Complexities of Free Speech: Some Approaches, Problems, and Reasons to Care

Hear this article read aloud here!

Author’s Note: this post based on an old Discord rant

Free speech comes up in arguments a lot today, and I have been disappointed by the way I often see it discussed. After some reflection, I think something like this is my key complaint: The conversations are usually about whether something fits the definition of free speech, or the definition of violating free speech. I rarely see people arguing whether something fits the reasons behind free speech or not, despite these clearly being what is relevant to discussing things on these terms in the first place (see also the noncentral fallacy).

This is one of the biggest reasons I still recommend, and it seems I’m not alone in this, John Stuart Mill’s “On Liberty” today even though it is older and contains mostly intuitive arguments on this subject (If you are ok with not reading it as a physical book, it’s public domain, so I would feel dishonest if I didn’t point out that you can also find it free and legal online as both an ebook and an audiobook). This book is still important because it lays the reasons for caring about free speech out. It lays them out like it’s trying to convince people, because it is.

Reading things like this may help encourage people to discuss these issues on a more basic and frank level, to break the taboo on asking what is actually at stake in free speech, and how the things at stake are not possible to insulate within a single definition we can argue over instead.

The case for free speech essentially rests on the idea that it is epistemically irresponsible to bury an argument because of the conclusion it supports. I won’t summarize the whole argument, but a wide range of pretty intuitive reasons is laid out. As an example, you need to be able to be persuaded that a view is wrong in case it, in fact, is wrong, but you also need to protect correct views from irrelevance or corruption by challenging the reasons behind them so they aren’t forgotten. Positions maintained by rote are vulnerable, both to sudden attacks by bad ideas, and to misuse when novel situations come up. Plus of course, if a view does not have a chance of being adopted, while the importance of defending it is reduced, so are the stakes of allowing it to be voiced. The chance that a view will never have a shot is functionally irrelevant since it doesn’t give you a reason to change your treatment of it.

These are some of the classic arguments in favor of free speech, and though it is worth examining, challenging, and adding to them, I think they provide a decent starting point. One thing recasting free speech as the product of reasons like these does immediately is to point out some of the conversations that don’t seem to fit with free speech at all:

  1. Some things people defend by calling them “free speech”, actually might not qualify for the type of consideration we give many forms of “speech” (an example here is the sort of “freedom to offend” discussions. Beliefs that might be offensive should be viewed as within the scope of free speech, arguments that might be offensive should too, but beyond that, free speech is a categorically different issue from whether it’s ok to, say, wear black face, or tell a cringey joke, or swear in certain contexts. The reasons behind free speech specifically concern arguments and positions. If these other offensive things are worth protecting for some reason, it is not for the same reasons as free speech).

  2. Some things I hear people argue don’t violate free speech actually might in a foundational sense (for instance when people say that they have a right not to listen to someone, or to boycott them, or to not invite them places because of their views, they are correct in a legal sense, but the reasons why free speech is uniquely important are still being ignored. As I’ve previously stated, the point is that it’s epistemically irresponsible to silence or dismiss an argument because of the view it supports).

Now, mind you, this isn’t all to say that free speech is unambiguously good either, we shouldn’t treat it as sacred when discussing it. The reasons I have given seem compelling, but stripped down, the arguments are predictive in nature. They make the case that an idea’s success is related to its merit in a way that can’t be reliably improved on with restrictions.

A particularly tricky challenge to Mill’s arguments comes from asking about a specific view you disagree with rather than a specific view you agree with. Mill’s arguments about uncertainty seem to be appealing in large part because of how specific our views are. You should accept that, given how many possible answers there are out there that other people accept, some or even most of your beliefs are probably wrong, and the truth, or at least some missing part of it, is among the views that disagree with yours. If the argument is, should we restrict speech that challenges our views, this makes plenty of sense, but it becomes harder to invoke when someone says that a particular view is worth restricting. For similar reasons to why you should think that your own particular view is incorrect in some way, you could say that this specific view you disagree with is also probably incorrect. Indeed, since the same argument about specificity applies to this other view, and in addition you currently believe that it is wrong, you have an unusual amount of evidence to say that a given specific view you disagree with is probably wrong. While you can’t apply this to every view you disagree with without removing all challenges to your current views, this does present a serious challenge to the standard arguments against restricting any specific view. The uncertainty argument in favor of free speech is in a big way an argument that you shouldn’t listen to only your favorite views, it does not do an exceptional job explaining why it should be wrong to only ignore your least favorite views.

