I write this response in reaction to two excellent posts that appeared on Medium that I read recently. One is by Kate Starbird. The other is by Danah Boyd. Both are worth reading on their own.
My take on so-called Fake News is different from theirs, and from what most other people seem to think as well, as far as I can tell. I do start from the same premise as what many others have pointed out: there is no way to objective define what is and what is not “fake” news and attempting to haphazardly enforce the “correct” news by, in effect, censorship, is counterproductive. However, I think there is a huge share of blame to be had for the conventional media outlets who peddle information, as well as various actors in politics, economy, and society who serve as the source of the information that they traffic in.
My point of departure is, I think, quite different: one particularly impressive paper that I recently read is a history master’s thesis by a Canadian graduate student at the University of Calgary, concerning the persistence of deeply misinformed and self-deluding assessment of Japanese air capabilities in 1930s. The problem faced by military intelligence is, at its core, fundamentally the same as that of any consumer of information. We do not know the truth, but we have some idea of what the truth looks like and we try to update our existing beliefs in light of the new information that we obtain, evaluated through the prism of our existing beliefs, with the caveat that the new information includes both good and bad. We do not necessarily know, a priori, distinction between the good and the bad, at least based on their content. However, we are able to evaluate the quality information specific to the domain that we are familiar with.
Pyke, the Canadian history student, has effectively conducted a two factor design: the control was provided by the Imperial Japanese surface fleet, which remained mostly closed to western observers throughout 1920s and 30s; the treatment took the form of the air arms, of both the Imperial Army and the Navy, which, in 1920s, was extraordinarily open to Western observers — because Japanese capabilities were in their infancy — but access to which became extremely restrictive in 1930s, as the Japanese began to develop their own approach to air warfare and technology. The evaluation of the Japanese naval capabilities among the Western military remained fairly accurate through the entire period, even if direct access to Japanese naval installations was denied. The evaluation of the Japanese air capabilities swung wildly, however: while the assessment was extraordinarily accurate in 1920s, even if tinged with stereotypes, they became laughably inaccurate, where wild stereotypes dominated whatever facts that made their way to USNI. Japanese surface fleet was never underestimated the way Japanese air power was throughout 1930s.
Why this difference? The answer lies in part with the state of USNI’s priors about Japanese military in 1920s. IJN surface fleet was already an established service, with well-established capabilities that demonstrated itself to be comparable to its Western counterparts. Even if very little detailed information about it was available, US could take the information it could obtain and filter them through its own practices as guideline: since we don’t know much about them, imagine that they are like us, and anticipate what they would do if we were in the same situation. This approach led to certain blind spots — e.g. IJN’s focus on night fighting, even without radar, and unusual reliance on torpedoes. But, overall, the assessment was fairly accurate. With regards the air arm, the situation was different. In 1920, Japan had very little background or capabilities in air warfare or technology and was indeed trying to put an air arm together by copying everything. It gave Western powers full access to its program precisely because it had nothing to hide — everything was indeed a copy. While the copycat nature of Japanese air services in 1920s might have shaped the mean perception of their capabilities, however, the outliers did not escape notice: where Japanese excelled, Western observers saw and noted. By 1930s, however, when good information dried up, the perception that,when it came to the air arm, Japanese were a nation of inferior copycats persisted, even in face of occasional contrary evidence.
In Bayesian terms, this is predictable. The conclusion that the Japanese air services are full of inferior copies was backed up by a vast amount of evidence of very high fidelity, obtained in the period of nearly unrestricted access to Western military observers. The contrary evidence obtained in 1930s were few and far between and were of far inferior quality, thanks to the restrictions placed by Japanese military authorities to Western observers, with nuggets of high quality information mixed in with absurd nonsense, with little means to distinguish between them, except far in retrospect. In face of few new observations, of very high variance, the priors persist. The new information provided little reason for USNI to update its beliefs about the Japanese air services.
The defining characteristic about the informational environment today is not simply that there is more information, but that there is a cacophony of information: it’s not just that the sample size has increased, the variance has as well. The reaction by the traditional media outlets to the increased competition was to withdraw resources from peddling high quality information, and instead sought to appeal to varying “tastes” of different market segments. So you have news for young people, old people, liberals, conservatives, or whatever. But where is the good news? Well, good news didn’t pay the bills, and, financially, that was correct.
The problem is that, if the intrinsic quality of the information declines, there is no value in listening to someone whom you might not trust. I might believe someone whom I would not normally trust if he can explain his viewpoint in logical terms that I can understand: I may not necessarily come to agree with his worldview, but I can certainly understand where he comes from, why he thinks as he does, and based on this understanding, we can at least agree on some common ground. In other words, quality of information and taste for a given flavor of information occupy different dimensions, but they are not quite orthogonal: it is possible that good information might be packaged in a manner that makes sense to different audiences — but only if the provider of the information makes a good faith effort, which will be difficult. Instead, what seems to have taken place is exactly the opposite: as the overall quality of the information declined, media outlets tried to make it up by wrapping poor quality information in colorful packaging, different one for different audiences.
The problem is that there is little difference in increasingly low quality information wrapped in ever more colorful packaging and even lower quality information wrapped in even more colorful packaging. New York Times may be still a better source of information than a conspiracy theorist, but, for someone to whom NYT does not really talk to, on topics that he is not very familiar with, why should NYT be trusted any more than a conspiracy theorist? This is not necessarily a choice to actively trust the conspiracy theorist over NYT, but where the noise generated by conspiracy theorists is actively obscuring what the truth is, and the mainstream media outlets, with increasingly limited resources and distracted by a desire to sell to different market segments, are themselves becoming too noisy. Amidst all the noise, there is little to believe, or, with which the consumers of information to update their priors with. I imagine readers of conspiracy theories actually consume, directly or indirectly, plenty of New York Times, too — conspiracy theorists cite NYT plenty, even if they don’t actively read them. But to them, NYT is just part of the noise, as long as it does not talk to them conscientiously. Even if they don’t actually trust the conspiracy theorists — who may well be just more noise — this is real problem: that for many people, there is no difference between Breitbart and NYT.
I used to cringe whenever I heard the slogan, “information wants to be free.” Free information, easy to obtain and easy to disseminate, is literally worth nothing. It is just a lot of noise. Pure noise does nothing to reshape people’s beliefs — although, in a way, it is fascinating to study. But, when good information is too expensive, whether in monetary or cognitive costs — why bother reading news that doesn’t talk to you? — bad but free information crowds them out. (Hardly a new idea: Sam Popkin, in his 1993 book, termed in “Gresham’s Law of Information.) If you want to drive out the bad but cheap information and cut through the noise, make the good information cheap. Not just in monetary terms, but in cognitive terms as well: try to actively communicate with the people who are not consuming them, in terms that they can understand, and rather than try to agree with your beliefs, try to find common ground, in terms of how the world works. This kind of information certainly will not be cheap and it will not want to be free all by itself. This will need a conscientious and costly effort to disseminate.