Free Press vs Free Expression
Plus! Apocalypses Averted, Bureaucracy APIs, Seeing Like a State, Prediction Markets, Carta, more...
I decided to make this edition free, because it’s a on a timely topic and because it’s important to me personally. Forwarding is encouraged.
In this issue:
The lead: Free Press vs Free Expression
The Financial Apocalypses That Weren’t
Big Tech Sees Like a State
Facebook’s Prediction Markets
Japan and Corporate Espionage
Free Press vs Free Expression
My favorite kind of writing is the sort that’s remarkable now for being unremarkable when it was written. In 1983, The New Yorker wrote a profile of Ponce Cruse Evans, better known by her pen name, Heloise. The article describes a kind of blandly debaucherous night interviewing Heloise and her husband: they start at a restaurant during happy hour (“you got two drinks for every one you ordered, so the table was quickly covered with margarita glasses”), then drive to another bar (“where it was also happy hour”), then they drive to Fuddruckers, then three more bars, and then the author gets lost and has some trouble with the law (“He told me to give him my license and sit down and shut up or he’d throw my ass in jail for public intoxication. I told him I hadn’t seen much else but public intoxication in San Antonio that night, and his handcuffs made a cricketlike sound as he took them off his belt.”)
All in good fun at the time, but now we’d view this behavior as unconscionably irresponsible. I’d expect even Vice to fire someone for this.
At the same time, it’s hard to condemn him so much: he put lives at risk for fun and didn’t care about it, but it took a multi-decade PR campaign to establish the norm that drunk driving was wrong. Since MADD was founded in 1980, traffic fatalities per mile driven have dropped by about two thirds. (Some of this was doubtless due to better safety, but tighter drunk-driving regulations, and stronger social norms, certainly had an impact.)
When technology changes, social norms need to change. Technology lets us do more with less, but “more” is a measure of magnitude, not direction.
So: should the New York Times publish the real name of “Scott Alexander,” the pseudonymous blogger who wrote Slate Star Codex? (Note: the blog has been deleted, except for a post begging the Times to reconsider, but archives are floating around.) Scott says he already gets death threats because of his blog, that irate readers have called his workplace trying to get him fired, that it’s a standard practice for psychiatrists like him not to let patients know too much about their lives, etc.
Here are some good reasons not to care at all:
This is a niche story, and not really newsworthy. “Area man used pseudonym” is not a big deal.
The article is allegedly a positive one, about how the online rationalist community Scott’s at the center of was worried about Covid-19 before the rest of the country.
The Times has a policy of publishing the real names of sources, which they only violate for edgy socialist podcast hosts, Banksy, members of ISIS, professional gamblers, and… a therapist who writes under a pseudonym because he’s worried his writing could interfere with his client relationships.
Scott did not try especially hard to preserve his anonymity. (For example: I’ve never met him, but I found his real name by accident by browsing on a site he’d linked to.)
There are many obvious reasons not to out him, but the most important is the least obvious: the Internet has changed how easy it is to communicate, and what messages get amplified the fastest. And the result is that legal protections for the media are in tension with the principle of free speech.
In an ideal world, laws start from moral principles (like “don’t hurt people” or “don’t restrict freedom without some offsetting benefit”), but what they actually cover are specific behaviors. “Don’t behave irresponsibly in such a way that you put other people at risk” is a vague precept that’s open to interpretation, while speed limits or laws limiting the mercury content of fish are pretty straightforward.
When producing and distributing information has a high fixed cost, the chokepoint for freedom of expression is the physical production of books and newspapers, or of radio and TV broadcasts. But online, almost anyone who can consume content can produce it; halting production is a losing game. In 2007, the MPAA tried to halt the distribution of a decryption key for DVDs and Blu-rays. They sued a few sites for publishing it, but Digg users revolted and repeatedly circumvented takedowns to post the key. The users won; the Wikipedia article about the controversy includes the full key.
Laws designed to protect free speech when distribution is the bottleneck will tend to have a strong bias towards letting publishers choose what to publish. Strict libel laws, for example, make it hard to call corrupt politicians or unethical business leaders to account. So countries that care a lot about speech will tolerate a fair amount of libel. But when the cost of publication drops, the cost/benefit analysis shifts: with a small number of media outlets, all of which have limited distribution, the odds of one of them writing something damagingly false, or misleading, are low. When everyone can publish whatever they want, someone will publish something terrible.