A difficult issue comes out of the predictive framing, while you should allow free speech if you think that the consensus reached will end up being the right one, even if it’s not the belief you currently hold, what if, even after carefully and impartially considering it, you think this isn’t true for a certain view? That if it enters discussion, it has a greater chance of being accepted and being wrong than of being accepted and being right for example (of course the full framework would have to be more complicated, for instance, would it be much worse to accept a view if it is wrong than to fail to accept it if it is right?)

Contrary to much conventional wisdom, the best classic arguments in favor of free speech seem to rely on the assumption that ideas are dangerous, not that they are safe (recall that the case in which views are truly hopeless can safely be ignored as long as there is a chance that the view is not hopeless. Indeed if discussion couldn’t change anything, there would be little reason to protect speech even if there would be little reason to restrict it either). This insight, arguably, provides a good deal of defense for freer speech on some level. For instance it requires more of free speech skeptics than a firm belief that their view is correct. It also requires them to think about how likely they think an opposing view is to be accepted, and whether, if it ends up being accepted, it is still more likely that it is wrong than right. Still, while I think that we should probably all be more uncertain, even about our most fundamental assumptions, you would have to be a hell of an optimist to say that a situation in which a bad idea is more likely to be wrong than right if it wins people over is entirely inconceivable.

There’s a thought experiment I have heard people make a few different ways, though I have had trouble tracking it down, which tries to bring home the intuition that some speech is intolerable in its danger, or at least that resistance is understandable. It goes something like this. Let’s say you are tied up, and in a room with you are a number of instruments of torture, and two people having a debate about whether to let you go, or to torture you to death. Perhaps this debate seems like a good thing if the person who has the ultimate choice is the one who wanted to torture you to death in the first place, but it is not so easy for you to say in your position that if the person with the ultimate choice was the one who wanted to let you go, that they should listen to the torturer for the sake of free speech. Since lives, freedoms, and wellbeing are all at stake in most political conversations, it is easy to feel used sometimes, like free speech is a game intellectuals play with real people as chips. I can appreciate that casual support for unconditional free expression might even feel suspiciously callous. This is ultimately the seriousness of free speech discussions, admitting the power of free speech is what gives both the critics and defenders a sense of urgency.

This is trickier, and involves the separate question of how bad it will be for other reasons to control speech in a certain way (for instance who your decision gives power/discretion, or an excuse for their own similar actions to. To build off of the last thought experiment, let’s say that instead of being the person tied up with a person who wants to free you debating with the torturer, you were someone tied up in the next room, waiting for the two to come in and not sure whether the torturer or the freer will have the final choice. All you know is that whoever chooses whether they should debate in the room before you will also be choosing whether the two debate when they get to you, or whoever happens to have the choice when they get to you immediately does what they want).

This is the level on which I think government regulation and simply refusing to invite someone to speak in the first place are different, not in that one is hurting free speech and the other isn’t, but that while both hurt free speech, one is just more likely to end terribly. Still, my current framing, unfortunately, just gives a set of questions and contingencies that could operate on someone’s current credences, rather than anything like a universal heuristic for how to treat free speech in a wide range of cases.

[Ed. Note: If you, the reader, were hoping for a more comprehensive solution in this article, go back and read the title.]

Different levels of faith or distrust in the ideas market could lead people to the views ranging everywhere from – we should default to empowering all arguments that present themselves, to – we should entirely halt, violently if necessary, all important topics of discussion. I think keeping the government out of speech as much as possible is one worthwhile rule, but how about how we treat discussion outside of a legal context? When we are just choosing who to listen to, who to invite to speak, who to boycott? There are some intuitive heuristics that seem appealing, but in my opinion they mostly fail.

