Many people wrote to me in response to my last column, “The Slippery Slope to Censored Speech,” disagreeing with my contention that people have a right to have their speech distributed and magnified through social platforms and personally ranked feeds.
These people assert that when social platforms suggest specific voices to follow or content to show a user in a feed, the platform becomes a party to—and should be responsible for policing—what is highlighted.
This line of argument leads to the belief that you can separate freedom of speech from freedom of distribution of speech. That implies the debate we face today isn’t so much about free speech, which most people claim to support, as it is about defining enforceable standards for selective highlighting and promotion of specific voices and content.
I understand how people come to this perspective. In fact, I think social platforms can primarily blame themselves for this narrative, having aggressively—and incorrectly—branded technology to deliver personally ranked feeds of social content as a form of news (as in the Facebook branding of its News Feed as a personal newspaper).
That said, conflating a ranked feed generated for a user by a social platform with publishing a newspaper is an inaccurate and fundamentally dangerous analogy. So, too, is the other often-referenced analogy of a ranked feed as a “space” like a bar or coffee shop, where the proprietor has some latitude over the rules and who may enter.
Instead, ranked feeds must be understood and treated as an important part of modern communication infrastructure. Social networks must stay out of the business of selectively moderating what flows through them.
If anything, we as a society should ban social platforms from using the substance of content as an input for selecting what speech to promote or suppress. We should guarantee end users more transparency and control over what they choose to see in their personal feeds.
What we should not do is encourage (or even allow) social products with ranked feeds to play an editorial role in selecting what users see or how they see it.
The right to speak freely is not the right to simply shout into the void what you will. The right to speak freely encompasses the right to say what you want, to hear from whom you want, and to be served without discrimination by technology that sits in the middle.
A Ranked Feed Is Not a Newspaper
Let me explain why standards applied to newspapers are not appropriate for personal ranked feeds.
A newspaper speaks as a single voice. A single organization controls an end-to-end content experience, which it delivers to all subscribers in the same way. That single branded voice carries the reputational and financial risk and reward of its work. The work starts with selecting paid journalists and assigning them areas to cover, includes setting editorial processes and standards for facts, and ends with selecting what gets distributed and managing physical delivery.
Contrast this with a ranked feed sitting on top of a social network.
In a social product, each voice is unique. The social platform provides only the distribution network to match content posted by people with those who want to hear from them. Social networks do not have a voice. The content comes from individuals who are unpaid and undirected. The network does not provide any editing or standards, and the distribution is personal and automatic.
This description may sound abstract, so think of what it means in practical terms.
The Economist doesn’t have bylines. If you read it, you are trusting the organization as a whole to hire great people, do great reporting, fact-check, edit well and select the right stories to feature as a single package. Most other publications don’t go to the Economist’s extreme of no bylines (largely because journalists want to build their own reputations). But even in other newspapers, like The New York Times, bylines are not on the homepage. The point is that readers trust the brand—not a specific reporter.
In the case of products like Facebook and Twitter, the name and photo of a story’s author is the first and biggest element on the page. In a social media context, the identity of the speaker is the most important part of any post. No one does (or should) trust anything posted on social services just because the technology platforms delivered the information to you as requested by the poster. The trust (or distrust) is embedded in the speaker’s identity alone.
Of course, far more differentiates ranked social feeds and newspapers. But the fundamental reality is that—if one can make any comparison—in the context of a social network, a newspaper is far more akin to a single trusted friend than it is to the technical infrastructure of a social feed delivering a posted message from one person to someone who asked to hear from them.
The upshot: Don’t argue that ranked social feeds are responsible for content because they are “like newspapers.” They are not. They are more like the postal service or a trucking network delivering what they are told to. And they are open to all.
A Ranked Feed Is Also Not a ‘Bar’
Sometimes, in more-sophisticated conversations, people abandon the obviously incorrect newspaper comparison, and suggest instead that socially ranked feeds are like a physical bar or coffee shop, where the owner doesn’t control the content but has a responsibility to curate the space and can choose whom to silence or remove.
These people argue that social products should be policing their content in order to create a warm and inviting space for their patrons. They contend that things like fake news and extreme speech transgress on this space, and give the social products the right (and the business interest, of course) to clamp down on such acts.
I think that the physical space analogy is a good one for things like administrated groups that you can choose to join within social products. In that case the administrator, not the platform, can set the rules and manage the community as they see fit.
But this is a very bad analogy for what happens in ranked social feeds, for two reasons.
First, we let bar and coffee shop owners kick out underdressed patrons or people who are being too loud because those patrons are hurting the experience of the people around them. In a social feed context, if I am getting content delivered to me that you don’t like but also don’t ever see, that has no negative impact on your experience of the product. My enjoyment or use of the infrastructure has no impact on how you use it, and vice versa—so there is no justification for any limitations.
Second, the idea of letting the owner of a bar or coffee shop set rules for their “space” is predicated on the idea that patrons have many different options to choose from that serve the same drinks. Giving owners of social media platforms broad latitude to include or exclude people doesn’t make much sense, though, when they are the only game in town.
