Epistemic status: written largely quite some time after I noticed the errors and from observing the UK and the US most. This is a draft, which I am very much looking for feedback on.
All crises are new, and their newness confounds us intellectually.
As societies become more complex and sophisticated, the ability of some groups to act decisively increases hugely. I’m thinking of both innovations such as Modern Monetary Theory (MMT) which tl;dr basically argues that the state can simply print money more or less ad infinitum.
All this power of action means that the Observe, Orient and Decide phases of OODA loop become counter-intuitively more important—the question is rarely if the state will be able to act, but how it can do the kind of information processing at scale to ensure that it does the right thing. I think therefore that as our powers increase, conceptual errors actually become more important over time and are likely to continue to do so. In other ways, of course, the hollowing out of the specifically neoliberal state after the collapse of the Bretton Woods agreements has meant that many of the other capacities of the state have been reduced. However, let’s park that for a moment.
The COVID-19 pandemic is perhaps the first time for a very long time that the ‘social whole’ had to be computed at once by large numbers of people.
This text is an attempt to state clearly some of the more persistent errors in thinking that I’ve noticed during the first year and a half of the pandemic. Like all complex, fast-moving events, pandemics put stress on our conceptions of the world. More importantly for the development of an adequate conceptual framework, such complex events place pressure on our models in a way that’s useful to respond to explicitly. Further such large-scale events will undoubtedly put further stress on our basic models, and so the process of updating now will hopefully pay off immensely.
They’re arranged in the order of when I started to notice them, which means they’re also approximately arranged in order of generality and obviousness from our vantage point now.
Presuming discreteness
This is the most obvious in the rejection of the connection between China and the rest of the world.
However, the real marker of this error is perhaps that the boundaries of what counts as a distant space are more or less entirely flexible, right up until you get the disease yourself. Italy, for example, for those in the UK, was still largely understood as somewhere very far away.
This is linked to the failure to grasp exponentially: the outbreaks were treated as very far away, and it was assumed that there would be plenty of warning before they got to an overwhelming level.
Presuming irrationality
This error, undoubtedly fuelled in part by racism, was that those who were closer to the action were the most likely to be making serious mistakes
A reason to be in favour of this argument is that people in complex situations often are genuinely unable to perceive the true dynamics of the situation they are involved in
The argument against it is that those who are closest to some phenomena are also those who have the clearest sense of the effects, especially when there is a great deal of novelty to those effects.
This was particularly prominent in the response to China, but it was also present in the response in the UK to Italy.
Contradictory downplaying
This one isn’t actually an error, but it looks like one. However, I think it’s actually plausible that it’s a sound argument given who the disease affects worst.
It’s most visible in the combination of two statements, both plausible, which I heard uttered sometimes by the same people: “Chinese people died because they’re unhealthy” and “Italian people died because they were kept alive by the excellent healthcare and therefore were old”
The prevention paradox
This is a commonly noted phenomenon in which people argue that measures which actively prevented a particular situation were in fact unnecessary because that situation did not occur.
In the early pandemic, this was especially prominent in the differing death rates of South Korea and Italy. It was clear that the virus was like the flu if and only if you treated it at the start like it wasn’t.
Failing to grasp exponentially
Everyone knows about this one: at the beginning especially, small numbers of cases were treated as a reason to be blasé about the disease as a whole.
A more subtle error comes in treating a 30% increase in virulence as the same as a 30% increase in contagiousness. The former being a linear measurement and the latter an exponential, the two are radically different after several iterations.
Rejecting the language of comorbidity
This one was consistent throughout the pandemic. Deaths were ascribed to more or less anything else except COVID.
Conversely, some people around me at least assumed that all deaths were due to COVID, as did I.
Presuming a universal form of disaster
Apparently empty hospitals (mostly revealed in footage of empty corridors) were taken as evidence of the unreality of the pandemic, even though the cancellation of elective procedures would be required by a highly contagious disease. Arguably, people thought these images were convincing because a much more familiar image of ‘disaster-movie’ overflowing hospitals could not be produced, as they had been in the wake of terrorist attacks or other mass casualty events.
Appealing to irrelevant expertise
This is particularly the case with appeals to medical doctors and nurses who were not epidemiologists.
This is probably the most understandable of the errors, given that medicine is often (wrongly) understood as a single discipline.
