Three Crises

On February 13, 2023, the CDC released a shocking report about teenage girls.  Nearly 3 in 5 had experienced persistent sadness or hopelessness in 2021.  30% had seriously contemplated suicide.

On March 6, 2023, the United States hit yet another awful milestone in its losing battle with gun violence.  By the 64^th day of the year, there had been 100 mass shootings in the USA – a feat that had taken till March 19 in 2021 and March 22 in 2022.

On March 10, 2023, Silicon Valley Bank collapsed after a bank run. It was the second largest bank failure in US history.  It could be a blip, but it could also be the state of a financial crisis.

On the surface, these three crises have little to do with each other.

Indeed, it could be in poor taste to associate them with another.  After all, each would seem to demand a different intervention.  Bankers might need therapy or benefit from gun control, but that’s it for common ground, right?

I disagree. 

These are not isolated crises.  They are crises that are arising at the same time for a reason.

Each of them – from spreading teen depression to financial panic – is a direct consequence of the information ecosystem that we have constructed for ourselves.  But it is not as simple as saying “social media good!” or “social media bad!”  Humans, it turns out, have always been vulnerable to outbreaks of hysteria.

Dancing Ourselves to Death

If you read enough history, you occasionally come across something that really fucks with your head.

“The year was 1374. In dozens of medieval towns scattered along the valley of the River Rhine hundreds of people were seized by an agonizing (sic) compulsion to dance. Scarcely pausing to rest or eat, they danced for hours or even days in succession. They were victims of one of the strangest afflictions in Western history. Within weeks the mania had engulfed large areas of north-eastern France and the Netherlands, and only after several months did the epidemic subside. In the following century there were only a few isolated outbreaks of compulsive dancing. Then it reappeared, explosively, in the city of Strasbourg in 1518. Chronicles indicate that it then consumed about 400 men, women and children, causing dozens of deaths.” (From John Waller’s A Forgotten Plague: Making Sense of Dancing Mania)

Who were these people that saw people dancing who couldn’t stop and then joined in themselves?

It sounds unbelievable.

Until you learn about what happened in January 2021.

Across America, England, Australia and Canada teenagers starting coming into emergency rooms with a similar sit of symptoms.  The kids, seemingly healthy, had all developed tics commonly associated with Tourette’s syndrome.  To the disbelief of their family, they had started uncontrollably swinging their arms and shouting loudly.  

The common thread, researchers realized, was this:  all of them had been watching TikTok influencers with Tourette’s. Their brains had seen the tics and subconsciously starting mimicking it.

Yes, really.

We already have a loose sense that information can act like a virus.  We call videos that spread quickly “viral videos” after all.  We talk about “spreading memes” or people being “vectors for the disease of misinformation.”

But we think of those phrases as metaphors.  Viruses might take over our body for their own ends, but information?  Our minds are our own, surely.

Except those metaphors are closer to the gospel truth than we might like to admit.

In his 1976 book, The Selfish Gene, biologist Richard Dawkins coined the term “meme” as a cultural analog to a “gene.”  A meme, like a gene, was a bit of raw information that contained content – an idea, a belief, a cultural norm, a value judgment – and like genes, these bits of information were subject to natural selection.

Any new idea or observation could be replicated by being spread – much like a gene in reproduction – and just like a gene, it would be tested by the environment. Many would die off, but the “most fit” would survive, replicate, mutate and mix with other memes to create the “fittest” ideas.

In the marketplace of ideas, we choose the ideas.  In the evolutionary ecosystem of the “meme,” it’s not the best ideas that win but those that are best able to demonstrate fitness – the capacity to spread through hosts.

But even if this is true, it’s been true forever.

So why would there suddenly be crises? We have millions of years of evolution-designed hardware to do battle with the fittest ideas.  And we have hundreds of thousands of years of developing social tribes and  governing systems to process information and discard those memes that are not useful.  We were well equipped for the Darwinian wars of yesteryear.

But we are not fighting the last war anymore.

We have removed ourselves from the pond of local social interaction, ideas filtered through local and comfortable norms that our ancestors inhabited, and dove head first into the ocean of a new information ecosystem that operates with a pace and ruthlessness that rivals the prehistoric savannah.

