The "Democratic Centralism" of COVID
The anonymous author of the satirical “Homeless Camping in Austin: A Modest Proposal” has also sent me this more serious guest post. The title is mine. “Democratic centralism,” you may recall, is the Leninist practice of demanding strict loyalty to a party line after a (usually perfunctory) debate. Printed with the author’s permission.
Well-read readers will note the parallels to Robin Hanson’s job market paper, “Warning Labels as Cheap Talk: Why Regulators Ban Drugs.”
Since very early in the pandemic, there has been a somewhat novel approach to information flow in the media and particularly on social media sites, at least compared to the baseline in the western world. Very quickly, there seems to have been a consensus that information gatekeepers should determine which opinions about the nature of the coronavirus and the appropriate policy response should be allowed to be widely disseminated. One of the notable early examples was when Medium, which is basically a website host, took down a piece written by Aaron Ginn arguing that the costs of lockdowns should be considered. There was no basis to argue that he was providing disinformation; his post was removed because it argued for a different position than what was being promoted.
While biased journalism is hardly new in the US or anywhere, the movement to close down the ability to distribute alternative opinions seems to have been novel at least within the United States. This was not like the New York Times refusing to publish opinion pieces that disagreed with its editorial stand; this was more like if the people in the olden days who sold bulk newsprint paper refused to allow anyone who dissented from the views of the newsprint providers to even obtain raw materials for printing.
This new attitude is puzzling given the novelty of the virus and the nearly intractable nature of the optimal policy decision, which must take into account the likely spread of the virus under various policies and the overall effect of the policies on the enormously complex and interconnected global economy. It is frankly absurd to think that by March or April all reasonable people had converged to the consensus view that the world economy should be locked down, but major press outlets and information platforms proceeded as if this was established fact. Given the extraordinarily poor performance of even the relatively simple virus models that were applied to the consensus view and the total inability to even begin to estimate the economic and human costs of the lockdowns, in retrospect this rapid convergence on consensus appears to be one of the single greatest acts of hubris in the history of mankind.
But, crucially, even at the time and without the benefit of hindsight this rapid collapse onto a single acceptable viewpoint by those who control the flow of information should have been seen as a colossal error. Modern information economics makes it abundantly clear that in the presence of biased experts whose objectives do not perfectly align with the people receiving advice, having multiple experts, each with their own different biases and preferences, is much better than having a single biased expert. This is true even if you could chose the least biased expert as your one expert. Adding another, highly biased, expert, will greatly improve the quality of information available to the person being advised in the richer versions of these “persuasion” models. The idea is that a single expert, even if he cannot directly deceive, will choose to report information in such a way as to influence the people being advised to act in the interests of the expert, rather than their own. Those being advised will know that the expert is engaged in such behavior, but even though they do their best to compensate for his manipulations the fact that he is the sole arbiter of what information does or does not flow gives the expert great power to influence the ultimate decision. Bringing in another expert with different biases improves the situation immensely (as shown by Gentzkow and Kamacia, Bayesian Persuasion with Multiple Senders and Rich Signal Spaces), as this other expert will release information to undermine the persuasion attempts of first expert when the two experts’ objectives disagree. Knowing this, the first expert will release even more information to undermine the persuasion attempts of the second expert, and this process can, in appropriate circumstances, leave the person being advised (us) with all the information we need to make the best judgements for ourselves.
Certainly no one could argue that the experts designated as the arbiters of truth by the information platforms were free of bias; the WHO and those who work for the WHO must constantly consider the risk of offending powerful countries, particularly China, and the US health bureaucracies are shot through from top to bottom with political biases. Notably, US public health authorities barely hid their political biases as they opposed lockdown protests and endorsed “racial justice” protests. So, there is no reasonable claim that Facebook and Google had magically found an unbiased expert. As such, the logic of information design should have pushed them to open their platforms to many different perspectives on the available data, if, as certainly appears not to be the case, their true interest was in promoting public health.
Even in our own school, a business school at a university that should be dedicated to the relentless pursuit and dissemination of knowledge, administrators pressured faculty to not present ideas contrary to the universities official views on coronavirus policy. This pressure was particularly interesting as there is no real sense in which the University should take an official position on coronavirus policy, and it is unclear that the university has done so. But, faculty are pressured not to contradict this view, and such pressure is acceptable at the highest level of the business school. Our school thus rejects the idea that there should be a competitive marketplace of ideas and instead embraces the much more totalitarian idea that once a consensus is reached by a designated inside group, all others should fall in line and not question these thought leaders.
As distressing as this move away from allowing ideas to freely compete is, the situation gets worse. It is essential to recognize the precise nature of the filtering that was undertaken by information platforms. The decision was made to designate certain opinions produced by certain authorities as the only acceptable opinions. This filter was not based on expertise; one could imagine a filter where only “qualified” individuals were allowed to opine on the coronavirus on certain platforms. But, Dr. Scott Altas was censored, despite being a medical doctor and a health policy expert. A more defensible filter than even one based on expertise would be a filter based on historic predictive ability; perhaps the platforms could have prevented those who made the worst predictions about previous pandemics and about the current coronavirus from further pontificating. But, Neil Ferguson, who made order of magnitude errors about virtually every recent pandemic including SARS-CoV-2, was never censored.
So, it appears certain platforms simply selected certain central authorities to fully back, regardless of actual expertise or a history of successful predictions. It is thus impossible to conclude that these platforms sought to filter information in order to promote better knowledge among the general public, which as discussed above would already have been a dubious proposition. Instead, it appears that the purpose of the information filters must have been control of information for the sake of control of information. The implications of this objective are disturbing.