Certain corners of the internet, an overlap of rationalists, GMU econ, and tech companies, have consistently outperformed prestigious institutions during the covid crisis. The outperformance was on almost every important margin, from predicting death tolls to mask wearing to approval of vaccines. Existing institutions ignored or ridiculed these successful subcultures.
As we are entering the endgame of the American covid experience, it’s worthwhile to stop and take stock. One response, aptly captured by the cartoon below, is to ignore the failures. Let’s refuse introspection and continue to put our trust in institutions.
Of course, in practice, it wasn’t hard to find information missed by the world’s top scientists and doctors. A 26 year old data scientist, Youyang Gu, managed to develop a model which was “consistently among the top”, and outperformed IHME model, which was being cited by Fauci and many news organizations. Youyang Gu won an Emergent Ventures Grant from Tyler Cowen and was widely read by the very online, but took months for his insights to be adopted by the credentialed class.
The lack of legibility from certain communities and the dominance of credentialed institutions meant consistently poor decisions from those in power. Most people are not very online. Even I was only vaguely aware of Gu, despite reading Marginal Revolution daily. It shouldn’t be surprising that it took months for Gu to be recognized, even if those months were deadly.
I interpret this slow response in part as a legibility challenge. Mainstream institutions are stuck in the 20th century. If you are legacy press reporting on covid, do you go ask for predictions from Imperial College London, the Institute for Health Metrics and Evaluation, or a guy on the internet with no epidemiological expertise? If you are a state agency making recommendations for tightening or loosening of the lockdown, do you trust reports from the New York Times, CDC, or a guy on the internet with no epidemiological expertise?
The very online have identified a new set of information curators who consistently outperform legacy institutions. The new information curators, however, are not legible to legacy institutions. Rather than evaluating predictive power of models, legacy institutions look for credentials.
While I have been pleasantly surprised by some professional communities, e.g. the epidemiological community beginning to embrace first doses and the infection modeling community improving their models, they did so because of immense amounts of attention and pressure. Institutions that aren’t facing a global crisis will likely continue decaying as they were before covid.
For America to thrive we must reverse this decay. The challenge for the 21st century is that 1) our institutions are failing and 2) most of those failures have slow feedback loops.
It should not be a surprise that GMU econ, rationalists, and tech were most effective at predicting and responding to covid. The cultures are overlapping and put an emphasis on correct predictions, understanding the impact of low probability high impact events, and prize truth telling over maintaining relationships. I suspect their success is not limited to covid. These communities likely hold better opinions, or can form better opinions, on many challenges facing America today.
Rebuilding American institutions therefore means elevating the GMU econ, rationalists, and tech communities into positions of authority. A key part of this is increasing the legibility of their insights. Different institutions have different ways of evaluating information. Legacy institutions discount blogs. Serious people write academic papers, serious people have credentials, etc. These norms are changing, albeit slowly.
Many people saw these circles of the internet being ahead of the curve. They will now read their blogs, listen to their podcasts, and follow their Twitter accounts. As the ideas reach an expanding audience, they have a positive impact on the organizations the audience participates in.
We are already seeing this happen. Marginal Revolution is one of the most widely read econ blogs, including by many in Washington DC. Tyler Cowen is a bestselling author. Ross Douthat was tweeting about Scott Alexander in 2018. Ezra Klein was mentioning Scott Alexander in Vox in 2014. The fact that such ideas are an important part of the discourse builds on years of work.
Another approach is to create new organizations that focus on translating relevant insights of these communities to a larger audience. A16Z is doing this by building a media arm. The media outlet will focus on technology, largely related to A16Z’s investment thesis that “software is eating the world”. Given the intended audience this will increase the legibility of the tech worldview.
Similarly, it’s possible to imagine an ‘Internet Bridge Institute’, aka IBI, for Washington DC. IBI would take arguments from internet subcultures, namely GMU econ, rationalists, and tech, and translate them for the DC policy world. This would increase the legibility of an important worldview for our times among policy makers. Additionally, there would likely be unforeseen benefits that come with such a collaboration.
Take education for example. Y Combinator, Lambda School, On Deck, and Primer are reinventing various aspects of schooling. They are sometimes hamstrung by the existing regulatory framework. Lambda School can’t offer ISAs in California for regulatory reasons. I’m no lawyer, but blocking a national platform from operating in a state seems to have implications for interstate commerce. These issues are a blind spot in current policy discussions. A search of AEI and Brookings, two of the pre-eminent DC think tanks, found only 2 mentions of Lambda School.
Covid shone a light on America’s troubled institutions. Reflection allows us to see what went right in addition to what went wrong. Unfortunately, the underlying challenges facing America are unlikely to get as much attention. Improving America means figuring out how to build better institutions, without the short feedback loops that covid afforded us. The sub-cultures that outperformed give us a good idea of where to start.