupling also suggests that the truth of a proposition should be divorced from the savoriness of its consequences. Some time ago, my home state of North Carolina was debating the wisdom of investment in our easternmost barrier islands, given that global warming trends may well leave most of those investments underwater within a century. A number of legislators proposed that these facts about climate change should be given little weight in debates, because if they were heeded the results would be economically devastating for coastal residents. Truth decoupling suggests that this stance was a mistake. It could well be the case that many coastal properties will soon be underwater, even if the consequence of recognizing this fact would be a sharp decrease in property investment and a ruinous hike in insurance rates.

For a final example, truth decoupling suggests that the fact that we want to belive, or are committed to believing some fact does not bear on its truth. It is well-known that most people believe they are good drivers, good lovers, and have above-average intelligence. Their desire to be correct in these assessments does not bear on the likelihood that they are actually correct. Similarly, many atheists have charged that religious beliefs are the benefits of a strong form of conservative bias. They allege that believers deeply wish for God to exist and are committed to God’s existence, so that otherwise-weak arguments for God’s existence are accepted because they accord with believers’ pre-existing commitments and desires. I do not want to pronounce on the truth or falsity of this charge, but it if were true, it would illustrate a tendency towards irrational belief among the faithful.

Truth decoupling is an important philosophical practice to which I am deeply committed. In the philosophy classroom, we encourage students to grapple actively with some of the deepest and most uncomfortable questions about life on Earth. Are we really free to choose our actions in a deterministic universe? If not, how is it possible to be morally responsible for our actions? Does morality even exist? Students are urged to look these questions straight in the eye and to defend stances on each question through the impartial exchange of reasons. The facts that it would be tragic for us to be unfree, for all moral facts to be false, or for no person to be morally responsible for their actions are not accepted as reasons in my classroom.

Effective altruists are often skilled at truth decoupling. This is, I suspect, why many effective altruists read my work despite knowing that they will encounter criticism, often strongly worded. And it is, I suspect, why effective altruists are often among the world’s leading forecasters. Effectively predicting future events requires putting aside personal prejudices to let the facts speak for themselves.

To be honest, many of the highest-profile failures of truth decoupling occur in criticisms of effective altruism. I am often struck by the speed with which critics of effective altruism move between the charges that (a) effective altruists have behaved badly, showing evidence of racism, sexism, and financial mismanagement, and (b) effective altruists make false claims about levels of existential risk or the cost-effectiveness of various interventions. Indeed, I often find myself required to remind skeptics that it is quite possible for effective altruists to be both (a) badly behaved, and (b) correct in their beliefs about philanthropy.

There is nothing wrong with truth decoupling. But truth decoupling must not be confused with a further practice of harm decoupling.

3. Harm decoupling

Harm decoupling involves decoupling the harms of a speech act from the questions of whether and how to engage in it.

Asking and answering questions are actions. Like all actions, these speech acts have consequences. Some consequences are good, others bad. A good consequentialist takes account of all of the consequences of her actions. It is no part of consequentialism to believe that the benefits of a speech act are confined to the amount of truth, knowledge or understanding gained. And it is certainly no part of consequentialism to deny that speech acts can have harmful consequences, or that these consequences bear on the moral status of the act.

Suppose I wanted to study the effects of concussions on the brain. That is a noble goal. Now suppose I were to go about it by punching my research assistants in the face, then measuring the effects. I suspect that my research assistants would protest. I could, of course, truthfully say that by punching my research assistants in the face I was putting myself in a better position to understand the effects of concussions on the brain. I could even add that it is very important for society to understand the effects of concussions on the brain. Yet my research assistants could rightly retort that by punching them I was causing them pain, creating a hostile working environment, and inducing the very concussions whose effects my research would show to be devastating in later life.

Truth decouplers would be correct to insist that none of these harms bear on the reliability of the data gathered by punching research assistants or the validity of inferences drawn from that data. However, harm decouplers would be wrong to insist that the harms of my research procedure should be ignored. Punching people is not a good thing to do, because it harms them. The fact that punching people may advance the cause of truth, knowledge, or understanding does not change the fact that I would be punching people, or that punching people is (typically) wrong.

