The Marketplace of Ideas
Why Free Speech is Broken
The “marketplace of ideas” is an interesting analogy because it has become less abstract over time. As initially conceived, the “marketplace” wasn’t a place at all - rather, it was a description of all of the various activities which people engaged in that involved exchanging and promoting ideas, from intimate parlor conversations with a few close friends to the international publication of literature which might reach millions of readers. Due to the slowness of media before the information age, and the small social graphs most people had before the internet, the “marketplace” was stretched out across space and time so that engaging with the “marketplace” was not at all like going to a “place” and browsing a selection of ideas. Perhaps the closest thing to an actual “marketplace” you might find would be an academic conference or perhaps a political debate, but these were small-scale events by today’s standards and were largely limited to a society’s elites. Most people never got to “visit” anything like a marketplace of ideas - if you weren’t an elite, the best you could do would be to visit the second-hand shop of ideas from time to time and see which trends from previous years had trickled down to your neck of the woods.
Like other markets, the marketplace of ideas has been somewhat democratized by the internet. Now you can go to social media and participate in ideas being formulated, circulated, debated, and filtered to legacy media in real time. Now if you’re on twitter you can watch an actual marketplace of ideas operate, often days before regular people hear a sanitized version of the outcome on TV news or weeks before they read about it in a magazine.
Because social media platforms operate on a much larger scale than legacy marketplaces of ideas (with millions of active users) but are also much more localized (both in time, and in interconnectedness), the characteristics of the marketplace stand out more. Market forces operate with greater power and efficiency than the prior networks of relatively tiny, decentralized marketplaces. The law of large numbers starts to apply and trends regress towards the mean, allowing us to really see what an actual “marketplace of ideas” look like with greater clarity and a greater sample size than any prior market. A social media platform is also much more concrete than “academia” or “the discourse”. Therefore, “the marketplace” is less of an abstraction and more of a description.
We also now know a lot more about markets than we did when the analogy of markets was first applied to human discourse about 100 years ago, and we know much, much more about competition in an open market than we did when the early liberal thinkers described ideas as being in competition with one another.
So, given the wealth of theoretical knowledge and empirical observation, and the expansion of platforms which allow us to more directly observe the operation of the marketplace of ideas, what can we say about how ideas compete?
Fast Food, Fast Fashion, and Fast Ideas
You know where this is going, right? In virtually every country in the world, local traditional cuisine is being displaced by unhealthy Western junk food, causing an obesity epidemic to spread from the US outward to the rest of the world. The more McDonald’s restaurants open in a country, the more its population slowly expands and begins to resemble Americans. Coca-Cola and other sugary, processed drinks bring diabetes. Don’t even get me started on tobacco.
Why is terrible, unhealthy junk food able to outcompete the food that local populations thrived on for generations? Why do populations continue to adopt American eating habits despite the clear evidence that these habits are terribly unhealthy? It’s because people aren’t good at judging what is good for them, and people evaluate what we consume based on factors other than our own long-term best interests. Given the choice we will often consume food that is bad for ourselves and our children, and billions of people will have a shortened lifespan and worse health as a result, and it’s not at all clear that, as a species, we even have the capacity to fight this phenomenon. As much as we all talk about healthy eating, we seem to be getting less healthy.
Fast fashion is problematic for the environment - this isn’t common knowledge, but it’s actually a huge problem. This is just the latest in a long history of market trends and industrial processes that have harmed the environment, and to be honest I mostly picked it because it had “fast” in the name. It could just as easily have been “coal power plants” or “plastic straws”. But the point remains - there’s market demand for fast fashion products, and so the market supplies these products, even though it is bad for the world.
Now, you might think all this is fine. Freedom of choice will always produce outcomes that are suboptimal according to some metric. If we prioritize our short-term enjoyment, we’ll suffer in the long-term. If we prioritize our long-term well-being, we have to make sacrifices in the short term. Perhaps you think we should still give ourselves that choice, rather than, for example, banning junk food, or limiting sugary drinks, or prohibiting alcohol, or regulating clothing factories. But what is clear is that a free market for food produces bad population-level effects. So it merits asking: does a free market in ideas produce bad population-level effects?
