“Never attribute to malice that which is adequately explained by stupidity”Hanlon’s Razor
Those looking for an explanation for the absurdly dysfunctional state of modern western society usually turn to one of two explanations. Either the problems are caused by individual leaders and their failings e.g. a senile Biden, a narcissistic Trudeau or a clownish Boris Johnson, or they are the deliberate attempt on the part of some group of masterminds to bring about a new world order; in other words, a conspiracy theory.
Part of the reason why these explanations are preferred is because they allow the possibility of redress. You can vote out an idiot leader and get a better one. And a conspiracy theory can be unearthed and held to justice.
The large-scale conspiracy theory is also quite flattering to the human ego. It implies a team of evil super geniuses who are so intelligent they that they are able to pull the strings of all nation states simultaneously turning the leaders of countries into puppets while hiding their nefarious intent from the general public. It’s pleasing to think that we are capable of that level of intelligence and competence. It’s also a comforting thought because the bad guys can be brought to justice. All we have to do is uncover their dastardly plot and bring them before the courts. The German lawyer, Reiner Füllmich, has been playing on this idea right from the start of corona with promises of convictions against those who pushed the “public health” measures. Unless I missed the news, none of his attempts have succeeded.
In this post we’ll sketch out an alternative explanation which is that the system itself is the problem. When the system is the problem it becomes really hard, maybe even impossible, for individuals to make a difference. I have seen such a dynamic with my own eyes in the form of dysfunctional organisations where new management was brought in to fix things. These were intelligent people who knew what was wrong and had a plan to make it right. But organisations are systems and systems have their own dynamic that is independent of any of the individuals involved. A system also has an external context that affects it. An organisation in a dying industry cannot be saved no matter how smart the people who are trying to save it. For these reasons and others, the system-based explanation is less gratifying and therefore less popular. But that doesn’t make it less truthful.
We’ll use the example of modern science to explore this concept because the question of whether science is corrupt or incompetent has become quite urgent in the last two years as we have watched the corona debacle unfold.
There are at least two underlying assumptions in our general culture when it comes to science. Firstly is the idea that anybody can do science as long as they have access to education. We can represent this graphically as follows:
That is, everybody has the same amount of innate ability to do science and the only thing preventing them from realising that ability is a proper education.
The second assumption is that all science problems are equally solvable which we can represent as follows:
Another way of saying this is that all science problems are equally complex. If you assign equal resources (time, money, people) you will get equal results.
Let’s look at a different model for both of these using the “Zipf curve” which has been shown to hold across numerous domains. Note that the Zipf curve matches the Marginal Benefit curve used in economics to capture the concept of diminishing returns.
What this graph aims to capture is the idea that the more people you train in “science” the less quality of scientist you get. This can be for reasons of nature or nurture (or a combination of both). The innate talents required to become a high quality scientist are not shared equally in the population and we would expect something like a Zipf curve to represent the distribution of those talents in the same way that not everybody has the collection of talents required to become a professional athlete.
When it comes to intellectual matters, you might argue that we could make up the difference in talent through education. But even if that were true, it would be more costly to educate the less talented people as they will need more time to develop their skills and knowledge. Once again you would run into a Zipf curve where the Marginal Benefit from education falls because the Marginal Cost rises. If we further assume that the talent pool of science educators is also a Zipf curve, then the quality of education would fall the more people get educated because there aren’t enough good teachers to teach them. Either way, you still end up with a curve of diminishing returns.
But what happens if the domain of “science problems” is also a Zipf curve? This would look as follows.
What this curve describes is the “low hanging fruit” dynamic. The problem domain of science is not equally distributed. There are a set of problems which are more simple and therefore more easily solved while the majority of problems are more complex.
If we combine these two concepts we get a story about the evolution of science. In a time of resource constraint such as in the 1800s, only the most talented people become scientists (assuming a relatively merit-based system of resource allocation). Those scientists will be working on the relatively simple problems in the field and therefore they produce the most valuable results. Everybody gets excited by the results and as wealth accumulates we throw more resources into the field expecting even more impressive results.
The extra resources would produce more results if the curves for ability and simplicity were flat. But they are not. They are Zipf curves. What happens, therefore, is that less capable scientists are put to work on more complex problems i.e. the intersection of the two diminishing Marginal Benefit curves. We spend more money to get fewer, less valuable results. But even though the Marginal Benefit falls, the overall cost-benefit equation might still be positive.
The problem of diminishing returns is exacerbated by a third consideration. More resources means more people are working on “science” and adding more people reduces the quality of communication. The following diagram is often used to summarise this problem.
Communication becomes more difficult as the number of people involved increases even if the quality of the information remains static. But if less talented people are working on more complex problems, we would expect the information quality to degrade leading to a situation where there is more communication of lower quality information. In short, the signal-to-noise ratio goes to hell.