The idea that free speech should defend itself against speech that threatens it, for example, isn’t exactly paradoxical, but feels hypocritical at least. Even a relatively coherent system for doing this based on infinite recursion, such as allowing only free speech that doesn’t threaten (speech that doesn’t threaten (speech that doesn’t threaten (… ))), uses as a central argument in its favor that the current rules of discourse can’t be perfectly trusted. This means that the very thing it is designed to prevent, opposition to free speech, gets its appeal from the same source as this heuristic itself. Since adopting this heuristic would provide support for a key reason behind the thing it opposes, it requires a huge amount of self-certainty, more than most rules, to justify its particular answer to these reasons.

[Ed. Note: Quick primer on why the commonly-used version of that argument, Popper’s Paradox of Choice Tolerance, is wrong.]

Another possibility is the idea that free speech shouldn’t argue against something really morally essential, like whether a group belongs in our moral circle for making judgments at all (this is my attempt to define a consistent standard of what is often called “hate speech”. As I hope to show, it doesn’t work that well, but without some consistent standard there is little reason to separate “hate speech” from speech that is worth restricting in general, and I believe the other ways of formalizing “hate speech” I don’t mention here have serious problems as well). This one also has a number of issues. One is that it either restricts all moral circle changes to our exact current one, or lopsidedly protects those in the current moral circle without stopping moral circle expansion. The first seems almost trivially bad if we look at history, what are the odds that we have no moral growth left to do? The other possibility isn’t great either though. If you admit to imperfections in the popular development of ideas, as you would have to in order to justify any restrictions on speech, this lopsided treatment nearly assures a permanent victory for the moral patienthood of everyone and everything to eventually enter the conversation, including almost certainly incorrect categories like rocks and fictional characters. Perhaps this is a result worth the price, if exclusion from the moral circle is bad enough. Given this type of defense however, the much greater weight it gives the risk of not including a group in our moral circle than the risk of including it, it seems like it would also indicate that we ought to jump the gun and accept the moral patienthood of rocks right now to be safe. This in itself is not a fatal problem unless you are an absolutist however, as you could say that society should just be much more careful about shrinking its moral circle than growing it, which does seem like a good idea to me (and to expand the previous logic, there should be much stronger arguments required to justify current rejects from our moral circle than current members of it).

There is something possibly even more concerning in this heuristic as well however. It is unclear to me how this could be turned into the type of universal standard that would justify specifically promoting it without making it effectively useless (this is where other definitions for hate speech may have fewer problems, though with important trade-offs in flexibility I won’t get into here). Characterizing a certain view as bigoted requires a value judgment unless the accused bigot openly says that they exclude the discussed group from their moral circle, so to make the heuristic plausibly universal, it would need to judge someone’s moral circle based on self-declaration. Making the restriction based on self-declaration would make the restriction useless even against a defender of slavery who says that they think slavery is what is best for their slaves (or to go back to the earlier thought experiment, it would make the torturer still just as untouchable in their position if they said that they thought you were a moral patient, but nonetheless thought torturing you to death would be good for you somehow).

Finally, what seems like one of the best possible heuristics, though still flawed, is to specifically empower those voices affected by the adoption of an idea. In particular if we return to the torture thought experiment, it seems like one of the best answers we can give to it is that, whoever is given the choice about how to treat us, whether it be the torturer or the freer, they should listen to us more than the other person in the room.

In the real-world this also creates incentives that help weaken bad arguments. This heuristic would no longer be content with the “slavery is good for the slaves” argument the moral circle one is vulnerable to, because if the slaver makes an argument where it is the good of the slaves that is relevant to the issue, they are functionally suggesting that you should listen to the slaves rather than the slaver. Instead, in order for the slave owner to dodge this filter, they would have to be honest, and say that what they care about is how they will be personally affected by the freeing of their slaves. The argument from the slave owner this heuristic would consider most valid is the one that is the weakest.