Are there ideas which are the equivalent of “fast food”? Easy to get, fast to pick up, tasty, but not very good for you? Well, yes. The idea of fast food, for one. Every physical product has a corresponding set of ideas - the recipe, the IP, the cultural significance, the marketing and branding that people recognize - so we can view product uptake as idea uptake. But that’s a trivial example.
A more interesting example is a stereotype. Stereotypes are typically simple and easy to understand. They’re quick to pick up. They may seem to have some explanatory power and so they may be attractive. They may be trendy - a bunch of people might express the same stereotype. Expressing that stereotype might make you feel knowledgeable or cool. And a stereotype is also “fast”, in the sense that it allows you to make a quick judgment about someone or something. But ultimately they can be bad for you, and bad for social groups, especially when they cause misunderstanding, mistrust, and conflict. If “fast” ideas can produce social conflict, then it’s worth considering whether the intensified marketplace of ideas we now have also intensifies the production of social conflict.
Thinking, Fast and Slow
Another problem with fast food is that it outcompetes and displaces “slow” food, which is often much more nutritious, but also more labor-intensive and perhaps less trendy. Do we have a problem with “fast” ideas displacing “slow” ideas? Yes. Let’s look at covid for an example.
Public health is a staggeringly complex topic, residing as it does at the intersection of public policy and human biology, both of which themselves are extremely complex. Often, our knowledge about covid reflects that complexity. For example, scientists might make a distinction between sterilizing immunity and functional immunity. Learning about this distinction is slow. The public hears “immunity” and thinks that means “invulnerability” or “safety”. This inference is fast. The result is that while people keyed into the science of vaccine development tried to communicate nuance and depth about how vaccines would save lives but we would still need layered protection, that communication largely failed and people largely came to believe that the advent of vaccines would mean we were all safe from covid. Politicians contributed by promising that the vaccines would end covid, or falsely claiming covid was “over” - the informational equivalent of spiking our drinks.
Now, however, it has become reasonably clear that we aren’t safe from covid, so we have a situation where many people believe that vaccines don’t work at all, causing a public health crisis, and many people believe we’re all safe from covid, causing its own public health crisis.
It was like this with all covid communication. Masks, lockdowns, social distancing, handwashing, public gatherings - every topic of public interest had a “slow” version that required some time and effort to understand, and a “fast” version which misled the public and produced endless conflict and confusion. The fast version spread quickly and was pushed in all of our faces by social media. The slow version often requires you to pay close attention to high-quality sources, or go looking for information with deliberate care.
Of course, legacy media didn’t perform much better than social media, so we can’t set the blame entirely on the marketplace of ideas here. However, as I hinted at earlier, a lot of legacy media these days is just regurgitated twitter gossip, partially because investigative journalism is underfunded, slow, and rare, which in turn is because of the economic market forces around media. So in a sense this is still the internet’s fault even though it’s also journalists’ fault.
In addition, the internet was instrumental in finding good information, for those of us who chose to seek it out - allowing us to view and share medical journals and statements from experts. I’m not saying the internet is all bad. Just as globalization allows me to access high-quality health food from around the world if I choose to seek it out, even as it tempts me with Chicken McNuggets, the internet allows me to get high-quality health information from around the world, even as it tempts us all with answers that are fast, easy, and false.
Goodhart’s Law, Engagement, and Truth
Goodhart’s law tells us that when a measure becomes a target, it is no longer a good measure.
Markets need measures to work. For example, we need to be able to measure the value of an idea in order to decide whether to engage with it. We need to be able to measure the truth of an idea in order to decide whether to adopt it.
Consider the first concern: we tend to judge the value of an idea to decide whether to engage with it. For example, if you tell me that Shelby’s Tacos is the best taco place in Springfield, that might be true or false - but before I start to care about whether it’s true or false I decide whether or not to engage with it. I don’t live in Springfield, so to me the idea has essentially no value - it doesn’t improve my life to know where to get good tacos in a city I don’t live in - so I don’t engage.
A social media company might look at that interaction - I saw “Shelby’s tacos is the best taco place in Springfield” but scrolled right past without engaging - and decide to use that statistic as part of a measurement of post quality. It creates an engagement score for each post, and promotes the “top” posts so that more people see them. At first this is great: I see more posts that are relevant to me, I engage more, the social media company is happy, I’m happy, etc. But then someone figures out that the platform is measuring quality by measuring engagement. They figure out how to mass-produce content which is low-quality but highly engaging: for example, clickbait, or its bastard child, ragebait. Now engagement is off the charts but we are no longer seeing high-quality content. Now we’re all bored and miserable all the time, but for some reason, addicted nonetheless.