It is important to recognize that there are many ways in which speech acts can harm. For one thing, speech acts can degrade individuals or groups. If I tell my research assistants that they are nitwits, or believe that they are unintelligent, my language and beliefs degrade them. By the same token, if I say that an entire racial group is genetically predisposed to contain mostly criminals and nitwits, or simply believe that they are genetically predisposed to be criminal and unintelligent, I degrade them. This degradation has tangible harms. I suspect that my research assistants would not appreciate being thought to be unintelligent, and I rather suspect that members of an entire racial group would feel the same.

For another thing, speech acts can reinforce harmful stereotypes. One way that this occurs is in question selection. Suppose I were to ask questions about the intelligence of groups. However, suppose that I took my questions directly from existing stereotypes. For example, I might ask whether women are predisposed to be more emotional than rational, whether negatively-stereotyped racial groups are predisposed to be less intelligent, or whether positively-stereotyped racial groups are predisposed to be more intelligent. In asking these questions and treating them as serious targets for investigation, I would be lending strength to the legitimacy of the stereotypes that generate them. I would also, given the limits of existing scientific methods and the potential for contamination by existing stereotypes, expose the targets of stereotypes to a very real risk that those stereotypes would be falsely supported by my results and analysis.

Another way that speech acts can reinforce stereotypes is through norms and practices of inclusion. Suppose, for example, that a certain discussion were to be open to all. However, participants would have to agree not to complain if they found the acts of others to be offensive and degrading. Furthermore, although the discussion was in theory open to all, it would be conducted in an existing community whose composition is highly unrepresentative of society at large, and in particular unrepresentative of stereotyped groups. This discussion would directly harm the targets of stereotypes by providing them an insufficient and uncomfortable opportunity to participate in discussions, and would likely further harm the targets of stereotypes by coming to false conclusions about them.

Stereotypes around race, sex, gender and other categories feature prominently in systemic patterns of marginalization and exclusion. We live in a society that is deeply structured by categories such as race and gender, and in which these categories continue to shape all aspects of life including income, housing, education, healthcare, and life expectancy. In this system, to reinforce stereotypes just is to contribute to systematic patterns of mistaken thought and poor behavior that continue to harm some of the most vulnerable members of society.

It may seem that mere participation in discussions of stereotypes does not directly harm anyone. That is unlikely, but also misses the point. The very stereotypes being reinforced form the backbone of one of the most oppressive, harmful, wasteful and frankly ridiculous systems of social organization that this world has ever known. In reinforcing these stereotypes, we reinforce the system. That is a grave harm that should be avoided whenever possible.

There is, in certain quarters of the effective altruist movement, a misguided belief that truth-seeking speech acts should be performed at all costs. This doctrine contributes substantially to the open engagement with race science discussed in my series on human biodiversity, as well as to many other questionable behaviors within the movement.

The doctrine that truth-seeking speech acts should be performed at all costs wrongly adopts harm decoupling in place of the correct practice of truth decoupling. I do not mean to suggest that the propositions being investigated are true – quite the opposite, I would urge readers not to convey the false impression that troublesome ideologies capture uncomfortable truths that must be suppressed for social benefit. However, what I want to stress here is that it can and often does make sense to avoid asking many questions in many contexts on the grounds that asking them would do great harm and convey little benefit.

Those who plead that they, unlike their interlocutors, are “high decouplers” miss the point, since they are actually bad decouplers. There is every reason to engage in truth decoupling, taking only truth-relevant factors to bear on the truth of a proposition. But there is no reason to engage in harm decoupling, taking the harms of a speech act to be irrelevant to its propriety. The harms of speech acts are real and important. To `decouple’ from these harms is to make the same moral mistake as one would make by decoupling from the harms of punching research assistants. In this regard, harm decoupling looks less like a virtuous epistemic practice and more like a way of willfully ignoring the consequences of your actions, then complaining that you are being held accountable for them.