This heuristic may be annoying for most people with strong opinions, because unlike with the other heuristics, there is bound to be some area where this heuristic says they should be listened to less, regardless of their ideology. Still, it seems as though this heuristic has unusual strength compared to the other two I have looked at, and if we must use a heuristic of this sort, this is the one I would promote. It also has a few weaknesses however, that lead me to doubt it as a useful enough solution to the problem at hand.

One stand-out is that it limits the power of expertise in discussion. I didn’t discuss expertise as a heuristic specifically, mostly because it has some things in common with this heuristic and I don’t want to talk about these different possible restrictions all day, but also because it seems harder to make work on its own without inheriting the problems of technocracy. Still, there are possible cases when the nature of the group directly affected in a discussion makes them less competent to choose the best policy.

Children are a key example here, they are usually recognized as serious concerns for moral discussion who are not as trusted in their contributions to these discussions (though for my own part it seems very plausible that this has distorted our moral view of children to the point where they are treated almost like regulated property. Some controversial arguments have even been made that they should be given considerably more voice than they currently are, though I remain skeptical). Children are a good example here because they provide a scale for the question as well. Perhaps you might say that 16 year olds should be trusted more than adults on issues concerning them, but do we trust toddlers more than adults on issues concerning them?

Allowing exceptions based on expertise creates serious vulnerabilities however. The slaver, more than anyone, would be thrilled to see “this group is not competent to speak for themselves” become an argument they can make to a racist society already disposed to believe something like this. Loopholes of this sort almost defeat the whole heuristic if they can’t be carefully checked in some way.

[Ed. Note: Party game: find competence criteria that everyone can agree on.]

Another problem comes from determining who is in the pool of those affected. While I think that this heuristic is a bit harder to troll than the moral circle one, and on many issues a rough pool can be effectively agreed upon, there are cases where people arguing in good faith may disagree. Even cases where this disagreement forms part of the substance of the arguments around the issue.

Abortion seems like a perfect example to me. Who should we include within the circle of those the discussion concerns? One possible answer, I think favored almost exclusively by the pro-choice side, would be those who can have an abortion, most and mostly women. Arguably what would be needed to separate this view from the other possibilities is the assumption that fetuses should not be in our moral circle. Right off the bat we have a disagreement that bleeds into the territory of another heuristic.

Others could argue that some or all fetuses belong in this group as well. Many pro-life activists would argue this case, but I think so would some pro-choice activists, since arguably the most famous argument in favor of the pro-choice position allows for the assumption that fetuses are moral patients. If this second grouping is correct, it leads to a very serious problem for this heuristic, because it is a case where two groups are affected, they have a conflict of interest, and one of them is voiceless. In cases like this, this heuristic no longer has its intended effect because a group is likely to win because it is possible to listen to them, not because they are necessarily in the right.

Finally, a third possible way to draw the line, which I see some pro-life people argue in favor of, is to include everyone. The reasoning here is that everyone was once a fetus, and it is in everyone’s interest whether they had been aborted or not. While this position seems to be one of the silliest to me, because most of us would agree that our parents should have had some choice that, had they made it differently, would have resulted in us not existing, it still seems like it is possible to make this type of case in good faith (a more generous interpretation of this would be if this group roughly agreed with the second grouping, with the difference that they consider you to meaningfully be the same person as the fetus you once were, and so a valid representative of an affected party). In fact, the question it raises of how we ought to treat or deal with the moral status of “possible people” in general constitutes some of the trickiest questions in modern ethics (see part 4 of Derek Parfit’s “Reasons and Persons”). Ethics concerning “possible people” once again runs this heuristic into the problem of affected voices it is impossible to listen to.

Since in cases like this, the relevant arguments about an issue are intimately entangled with the correct way to use this heuristic, there seems to be a serious value-judgment issue baked into the fabric of this rule.