Some argue that there are enough social platforms out there that they should be able to set rules as they choose. You can always just go use another speech platform if you like.
This, however, doesn’t really reflect the reality of network effects. There are a handful of key social platforms, and what they offer isn’t commodity drinks that you can buy anywhere—it is entrenched access to communicate with the user base that isn’t going anywhere. Every once in a blue moon a new “bar” will crop up. But when the vast majority of people go to just a tiny number of venues, the validity of the idea that those venues can choose to refuse service decreases dramatically.
Ranked Feeds Are Fundamental Modern Social Infrastructure
So, if the oft-cited newspaper analogy is dead wrong, and the bar or venue analogy is also incorrect, that begs the question—what is a ranked social feed and how should we think about setting rules to govern it?
The answer is that a ranked feed is a system that works on behalf of a user to surface the content they are most interested in from the global set they have access to. The algorithms work on behalf of the end consumer, reflecting their interests alone.
Access to freely use this type of technology, in my mind, is a fundamental right.
We often think about free speech in the very narrow context of my right to say something. But as technology gets more powerful, it becomes critical to articulate that free speech must also include the freedom to hear what I want to hear. The freedom to hear, in a world of infinite scale and a cacophony of voices, requires that I retain my right as an end user to leverage technology that highlights for me what I want to see.
If I don’t have the freedom to select the voices I want to listen to, and then prioritize what I want to hear from those voices, that infringes upon my right to hear.
The retort people often like to insert here is that “no one is saying you can’t have free and unlimited private messaging—we are just talking about ranked feeds of posts.” But this misses the point. Chronological messaging only scales up so far. After a few hundred voices or conversations, the threads become impossible to follow.
The innovation of a ranked feed is that it is infrastructure that helps you as a person hear from more sources and prioritize your focus and time better than you can without those tools.
So the question is, does the state or society (or a private company, of its own volition) intend to limit my ability as an individual to use technology to find the voices I am most interested in and the most interesting content from those voices? I think to assert rules over this ability would be a major violation of rights.
In the end, the thing to take away is this—ranked feeds and suggestions are a fundamental social infrastructure and need to be treated as such. You can’t claim to believe in freedom of speech if you don’t also believe in freedom of listening, and in allowing people to use technology to prioritize what they choose to see.
Fundamental Medium for Defending Speech Rights
Sometimes technologists, in particular, argue that the problem isn’t the idea of ranking systems and feeds; it is the challenge of closed ranking systems on top of walled-garden social networks.
They say that people should, of course, be free to use whatever technology they want to scan the whole internet and choose what to read and where to interact. But they also say that the end-to-end integration of identity (profiles) plus speech (content) plus distribution (privacy and reach) plus socially ranked feeds, all in single, vertically integrated products, is the problem.
Abstractly, I think it would be great if the open internet could be the solution. Unfortunately, several early design decisions about how to implement the web left it hobbled as a native speech platform (though these same choices, it can be argued, were critical to its growth).
Instead, it took the advent of social networks to layer the necessary identity and privacy functionality on top of the “base” open internet to make it a broadly usable and useful medium for speech at scale.
Speech is the combination of identity, audience (privacy and distribution), and content. Social networks, not the base internet, are the platforms that provide this mix, and sadly it is very hard to decouple these functions because of how they reinforce each other.
This means that rather than having rules of free speech embedded in the fundamental guts of the internet’s design, we need to navigate the dangerous reality of having human speech consolidated on a handful of platforms. It means having identity, content, distribution and prioritization all mixed together (as they are in the real world).
Many might want these rules fundamentally encoded and the keys thrown away, but we are going to have to wrestle with these rights on a human level, in negotiation with private organizations—there is no technical solution.
Profit and the Objective Function of Ranking Problem
Some people believe it’s a good idea to have social platforms providing ranked feeds and think the platforms need to be open and free, but balk at what they see as the natural motivation of private companies providing these services.
Specifically, people argue that the problem is that ad-supported social networks are not reflecting what users want to see in their ranking, but instead are giving a higher ranking to bad voices that maximize engagement and profit.
This is just wrong.
In most scenarios, it is precisely the people most concerned with maximizing profit who are most willing to do more content moderation. Sure, content moderation is expensive, but advertisers don’t want to be associated with extreme speech. And even if they were, the regulatory risk of supporting freedom of speech is too high. People don’t support free speech for profit—they support it despite profit. People don’t understand that advertising models tend to naturally dive towards more centerism and moderation, not extremism.
Consider the case of Tumblr. Tumblr is in many ways one of the most open speech platforms on the Internet, with nearly zero enforced rules. Because of this openness, it has nearly an impossible time monetizing its content. Contrast that with a service like TikTok, which is known for heavily curating what is shown and aggressively taking down content that violates stringent content guidelines. TikTok is likely to rapidly become a financial juggernaut.