The same problem occurs when we try to measure truth. We don’t have an oracle, so we can’t directly measure truth. Instead, we measure truth using the assumption that the ideas which many people come to agree on after an open and rigorous debate are the true ideas. They’ve survived challenges and been judged true. This assumption is one of the foundations of the liberal argument in favor of free speech and open discourse - articulated famously by Mill in “On Liberty”. We “know” what’s scientifically “true” by looking at the consensus of scientists after the scientific method has been applied and then checked by other scientists. So you could call our measurement of truth something like “consensus of experts after debate”.
Enter Goodhart’s law. Now all we need to do to make ideas “true” is generate “experts”, have them “debate”, and then measure the “consensus” among them. If you look at the media ecosystem on the internet right now, you can see that this model actually explains a lot of otherwise strange tendencies of the internet.
It explains the impetus for everyone to suddenly gain expertise in whatever new issue crops up at any given time, and also why expertise itself becomes a subject of debate, e.g. recently when Tucker Carlson challenged Ted Cruz’s expertise on Iran, or Douglass Murray challenged Dave Smith’s expertise on Israel
It explains “debate bro” behavior - the desire to use debate as a tool to create legitimacy for fringe ideas, and subsequently the converse urge to avoid debate because it platforms and legitimizes such ideas
It explains why people who want to manufacture their own truth use influencers and bots to create the image of consensus
Expertise, debating, and consensus-forming have gone from an indication of a valid knowledge production process to a set of techniques used in the manufacture of truth - or, the appearance of truth, for, as Goodhart’s Law reminds us, these things are no longer a good measure. You can no longer judge an idea to be true just because it has been “debated” by “experts” who “agree” on its merits. Those concepts have been so thoroughly co-opted that they are now useless. We’ll have to measure our truth some other way.
It gets even more bleak when you realize that LLMs are essentially giant consensus-measuring machines - when they munch all of the words that have ever been written, they’re essentially taking the most accurate measurement of consensus available to humans right now - and that people are currently “fact checking” by ‘@’ing Grok on twitter, and that every time Elon Musk notices that Grok has discovered a truth he finds inconvenient or objectionable, he reprograms Grok to lie for him. We are watching the real-time corruption of the most accurate truth-measuring machine humanity has ever built. Let that sink in.
So, the marketplace of ideas can no longer tell us which ideas are “best”, for any reasonable definition of “best”. It can no longer tell us which ideas are “true,” for any reasonable definition of “true”. That seems bad.
Of course, we still have individual judgment, although even that is being eroded by the rapid increase in people who simply ask AIs to do the thinking for them. Also, judgment has to be based on something, and with the increasing difficulty of knowing what is true, we have an increasing problem of knowing what to base our judgments on. We’re not yet at an impossible situation - I do still think that knowledge is possible - but we can’t ignore that our information environment is getting steadily worse, as is our ability to cope with that problem, and there’s no indication of an endpoint to these phenomena.
Even censorship won’t really work, because the measure will just become “information that can get past the censors” and so even if you somehow managed to find or create competent, benign censors, the new measure would be “content that can get past the censors”, and soon enough that wouldn’t measure anything in particular. We can already see this going on with platforms that monitor content that advocates harm. These platforms look for certain words or phrases associated with violent or aggressive speech, and content creators therefore just avoid these words. Now we have a set of neologisms like “unalive” instead of kill, “grape” instead of “rape”, “seggs” instead of “sex”, etc. which are designed to get around censors. There’s even a particularly cruel innovation where someone will post a still image from True Detective. What does this mean? Turns out it’s a screenshot of a scene where one character tells another that they should commit suicide.
The System is Down
We know that in some circumstances - such as the context of scientific development over time - the free exchange of ideas can produce true and valuable knowledge which can improve human lives.
We also know that in the current global marketplace of ideas, it isn’t doing that with any degree of reliability. As it stands, free speech is broken. The internet broke it. Social media broke it. We have an invisible hand, but the invisible person it’s attached to is a deranged lunatic who eats AI slop and poops ad revenue.