One particularly striking example of this fracturing I saw was on the podcast ‘This Week in Virology’ a group of five very knowledgeable virologists admitted that they had no idea about how the virus would evolve over time because they were not sufficiently expert in viral evolutionary dynamics. They committed to getting an expert in viral evolutionary dynamics on the show to talk about specifically this.
More contentious was the consistent appeal to authorities who had arguably drastically reduced their own credence by being conspicuously wrong and misleading in the early stages of the pandemic.
Lesson: the fracturing of contemporary expertise is extremely deep. This means that the number of experts is often much smaller than it appears, which reduces the ‘wisdom of crowds’ effects that science normally benefits from. However, it probably can’t be solved by training very good generalists.
Presuming stupidity on the part of the population
This error marked the communication efforts of medical authorities in the early pandemic and produced all kinds of distorting effects later on.
It was also evident in the patronising way that the use of ivermectin was discussed.
Conflicts between bureaucracy and urgent health measures
Most striking is the distinction between public health legislation and medical technology legislation in the US. This meant that rapid COVID tests fell into a gap between two forms of legislation and therefore could not be adequately approved.
Confusing views of discrete scales and preferring the most useful
This is perhaps most obvious in the relationship between the clinical view and the epidemiological: early in the pandemic, doctors’ expertise was relied on by some despite its broad irrelevance.
Presuming an unchanging set of causal drivers
In the UK, there was a consistent line from the left that the reason the pandemic had gotten out of control was that the government had failed.
Arguably, this was true in the very early stages of the pandemic, but within a few months, the main causal drivers had shifted to a much more complex mixture of personal action and policy.
Confusion of symptoms with infectiousness
Given that a great deal of the contagion happens
Exclusive focus on COVID-19 at the expense of other problems
Tension between the epidemiological view and the virological view
This comes in two forms: the first is the discussion of a given viruses’ virulence
The second is in treating epidemiological data (the R0 rate) as evidence of changes in the virus
This error is the same as not recognising the difference between the past and future behaviour.
Failure to grasp that one’s own actions are producing the landscape for future action
This is most particularly obvious in the case of the vaccine, where the vaccination program
The assumption that the science could be easily turned into a set of policy proposals
There is, unfortunately, no clear mapping between discoveries in science and policy.
The assumption that science as a whole will be unchanged by these events
Failure to appreciate the need to rollout new technologies
The number of steps
False dichotomies and silver bullet thinking
This is most apparent in the relationship between vaccines and medicines, or vaccines and masks.
Failure to understand the relationship between collective and individual benefit
The argument for everyone, regardless of risk, to be vaccinated, is that this will help drive the reproduction rate of any future strain below 1.
The presumption that COVID had replaced other threats
This was an error that I committed a lot.
Presumption of similarity of experience
From very early on, it was clear that COVID presented itself in a huge diversity of ways, in terms of symptoms, time to onset, and so on.
However, instead of a statistical understanding of the diversity of symptoms, people translated the experiences of those around them into a likely personal experience and disregarded other evidence (this may be a very localised kind of error).
The most egregious example of this is the UK government’s failure to update the symptom list from the ‘classic 3’ some 500 days after it was known that they were not the only symptoms to be aware of.
Presumption of almost infinite variability in experience
The direct opposite to the above error, some people assumed that more or less any kind of experience could be evidence of COVID. Although there were a huge variety of symptoms identified as being evidence of COVID (the ZOE COVID study being the most complete list that I know about), the list did not include every possible symptom.
This error often seemed to imply a belief that COVID had replaced other threats.
There are probably also lots of failures to think about complexity and scale in relation to the supply chain issues of the vaccine rollout, but I haven’t been able to itemise them in the same way.
These are lots of errors, and this is only a very scattered list that others are very welcome to improve on or edit. It seems implausible that there could be a single solution to this list of errors, but hopefully itemising them can allow some greater awareness. Given that some of the errors are in tension with others—that is, solving one of them can lead to problems with another—vigilance may be the only option.
Correctives and strategies
Use other people’s average gut feelings as a proxy. I think it’s reasonably safe to assume that most people are proceeding with relatively limited information—indeed, almost everyone is the vast majority of the time in novel situations. It’s possible to read the consequences of this average gut feeling insofar as it expresses itself in behaviour from the changes in case rates and so on. Then, by sampling the people around you and adjusting for how cautious they are relative to the average, you can compute how cautious you should be.