Why Information is like COVID

In the last three years, we have all become relatively comfortable with understanding the spread of viruses.  So let’s apply that model to our information ecosystem.  After all, a person – when confronted with a bit of information (whether virus or meme) can reject the information before it takes root, we can host it but defeat it before spreading it, or become a vector for spreading it ourselves.

Thus, the same models we use for understanding the spread of an infection can be used to model the spread of an idea.

In epidemiology, this model is defined by two variables:  R₀ and R_E. 

R₀ measures the propensity of a virus to spread among a naïve population that has no immunity.  If a virus has an R₀ of 2, then each person can be expected to infect 2 others.

R_E measures the propensity of a virus to spread among a population in real life.  If a virus has an R₀ of 2, but half of the population is immune to it – from prior exposure or vaccination - then its R_E is half of its R₀ and each person can be expected to infect only 1 other.

Information spread can be modeled in the same way.  A particularly “infectious” idea – whether a really funny video, a signal of virtue, or a fear response – should have a very high R₀.  Even more dangerously, repeated exposures to an idea – if not corrected – makes it more likely to stick. It’s as if every time you got COVID, you became more infectious.

Now, we fought COVID the way we fight all viruses – by making it harder to spread.  The thing is – we’ve been doing the exact opposite with information.

In the early days of humanity, ideas spread orally, so you could only learn of them from someone speaking to you directly.  Then we had writing, which allowed a bit bigger of a vector for spreading ideas.  But the internet has lowered the cost of putting content into the world to zero – and it is literally available for anyone on the world to see or consume.  We simply have a far larger amount of memes out there than ever before!

And then we needed a way to sort through all this spreading information to find the ones that we cared about.  So we deputized giant machine learning models to help serve the role that our social tribes once played – filtering through the noise to find our personal version of signal.  And in doing so, we’ve created an infrastructure that maximizes the speed at which information can spread and that can ensure the fittest ideas get maximum exposure. 

Trending algorithms, like those on Twitter, find the memes that are most viral at any given time and funnel more people to them.  It’s like if there were a COVID outbreak in a town so rather than isolating them, we tried to cram as many people into the space as possible. 

Ranking algorithms, like this on TikTok or Instagram, find the memes that are most likely to take root in an individual and deliver them directly to a host. 

This need not bad.  Most of the information we share is good.  We share the things that make us human – our families, our struggles, our hobbies.  We also share causes that can change the world for the better.  Human rights are a pretty good meme and we should spread it.  Democracy is also pretty solid, IMO. 

But evolution – whether biological or memetic – cares little about what is “good” or “true” or “helpful” to its hosts.  It cares about what can survive and spread. It generates strong ideas that can exploit the hosts most vulnerable to its fitness adaptations.

So when it comes to questions of identity or insecurity or social isolation, it finds a willing population of teenage girls to spread among.  When it comes to violent crime, we send news about violence to mentally unstable individuals.  When it comes to financial panic, we send it to Very Online VCs like David Sacks.

And just like a virus – the meme’s success does not depend on the thriving of its hosts.  It is happy to infect, spread and move on.  If the human being that listened is left damaged, well… that’s just nature baby.

It's not an accident that we use the same word “contagion” to describe how suicide begets more suicide in local areas, or how mass shootings spawn imitators or how a bank run at a single bank seemingly spawns similar runs at totally unrelated banks.  Human see, human do.

We are not powerless in the face of nature.  We escaped the War of All Against All once and we can do it again.  Friedrich Hayek famously wrote that capitalism succeeds because it does what no other system can – it enables the effective use of information by all members of society.  That is our challenge once more, to build systems that can turn a waterfall of information into a power source rather than a devastating flood.

Viral Interventions

In the American and Chinese response to COVID, we can glimpse two paths for civilizations to confront threats that spread like viruses and grow exponentially.

China used a top down system as a blunt instrument. They shut down dissent and shut down vectors of transmission. They succeeded in the short term, but with a heavy long term cost. China’s population remained utterly naïve. Because they had not been exposed to the virus, when they finally “let it rip,” it may have killed up to 1.5Million people.  All while threatening their economy and political stability. Authoritarian interventions work in the short term, but fail to build any kind of durable resilience.