What about if we just allow people to determine their own standards for morally acceptable speech, and which means of restriction of this speech are worth their dangers? One problem is that it would probably make a society more internally hostile, and ultimately more miserable, the more it disagrees. This would create incentives for a society to reach potentially hasty consensuses on important issues that deserve more caution. There is also a smaller scale potential problem within subcultures. Since many subcultures seem to reward commitment to the cause with more influence and status within the subculture, if you encourage a self-evaluated standard of speech restriction that is based on one’s credence in the likelihood of an opposing view being right, members of the subculture may have an incentive to be more and more restrictive against listening to other views, as a signal of this commitment.

[Ed. Note: See also, blue lies.]

In general, it seems to me that the following two things about free speech are probably true:

  1. If we are in a world where free speech leads to the best ideas having the best chance of winning, protecting free speech is more easy and desirable than protecting against free speech would be if we lived in a world where this wasn’t the case.

  2. We probably live in a world where, at least in some situations, this is not the case.

While number 1 is maybe not as obvious given my expansive definition of free speech, it isn’t too hard to see the reasons why it is still true. A shared respect for free speech, especially on this discursive level, gives people more reason to get along even if they disagree, and gives people who disagree a common goal for how to promote truth, and so this method for promoting truth is easier to make large-scale progress on (even if everyone agrees that free speech should be, on some level, restricted, the fact that they disagree on what is right will make it so that none of them make much progress. They are acting on this shared idea about discourse in different ways of necessity). Number 1 seems to be especially true for a society in which just about everyone disagrees on some important topic, and most of them are probably wrong looking at how historically idiosyncratic their views are.

More and more people however, it seems to me, are recognizing point number 2, leaving free speech in a bit of a crisis. Point number 2 is ugly, there’s not much to do with it, because it hampers coordinated action by its nature, it makes each person feel a good deal more powerless and more alone. I tend to think that many people overrate the extent of number 2, looking at history (though keeping in mind the bias that I have a present-centric view of the past, and no present-centric view of the future, since I have no view of the future. I have looked at many of the views of people from history who have had similar fundamental values to me, and it seems to me that discourse tends in mostly pretty good directions. I would encourage other people to do similar research, perhaps not everyone should have the same confidence), it seems like good discourse has a tendency to win strong enough that speech restriction is rarely worth its problems.

Allowing that number 2 is compelling however, it seems to me like the best solution might be to change number 2. To increase our trust in free speech, as we improve the ability of discourse to produce good answers under the condition of free speech. This seems like the best course of action to me, because although we disagree on many matters of importance, it seems like there are many ways we tend to agree much more on fundamental things about discourse.

We should try to research our positions, and provide sources behind our research when we support arguments with it. We should not omit this research when we find it convenient for what we are trying to ultimately defend. We should be self-conscious about what changes our minds, so we can avoid changing our minds for reasons we wouldn’t like. We should look for inconsistencies in our thought. We should learn how to read the story data tells, and avoid reading any further into the data than its scope. Possibly most importantly of all, we should try to hold each other accountable to these types of standards.

The social takeaways of this aren’t always obvious, though I think that there are popular practices in discussion they seriously implicate. For me the stand-out is live debate, which I have to admit to personal bias on because I don’t enjoy or feel like I get much from debating people live myself, and much prefer written exchange. Still, I don’t think it’s hard to see potential problems with treating discussion of issues we take seriously for their practical consequences as real-time sporting events. If both parties go in with a full appreciation for each others’ knowledge and arguments, the interaction is essentially scripted anyway. If they don’t, they will either get wildly off-topic from each other, or have to respond to new ideas and information way more quickly, and in a more motivated way because of the competitive setting, than they might have if the exchange were framed differently.

The idea of improving the conditions of free speech is not a comprehensive answer to the larger issue of free speech. While it could make free speech a better idea, perhaps it will always be a bad idea to personally invite Nazis to speak at events for instance. There also isn’t perfect consensus even on the level of speech and reasoning norms I’m describing. Still I think there is enough consensus, and reliable enough consensus, that improving the conditions of free speech while increasing how well regarded free speech is presents the best large-scale path available on this issue.

, help us write more by donating to our Patreon.