Cognitive Errors in the COVID-19 pandemic
Epistemic status: written largely quite some time after I noticed the errors and from observing the UK and the US most. This is a draft, which I am very much looking for feedback on.
All crises are new, and their newness confounds us intellectually.
As societies become more complex and sophisticated, the ability of some groups to act decisively increases hugely. I’m thinking of both innovations such as Modern Monetary Theory (MMT) which tl;dr basically argues that the state can simply print money more or less ad infinitum.
All this power of action means that the Observe, Orient and Decide phases of OODA loop become counter-intuitively more important—the question is rarely if the state will be able to act, but how it can do the kind of information processing at scale to ensure that it does the right thing. I think therefore that as our powers increase, conceptual errors actually become more important over time and are likely to continue to do so. In other ways, of course, the hollowing out of the specifically neoliberal state after the collapse of the Bretton Woods agreements has meant that many of the other capacities of the state have been reduced. However, let’s park that for a moment.
The COVID-19 pandemic is perhaps the first time for a very long time that the ‘social whole’ had to be computed at once by large numbers of people.
This text is an attempt to state clearly some of the more persistent errors in thinking that I’ve noticed during the first year and a half of the pandemic. Like all complex, fast-moving events, pandemics put stress on our conceptions of the world. More importantly for the development of an adequate conceptual framework, such complex events place pressure on our models in a way that’s useful to respond to explicitly. Further such large-scale events will undoubtedly put further stress on our basic models, and so the process of updating now will hopefully pay off immensely.
They’re arranged in the order of when I started to notice them, which means they’re also approximately arranged in order of generality and obviousness from our vantage point now.
Presuming discreteness
This is the most obvious in the rejection of the connection between China and the rest of the world.
However, the real marker of this error is perhaps that the boundaries of what counts as a distant space are more or less entirely flexible, right up until you get the disease yourself. Italy, for example, for those in the UK, was still largely understood as somewhere very far away.
This is linked to the failure to grasp exponentially: the outbreaks were treated as very far away, and it was assumed that there would be plenty of warning before they got to an overwhelming level.
Presuming irrationality
This error, undoubtedly fuelled in part by racism, was that those who were closer to the action were the most likely to be making serious mistakes
A reason to be in favour of this argument is that people in complex situations often are genuinely unable to perceive the true dynamics of the situation they are involved in
The argument against it is that those who are closest to some phenomena are also those who have the clearest sense of the effects, especially when there is a great deal of novelty to those effects.
This was particularly prominent in the response to China, but it was also present in the response in the UK to Italy.
Contradictory downplaying
This one isn’t actually an error, but it looks like one. However, I think it’s actually plausible that it’s a sound argument given who the disease affects worst.
It’s most visible in the combination of two statements, both plausible, which I heard uttered sometimes by the same people: “Chinese people died because they’re unhealthy” and “Italian people died because they were kept alive by the excellent healthcare and therefore were old”
The prevention paradox
This is a commonly noted phenomenon in which people argue that measures which actively prevented a particular situation were in fact unnecessary because that situation did not occur.
In the early pandemic, this was especially prominent in the differing death rates of South Korea and Italy. It was clear that the virus was like the flu if and only if you treated it at the start like it wasn’t.
Failing to grasp exponentially
Everyone knows about this one: at the beginning especially, small numbers of cases were treated as a reason to be blasé about the disease as a whole.
A more subtle error comes in treating a 30% increase in virulence as the same as a 30% increase in contagiousness. The former being a linear measurement and the latter an exponential, the two are radically different after several iterations.
Rejecting the language of comorbidity
This one was consistent throughout the pandemic. Deaths were ascribed to more or less anything else except COVID.
Conversely, some people around me at least assumed that all deaths were due to COVID, as did I.
Presuming a universal form of disaster
Apparently empty hospitals (mostly revealed in footage of empty corridors) were taken as evidence of the unreality of the pandemic, even though the cancellation of elective procedures would be required by a highly contagious disease. Arguably, people thought these images were convincing because a much more familiar image of ‘disaster-movie’ overflowing hospitals could not be produced, as they had been in the wake of terrorist attacks or other mass casualty events.
Appealing to irrelevant expertise
This is particularly the case with appeals to medical doctors and nurses who were not epidemiologists.
This is probably the most understandable of the errors, given that medicine is often (wrongly) understood as a single discipline.