The American System, in contrast, is best defined by – well the lack of a system.  America is both small-c and large-C constitutionally incapable of a Chinese response.  We do not abide being ordered.  The results of our non-system were not pretty.  Over 350,000 Americans died in just the first year of the pandemic.  But the longer term results – well – are a bit more balanced.  Americans developed resilience in the face of the pandemic.  Getting COVID in America today is inconvenient, but it is hardly catastrophic for the lion’s share of citizens.  Our population developed interventions – vaccines, new treatments, masks and personal spacing – and life returned to normal.

China’s control works in the short-term, but breeds long-term fragility. America’s is short-term painful, but breeds long-term resilience.  The lesson for our information ecosystem is also clear – we want to minimize short term pain while maximizing the gains of long-term resilience.  And we have tools to do that.

We can start by getting eyes on the potential spread of dangerous memes. Today our best view of the spread of “memes” comes from Twitter’s trending page.  But this is personalized to us, and lacks huge amounts of important context on the spread of a meme – where is it spreading? Who is spreading it? How quickly is it growing?  And just as importantly, is it causing dangerous consequences? 

Viral video of a banker falling on a street – good.  Viral video of a banker falling as they are overrun by fearful depositors – maybe, less good.

If we develop a set of open and transparent tools for viewing the emergence of information, we will be able to understand when interventions are worth deploying.  I don’t pretend to know who should deploy them – the state, the companies or the emergent digital organizations in blockchain-land building a collectively-owned commons.

But whoever gets their hand on the lever should know that they have a few tools in their toolbox: interventions, inoculations, circuit breakers and cooperation.  Not every tool is right for every situation. Many can be abused. But the goal in managing any sufficiently complex system isn’t to find a silver bullet – it’s to have a variety of scalpels handy.

We need to propagate interventions that work to empower and build resilience in populations. Think of these as our “medication” for information viruses.  Today, good interventions are lost in a sea of bad information.  Today social media content on mental health is driven by the “trauma influencers.”  These accounts, some pseudo-therapists, some just nihilist meme-makers, emphasize emotional validation rather than actual existing constructive therapies that, while they share some of the language around trauma and triggering, ultimately move beyond validation to teach coping skills.  Effective therapies exist – I’m a personal fan of ACT, CBT and dynamic modes-- but these interventions are more difficult to put into a viral meme. 

We need to inoculate individuals against harmful memes.  Studies on misinformation show that individuals exposed to a “weaker” version of an idea can build up resistance to that meme. So that when they are exposed to it, they are less likely to adopt and spread it widely.  We should be teaching people about the types of memes that play on their built-in biases to trigger emotional responses that are unhelpful to their personal goals – panic, fear, hate, depression, etc.

We also need to consider when it might be helpful to implement a variety of circuit breakers in our system.  Today, when the stock market falls rapidly, circuit-breakers are triggered that temporarily suspend trading.  Individuals remain in control of their equities, and they can begin trading again after a pause.  But a “circuit breaker” response to overheating memes that temporarily restricts or delays sharing, could prove helpful for breaking panic loops that trigger things like bank runs.  (Though admittedly, this capacity has a great risk of being abused).

Finally, we might look to our ancestors. 

They understood that one of the best systems for confronting dangerous new ecosystems was to band together.  Tribes are an evolutionary adaptation so that humans can cooperate to put a stop to threats.  But our modes of cooperation are often slow and involve too much counterparty risk.  For example, during the bank run, many VCs tweeted about standing by SVB, but without the ability to credibly organize, each had to fear that another VC would defect and they would be left holding the bag.  Smart contracts – with their ability to create self-enforcing mechanisms for cooperation online – should fix this by enabling these banks to coordinate and credibly commit, in immutable code, to stand together if sufficient thresholds are met.

Then, whatever memes may come in our new world – whether they make us dance, make us depressed or make us withdraw all our money from the bank – we can at least have a fighting chance to do what us humans have always done: work together to survive.