4. Standpoint

One of the reasons why harm decoupling may seem superficially attractive to many is that their standpoint makes it more difficult to see the harms being decoupled from. To those not forced to deal with the daily effects of systemic sexism, racism and other forms of discrimination, these harms can feel distant, vague or even non-existent. From this perspective, a discussion about race, gender or similar topics can come to feel like merely an academic truth-seeking exercise among friends. What could be wrong with that?

To those exposed to the daily costs of these discussions, their harms are more immediately apparent. There is little temptation to treat the very discussions which fuel and intensify discriminatory systems as mere academic matters. Unjust systems are created and sustained through biased beliefs, and those beliefs are created and sustained through the very types of conversations about race and gender being challenged.

One example of the importance of standpoint comes in a recent discussion of events at Manifest. A commentator proposed that the events at Manifest should be understood along the lines of the paradox of intolerance: “the more open Manifest is to offensive, incorrect, and harmful ideas, the less of any other kinds of ideas it will attract.”

Oliver Habryka, the CEO of Lightcone Infrastructure, wrote:

I don’t really see the paradox in this case. I am totally OK attending a conference with people who I strongly disagree with or who I think are generally being harmful for the world in a bunch of different ways … The paradox IMO only applies if for some reason the different groups actually start competing or interfering with each other, but I had no run-ins with any of these people, and nobody seemed to do any kind of bullying or threaten violence or anything else that seems like it would actually pose a fundamental conflict here.

Habryka may be quite right that no direct threats of violence or intimidation were present at Manifest. But Habryka is dead wrong to assume that the targets of discrimination will not be driven away from an event absent direct threats to their person. Further, this reply does not engage with the systematic downstream effects of discriminatory rhetoric: the cost of this rhetoric is not merely that minorities may be marginalized at Manifest, but also that they will be marginalized throughout society at large.

Here Shakeel Hasheem, former director of communications for CEA, responded curtly:

Have you considered that the reason you don’t see a paradox here is because you are not one of the minorities targeted by the abhorrent views you and your organisation seek to platform?

What is at issue in this exchange is an appeal to standpoint. Habryka finds it genuinely difficult to see the myriad ways in which racist rhetoric may be harmful. For Hasheem, these harms are so evident that it is genuinely perplexing how Habryka could fail to see them.

Here it is important to distinguish two ways in which our position may blind us to harms. The first is that standpoint may make harms difficult to see. The more that we are shielded from the effects of discrimination, the harder it is to observe these effects. The second is that our perspective may make us actively resistant to seeing the costs of systems from which we ourselves benefit.

A commitment to truth-seeking and harm mitigation should tell especially strongly against active resistance to being shown the harms of our actions. If, indeed, our standpoint makes it more difficult for us to see these harms, then we should be more rather than less enthusiastic about investigating them. And if, indeed, these harms are significant, then a good consequentialist should be still more concerned to understand and mitigate them.

5. Conclusion

Today’s post distinguished two types of decoupling. Truth decoupling involves decoupling the truth of a proposition from truth-irrelevant factors. Truth decoupling is almost definitionally correct, since only truth-relevant factors bear on the truth of a proposition.

By contrast, harm decoupling involves decoupling the harms of a speech act from the questions of whether and how to engage in it. Harm decoupling is a surprising practice in a movement dedicated to effectively promoting good outcomes and avoiding bad outcomes. The harms caused by speech are no less harms for having been caused by speech, and they can no more be ignored than any other harms.

To defend harmful speech as a virtuous form of high decoupling is to conflate the mistaken practice of harm decoupling with the correct practice of truth decoupling. We have no license to ignore the harms caused by our speech, and it is no virtue of decoupling to do so.

One of the difficulties in discussions of decoupling is that standpoint may make harms difficult to see. In these cases, it is especially important to listen open-mindedly for evidence that speech may be harmful. In particular, if many members of the group targeted by your speech vocally complain that the speech is harmful, that may provide good evidence that the speech is indeed harmful.

I hope that future discussions of decoupling will be careful to introduce and defend the precise version of decoupling being invoked. If I am right, many attempts to use decoupling norms to defend harmful speech rely on mistaken forms of decoupling.