Pro-publisher norms interact with mildly-autistic writers in a very dangerous way. The writers are productive, and have niche interests, so they write a lot without getting much traction. Over time, if they’re interesting, they develop a following. Eventually, they have haters. And then two scale factors kick in:
If you’ve written a million words, the most damning 0.005% of your oeuvre fits in a tweet.
If 99% of people are basically decent, and there are 4bn Internet users, that means there are 40m truly terrible people, and only one of them needs to quote something out of context to kick off the outrage cascade.
Slate Star Codex has a lot of verbiage. It may not be upt to the million-word count just yet, but I wouldn’t be surprised if it had hit 500k or so. A pseudonymous blog is a way for a writer to interact with ideas, and explore thoughts that might be weird, uncomfortable, or dangerous; pseudonymous speech is easy to un-say. Whereas writing under one’s own name is subject to a veto from anyone in your life who has something to lose by associating with you: employers, friends, landlords, even family.
Fame has always had upsides and downsides, but now the downsides scale faster unless you intend to spend all your time being famous. If you want to get interviews, get introductions, give speeches, endorse products, or work as an actor, having those opportunities but dealing with death threats and stalkers might be a fair trade. But if you mostly write about books and weird theories, it’s terrible. So a world in which anyone can make you famous is a dangerous one: at best, it’s a world where you have to switch jobs; at worst, it’s a world where you’re unemployable.
Connecting a pseudonym to a real-world identity in a verifiable way unleashes potential energy proportionate to how interesting the pseudonymous work was. A little bit of that energy is beneficial—Joe Klein probably got more TV appearances after he got outed for writing Primary Colors), and J. K. Rowling’s pseudonymous book went from 4,709th to 1st on Amazon’s bestseller list when she was revealed as the author.
But mostly, people who become instantly accidentally famous for something aren’t happy about it, and rightly so. Some of them become famous for something they did that was legitimately bad, but the Internet can’t handle proportionate response: either nobody cares, or everybody does. It’s as if English had a term for “Yes, please,” but the only equivalent to “No, thank you” was punching someone in the face. It’s just a matter of virality: promoting news about a minor slight won’t get much traction, but if a negative story starts to grow, it keeps on growing—and a tweet that prompts 10% of its readers to retweet it might prompt 0.1% of them to send a death threat, try to get the subject fired, or worse.
Social media doesn’t have an especially strong chilling effect on people who misbehave in the real world, because they’re usually not wondering if their actions are being streamed on Facebook live. But it does have a chilling effect on people who write online. They face “grep risk”: the risk that someone will take a dislike to them, and search through everything they’ve ever said for the most injudicious bits, or even the bits that sound worst out of context.
(There are at least three prominent cases I know of where someone got fired because they parodied people they didn’t like in a way that could be misread as endorsing the beliefs they were mocking. This has been happening for a long time: Darwin gets misquoted all the time because his writing is full of hedges and caveats that, out of context, sound like concessions that evolution doesn’t make sense; “We are all Keynesians now” gets quoted a lot more often than the full quote, which was “In one sense, we are all Keynesians now; in another, nobody is any longer a Keynesian”; “Information wants to be free” was one sentence in a paragraph that started with “Information wants to be expensive,” and so on. When you see a surprising pull-quote, prepare for a boring transcript.)
In fact, one reason for this is that the ideas that get amplified the fastest are not the ones that obviously benefit one side, but the ones that both sides can say confirm their priors. There’s a great blog post on this called “The Toxoplasma of Rage,” which I’d link to, but it was on Slate Star Codex.
It’s very good for the world if people who are interested in ideas rather than people can write about ideas without facing personal consequences. In the space of possible ideas, the vast majority are nonsense, most of the rest are wrong, but every future discovery is somewhere in the space of possible ideas, so tolerating the wrong or crazy is basically the only way to accelerate the discovery of the good and novel.
Scott will probably be approximately fine, whether or not the Times links his pseudonym to his real-world identity. Perhaps he’ll be a full-time writer instead of a psychiatrist with a fun hobby. That would be nice for me, and for his other fans. His life will be worse, since he’s lost an important freedom, but it won’t be over—he’ll get more death threats than usual, but online death threats very rarely get carried out.
It’s bad to live in such a binary world, where Scott’s future is up to a journalist on a deadline rather than the person who lives with the consequences. Pseudonymity is important because it lets someone explore a body of intellectual work long enough to build up to a useful conclusion, without the risk that their false starts, dead ends, and mistakes will harm their real-world ability to earn a living. Anonymity is more complex; it has its uses, sure, but it’s a call option on recreational harm. And nobody is more anonymous than one more participant in an angry mob.