You can see this dynamic even in small groups. Take a musical band, for example. If there are five people in the band, it only takes one person to be “out” for the whole band to be “out”. Similarly, on a small engineering team if one person doesn’t understand, this effects the overall communication flow because erroneous messaging is introduced and more time needs to be spent correcting the errors. If the person is a line level worker, it’s usually possible to work around them and try and exclude them from communication. But I have been on teams where the person who didn’t understand was the senior manager on the team. It’s a lot harder to move a senior manager out of the way so that things can get done.
A key thing to bear in mind is that a low signal-to-noise ratio won’t appear to be obviously “wrong”. This is true both for the people on the inside doing the work and also to external participants. Noisy communication is worse than the case where communication is “wrong”. Wrong communication is almost as useful as right communication. If you know somebody is always wrong, you just invert whatever they say and now you have truth. You can’t do that with noisy communication. Noisy communication is ambivalent, unclear and confusing. Again, the musical group example is a useful here. A somewhat incompetent band doesn’t sound “wrong” but rather “blah” or “meh”. You shrug your shoulders and say something like “it’s not bad but it’s not good either”. This is in contrast to a band like Nickelback who are technically proficient musicians that happen to make bad music.
When the signal-to-noise ratio is low, it becomes far more difficult to show that something is wrong because there are no clear and obvious errors. There is no smoking gun that will set the record straight and restore order. Rather, there is an accumulation of numerous small errors which are much harder and more time consuming to identify and correct. In a small group such as a band, it’s possible to find the weak link (usually the drummer) and get rid of them. In larger groups it becomes far more difficult and in really large organisations like corporations and government departments it’s as good as impossible.
It’s important to understand that this dynamic of noise accumulation occurs before politics, commercial money and the enormous egos of billionaires and celebrities gets involved to make things even more confusing. Corona provides a useful case study. There was never any reason to believe that the mRNA vaccines would work to end a pandemic. The science had not proven the matter one way or another. To put it in terms we have been using, the science had a low signal-to-noise ratio. This meant it was possible to believe that the vaccines “might” work. After all, anything “might” happen. Once upon a time, science was about “laws” and was founded upon hard-nosed cause and effect relationships that had been empirically proven. That’s the kind of science you see at the “simple” end of the Zipf curve. But as complexity increases, the clarity of understanding diminishes and you no longer have “laws” but “guidelines”.
Once the vaccine question became political, the political imperatives took over and politicians had to gloss over the inherent ambiguity in the science. Thus, we were assured the vaccines were “safe and effective”. Meanwhile, corporations which exist to maximise shareholder value were happy to sell a product when governments indemnified them against legal liability.
It’s not a coincidence that the corona event took place in the domain of viral disease as this is arguably one of the more complex scientific domains. I would place it somewhere about here on the graph. In other words, highly complex.
Note that viral disease as an object of study also has a built-in communication problem because it runs over three separate scientific disciplines: virology, epidemiology and medicine and that’s before you consider the mathematical epidemiologists, the immunologists and other sub-sub-disciplines. Viral disease is firmly in the category of study that the systems thinkers of the 20th century posited was not amenable to reductionist science which means it cannot be simplified to the point where calculation can be done. The best we can do is assemble cross-disciplinary teams to undertake research aimed at obtaining general principles of action. Those general principles were exactly what constituted the public health guidelines that were the accepted wisdom of how to deal with a pandemic prior to March 2020.
The post war period has seen huge amounts of resources pumped into science and yet we have ended up with the “reproducibility crisis”. The reproducibility crisis is just another word for the noise generated by the intersection of multiple diminishing returns. No amount of extra education and training and money will solve the problem. The result is not error but noise and when the noise gets raised to a high enough degree you have a situation where anybody can read into it whatever they like. At that point, science becomes a giant Rorschach Test.
The problem of a low signal-to-noise ratio is not limited to science. Most things in the modern world suffer from it. Everything is “blah” and “meh”. It’s the paradox of success. We have huge resources to apply to problems and we invest those resources into new ventures. It works for a little while but the law of diminishing returns means that everything quickly turns to mud and the quality of everything falls sharply. This is true in the consumer economy, in the political sphere, in the media, in the arts and in science and technology. Rather than accept this as a fact of life, we pump more resources in until the returns turn negative and that leads to inflation and the debasement not just of the currency but of political, social and cultural capital. We’re pretty far into that dynamic right now and it’ll probably get worse before it gets better.
It’s partly for this reason that societies and cultures seem to peak when strict resource limits are in place. Without limits, the signal-to-noise ratio falls and everything becomes saturated and over-exposed. The noise floor steadily rises until and only those who can shout the loudest get heard. To quote the New Zealand Prime Minister during corona, “we (the government) will be your single source of truth.” The words that usher in the age Caesarism.