I don’t really have any large structural ideas on how to address this problem. I’m not sure it can be addressed in a top-down manner even in principle. I think the best we can do is cultivate norms, practices, and institutions to serve as producers and caretakers of good, important ideas, and of people who are capable of recognizing and using them. We might not be able to ban Coca-Cola - and maybe we even shouldn’t - but we can make healthy choices and teach others how to do the same and work towards making sure that healthy alternatives are as widely available as possible.
I think we need to learn to recognize the characteristics of a truth-producing institution and the characteristics of a slop-producing institution, and spread awareness of the distinction and how to spot the difference. This is a prime time to do this task, while the distinction is still fairly obvious to many of us. Later it will be harder.
We need to get better at recognizing quacks and promoting norms that sideline these people. Whatever cult mindset that got us RFK Jr. in a cabinet position should be rooted out and eradicated. I’m not an expert on cult deprogramming and I don’t really know if this is even possible, but if it is, we should do it.
We need to make the public understand that fast ideas are junk food - that most things worth knowing require careful study and deep understanding, and that sound bites and takeaways are likely to mislead you if you aren’t careful. We need people to look at the process of knowledge, rather than just the product of it.
We need to promote a distinction between critical thinking and just asking questions. It’s always possible to formulate a question that promotes junk ideas or displaces nutritious ideas: a prominent example in the current discourse is “do you support Israel’s right to exist?”, which is a question the sole purpose of which is to stop you from thinking about what Israel is currently doing with its existence. Critical thinking occurs when we think about an issue in light of its context, considering the merits of different views on the issue, and thereby gain a deeper understanding of the issue and the views about it. Critical thinking does not occur when we think about an issue in isolation, consider only one side of the issue, or ask biased, leading questions designed specifically to guide the listener towards a specific, predetermined conclusion.
We need to stop debating bad faith actors, and get better at recognizing them sooner so we can tell them where to shove it when they try to lure us into a pointless argument. Debate is at least as likely to legitimize a bad idea as it is to promote a good idea, and people who are debating in a public forum are just as likely to be playing to an audience as they are to be genuinely trying to arrive at the truth. Just call these people out, right away. Tell them they’re full of it and not worth your time. They will throw a tantrum. Laugh at them and walk away.
We need to be able to call out bad and dangerous ideas as bad and dangerous. Anti-vax propaganda? Bad and dangerous. Pro-genocide rhetoric? Bad and dangerous. In the before times, anyone promoting these ideas would be socially isolated until they cut it out. Of course, that system had its flaws because some people thought that belief in a round earth was bad and dangerous, and many good ideas were unjustly punished. We should strive for accuracy, be restrained in our punishments, and refrain from trying to “cancel” someone by destroying their career or livelihood over one bad idea. But we should still have social censure, at minimum, against bad and dangerous ideas, and I would say that the radical libertarian US is not the ideal balance of freedom of speech vs. censorship. I’d be happier to live in a place where, for example, Nazi symbolism and gestures are legally punishable, and where platforms which publish “alternative” medical advice can be held liable if someone dies due to following that advice. But I don’t have the power to change US law. I do have the power to call Nazis, anti-vaxxers, and genocidal maniacs all sorts of names, and I make no apologies for doing so. I do have the power to promote platforms with stricter terms of use, and to register my protest when platforms change their terms of use to allow hate speech or medical misinformation, and I make no apologies for that either.
And finally, I think we need to step out of the marketplace once in a while and go and visit that second-hand shop of ideas, or the local library of ideas. We need to look at the classics - the ones that managed to get published back when there were still gatekeepers who could measure truth and quality, before the internet Singularity broke our collective ability to do that via Goodhart’s law. We need to talk to each other face to face, outside of the view of the consent-manufacturing, truth-pantomiming internet panopticon - we need to talk in the real world, where our main incentives are to connect to each other and forge stronger bonds of collective understanding. Sometimes I feel like I get more truth in a taxicab than I do by reading the front page of the newspaper, because the taxi driver is being his authentic self, and the newspaper is being whatever it thinks it needs to be to compete with TikTok. I guess what I’m saying is we need to touch grass.
“You can’t believe everything you read” has never been more true, and we need to cultivate critical thinking, moral reasoning, sound judgment, and social norms which promote the creation and spread of true, valuable knowledge. In the marketplace of ideas, we need to be the people who buy broccoli instead of burgers.