One particularly striking example of this fracturing I saw was on the podcast ‘This Week in Virology’ a group of five very knowledgeable virologists admitted that they had no idea about how the virus would evolve over time because they were not sufficiently expert in viral evolutionary dynamics. They committed to getting an expert in viral evolutionary dynamics on the show to talk about specifically this.
More contentious was the consistent appeal to authorities who had arguably drastically reduced their own credence by being conspicuously wrong and misleading in the early stages of the pandemic.
Lesson: the fracturing of contemporary expertise is extremely deep. This means that the number of experts is often much smaller than it appears, which reduces the ‘wisdom of crowds’ effects that science normally benefits from. However, it probably can’t be solved by training very good generalists.
Presuming stupidity on the part of the population
This error marked the communication efforts of medical authorities in the early pandemic and produced all kinds of distorting effects later on.
It was also evident in the patronising way that the use of ivermectin was discussed.
Conflicts between bureaucracy and urgent health measures
Most striking is the distinction between public health legislation and medical technology legislation in the US. This meant that rapid COVID tests fell into a gap between two forms of legislation and therefore could not be adequately approved.
Confusing views of discrete scales and preferring the most useful
This is perhaps most obvious in the relationship between the clinical view and the epidemiological: early in the pandemic, doctors’ expertise was relied on by some despite its broad irrelevance.
Presuming an unchanging set of causal drivers
In the UK, there was a consistent line from the left that the reason the pandemic had gotten out of control was that the government had failed.
Arguably, this was true in the very early stages of the pandemic, but within a few months, the main causal drivers had shifted to a much more complex mixture of personal action and policy.
Confusion of symptoms with infectiousness
Given that a great deal of the contagion happens
Exclusive focus on COVID-19 at the expense of other problems
Tension between the epidemiological view and the virological view
This comes in two forms: the first is the discussion of a given viruses’ virulence
The second is in treating epidemiological data (the R0 rate) as evidence of changes in the virus
This error is the same as not recognising the difference between the past and future behaviour.
Failure to grasp that one’s own actions are producing the landscape for future action
This is most particularly obvious in the case of the vaccine, where the vaccination program
The assumption that the science could be easily turned into a set of policy proposals
There is, unfortunately, no clear mapping between discoveries in science and policy.
The assumption that science as a whole will be unchanged by these events
Failure to appreciate the need to rollout new technologies
The number of steps
False dichotomies and silver bullet thinking
This is most apparent in the relationship between vaccines and medicines, or vaccines and masks.
Failure to understand the relationship between collective and individual benefit
The argument for everyone, regardless of risk, to be vaccinated, is that this will help drive the reproduction rate of any future strain below 1.
The presumption that COVID had replaced other threats
This was an error that I committed a lot.
Presumption of similarity of experience
From very early on, it was clear that COVID presented itself in a huge diversity of ways, in terms of symptoms, time to onset, and so on.
However, instead of a statistical understanding of the diversity of symptoms, people translated the experiences of those around them into a likely personal experience and disregarded other evidence (this may be a very localised kind of error).
The most egregious example of this is the UK government’s failure to update the symptom list from the ‘classic 3’ some 500 days after it was known that they were not the only symptoms to be aware of.
Presumption of almost infinite variability in experience
The direct opposite to the above error, some people assumed that more or less any kind of experience could be evidence of COVID. Although there were a huge variety of symptoms identified as being evidence of COVID (the ZOE COVID study being the most complete list that I know about), the list did not include every possible symptom.
This error often seemed to imply a belief that COVID had replaced other threats.
There are probably also lots of failures to think about complexity and scale in relation to the supply chain issues of the vaccine rollout, but I haven’t been able to itemise them in the same way.
These are lots of errors, and this is only a very scattered list that others are very welcome to improve on or edit. It seems implausible that there could be a single solution to this list of errors, but hopefully itemising them can allow some greater awareness. Given that some of the errors are in tension with others—that is, solving one of them can lead to problems with another—vigilance may be the only option.
Correctives and strategies
Use other people’s average gut feelings as a proxy. I think it’s reasonably safe to assume that most people are proceeding with relatively limited information—indeed, almost everyone is the vast majority of the time in novel situations. It’s possible to read the consequences of this average gut feeling insofar as it expresses itself in behaviour from the changes in case rates and so on. Then, by sampling the people around you and adjusting for how cautious they are relative to the average, you can compute how cautious you should be.