Looking beyond the narrow profit issue, there is a legitimate question of how, as a service, you reflect what individual users want to see. Beyond that, there is a separate question of how, even if you are showing what they want, you prove to a user that your selections are correct (this is a classic ranking problem).
These unsolved problems are still being worked on. When people see things they don’t like highlighted specifically for them, or feel bad once they have looked through what the feed has suggested to them, everyone loses. So the question is how to get better at showing people the things they really want.
Personally, my answer is always that if you can’t figure out the provable right answer, let people decide for themselves. Since social products can’t possibly know exactly what users want, nor can they prove to those users that what they are showing is good, the answer is to put control as much as possible in the hands of the people and let them make decisions for themselves. The platforms are generally founded on this principle, where users select the voices they want to hear from and tune the algorithms based on what they choose to interact with or remove. But I think there is always more to do to give transparent control to users.
Rules for Ranked Feeds
As I wrote in my last column, I worry that we have already gone way too far down the path of content moderation and are on a slippery slope to censorship.
Many social products have already caved in to those who want to limit freedom of expression or see a specific viewpoint highlighted above all else.
Too many teams are moderating illegal speech with tools that too easily adapt to censor more and more types of content. This has already sufficiently broken the social norms around freedom of digital speech so as to normalize limiting speech—making it easy for those limitations to balloon dangerously.
That said, we should consider encoding as law an alternative package based on fundamental rights for digital speech when it comes to ranked feeds. Here are a few ideas.
First, we could discuss whether users should have more direct access to view and modify directly the ranking algorithms social products build for them. If I want to see more or less of something—or specific viewpoints—I should be able to express those desires, and the system should have to respect them. This algorithm openness debate is an old idea, but it needs to come back for serious consideration. It involves practical challenges, but intellectually and emotionally it is a standard I believe in strongly, since these systems work on my behalf.
Second, while this is more challenging to encode into a set of rules, one can make a strong argument that social platforms should be banned from evaluating the content of a post when choosing whether to feature it. To rank a post using heuristics like the source of the post, or metadata such as who has commented on it, makes sense and has natural-world equivalents. You could, however, tell services that they cannot look at the content of the speech itself as a signal for how to rank a post. As services move toward true end-to-end encryption, this is the natural direction anyway. In the future, systems won’t actually know the substance of the content they are delivering on behalf of users. But ahead of that reality, it wouldn’t be a terrible idea to set regulations that ban using content as an input for ranking. This could tie the hands of social products in a healthy way and prevent them from blocking certain words or concepts from flowing through their systems even if many people want to see that content.
Other rules may make sense to add. But the spirit of any such new rules has to be to extend and deepen free speech in the digital age. That means protecting my ability to have an identity and to speak what I want to the audience of my choice, allowing people to freely hear from those they want to hear from, and enabling people to use technology freely to sort out what to listen and read from the corpus of things they find interesting.
How to ‘Fix’ Speech With More Speech
Ultimately we have to acknowledge three things that are in tension.
First, the Narcissus problem is real and a threat to society. There is no doubt that the blurring of fantasy and reality is a danger because most people, if given the choice, will choose to read and watch an idealized and entertaining reflection of themselves rather than reality. Today human characters and brands produce that reflection, but soon enough it will be pure artificial intelligence. As I have written over the years, this is an existential threat, as technology allows us to more deeply personalize our information and entertainment spaces and disconnect from reality. Personalized feeds are in many ways the leading visible edge of this challenge.
Second, in facing that threat, the seemingly easy solution of manipulating the technology platforms that provide the rails for speech is a very dangerous tactic. We need to leave the fundamental rails of communication free, open and trusted. Otherwise we risk serious long-term manipulation and control. Even short of that disaster, we risk devaluing what we see and hear from others as a form of manipulated or limited Kabuki theater, destroying our trust in each other and open debate. Many people try to argue both sides, asserting that they believe in free speech but think ranked feeds must be editorialized and moderated. This is not a split you can make, however. Leaving the rails of communication free and open means leaving ranked feeds unencumbered to deliver to end users what they ask for.
Finally, it is very hard to defend rights and systems for protecting the long-term vibrancy of society in the face of immediate pain and challenges. That said, our ability as a nation and society to hold ideas and rights above all else is a big part of what makes us strong. As we face real suffering today, it is easy to fall into the trap of making small sacrifices in terms of our rights or the correct framework for speech. Doing this might feel good in the moment, but it is exactly these types of trade-offs—made in small ways, over and over—that erode a society.
In the end, if you believe in free speech, you have to also believe in freedom to select the voices you want to hear and what you hear from them. If you believe in that, then you should strongly agitate for stronger rights that protect your use of technology to highlight the voices you choose to listen to freely and that prevent manipulation and moderation.
Where this leaves you is that the solution to speech you don’t like has to be more speech. We should welcome having all ideas and viewpoints out in the open and for discussion. We should deeply fear a world where common spaces become hostile, pushing voices underground or into the fringe, where ideas can fester. This has always been true, but it is more true today than it ever has been in the past.