So I hope the prospective outing of Scott Alexander is a Schelling Point for reconsidering social norms and legal rules around media and privacy. The media can make someone famous, whether they like it or not; social media can make them infamous in a matter of hours. Most people don’t play the lottery, because they recognize that the odds are against them, but viral media enroll everyone in a downside lottery. Your ticket probably expires worthless, but every day there’s a new big loser.
In the interest of full disclosure: I don’t have any direct relationship with Scott, although we have a few dozen acquaintances in common. I’ve linked to SSC in this newsletter a few times. Scott has also linked to something I wrote once—under a pseudonym.
The Apocalypses That Weren’t
Two interesting bear theses a month or two ago were 1) that online advertising would be hit harder in 2020 than it was in 2008, and ad-driven tech companies would see lower profits and multiple compression, and 2) that renters wouldn’t pay rent, leading to a spiral of defaults. Neither one played out, but for every different reason. In digital advertising, the Covid crisis dragged the future forward, making more retail spending either online-first or online-mediated. And so far, PPP and unemployment insurance have kept renters current. But these are very different outcomes: advertising was a step-function change that’s likely to be somewhat sticky; shoppers accustomed to curbside pickup will be more likely to browse inventories virtually even if they pick up goods in person. But the rent situation appears to be a function of fiscal policies that have a definite expiration date.
DoNotPay has raised $12m for its “robot lawyer” product, which helps customers contest parking fines, file for unemployment, cancel gym memberships, and otherwise replace paper chases and hold music with a simple interface. Some of the systems DoNotPay assists with are designed to be abusive; gyms and the New York Times clearly make cancellation hard to reduce their churn rate. But in other cases, DoNotPay is helping institutions skip a generation: unemployment claims were automated decades ago, so they run on un-maintainable legacy codebases. Building an API that gracefully handles the system’s quirks may be easier than replacing the system itself.
(As a side note, DoNotPay’s founder, Josh Browder, is the great-grandson of the former general secretary of the American Communist Party. I’m glad the family has continued the tradition of fighting for the little guy, albeit in a more effective way.)
Big Tech Sees Like a State
A long-term Diff theme is that big tech has similar incentives to big government: make reality richer (so there’s more of it to tax) and legible (so taxation is easier). Two recent headlines in this vein:
Fandango, like Yelp, makes it easier to search for locations that comply with distancing measures. Fandango’s search defaults don’t have legal force, but their effect is equivalent: whatever they default to becomes the de facto standard.
Google wants to eliminate the password. This is clearly useful for them: passwords were an okay idea when most people only had one account, but with multiple accounts it’s hard to maintain good password hygiene, and password managers have low (but growing!) adoption. Anything that makes the Internet more secure increases online spending, and AdWords captures a chunk of that.
Facebook’s Prediction Markets
Capital markets are the best prediction market in the world, because they combine three important questions: are you right, are you right about how wrong everyone else is, and are you right about something that matters? But prices project a high-dimensional reality onto a number line (or, if you want to get fancy using options, the two-dimensional space of price and timing). So markets are not a great place to find specific predictions about exactly what will happen. Facebook is trying to conquer this particular curse of dimensionality with a reputation-based prediction market. Google has tried this internally in the past, though it doesn’t appear to be active today. Prediction markets are a challenge in part for the dimensionality reason above: if you don’t control for how important and contrary a bet was, you select for traders who have an edge because they’re either good at estimating the consensus or good at predicting events that ultimately don’t matter.
The most impressive thing about Carta is that to compete with them today, you have to somehow outrun a network effect they’ve been building for years. (See my writeup from May for more.) AngelList may be able to do just that: they’ve launched a cap table management product, targeting the Indian market. It’s free, it gives AngelList information they can use for their core product, and it means that in at least some geographies, they’ll be ahead of Carta if they try to make a market in private company stock.
Japan and Corporate Espionage
Japan is requiring universities to disclose foreign funding before they get research subsidies from the government. The US has had a few cases of professors who received undisclosed funds from the Chinese government. (Everyone has a weakness. For some of us, it’s competitively growing enormous pumpkins.) China’s increasingly autarkic technology policies make sense in light of deteriorating relationships with the US and (to a much lesser extent) its allies, but they give the CCP a very strong incentive to keep up-to-date on other countries' research.