I’ve been trying to find a word for this? Structural isomorphism? Methodological similarity? I’m very open to suggestions! How about CONCEPTUALLY ANALOGOUS?
Epistemology: The theory of knowledge, especially with regard to its methods, validity, and scope. Epistemology is the investigation of what distinguishes justified belief from opinion.
“there are a lot of meta-features of different kinds of events that create conceptual tool-kits that can be brought over, even though on the surface, the event is a different category. I’ve been trying to find a word for this? Structural isomorphism? Methodological similarity? I’m very open to suggestions!”
Structural isomorphism is too restrictive. Think of oat milk. It’s not cow milk (or goat or sheep) but many people want them to behave similarly. When they don’t behave in the same way, people try to find work arounds. That makes it functional or behavioral isomorphism, if isomorphism itself is the right word.
Methodological similarity is similarly too restrictive, since this is behavior that people exhibit under everyday circumstances, not just academic research. Again, oat milk.
It’s more a form of analogy: using an example you know well as an analog for what you don’t know. It can often get you into the new example because there are similarities, or can yield incredible insights because of how the two differ,
$0.02... whether 'analogy' is a better fit than, say, 'homology' depends on whether you want to emphasize the disjointness of the subject matter domains, or the behavioral toolkits we execute over them, which have a lot in common and a familial relationship. I tend to think the latter is key here, but everyone's mileage may vary...
I see your bid and I raise you to $0.03. An avian wing is homologous to a human arm. A bee’s wing or an airplane wing are analogous to a bird’s wing. And then there are bats, whose front legs are homologous to human arms, and whose wings are analogous to birds’ wings.
Clearly the word “similarity” is too vague. We can solve that through obfuscation by using the homologous Greek word “homoiotis.” Sounds technical and fancy that way.
Imho "solve ... through obfuscation" is worth at least a buck :) But the arm-wing-leg thing is exactly on point. If we'd like to leverage what we know about plane crash analysis (or whatever) to optimize pandemic response (or whatever), and we're looking at some aspect, let's say stakeholder withholding of information (whatever), then drilling into whether the similarity is deeply rooted or merely convergent is a key part of what we're talking about. I think.
I’ve run through a number of scenarios from the many disciplines I have any experience in and have found the pattern seems to be that “deeply rooted” correlates to “structural” and “convergent” to “functional” or “situational.”
Rereading the whole conversation, however, I now see something I hadn’t seen yesterday. You’re addressing issues about the architecture of the problem space: How can we look at this problem in a way that we can make some headway to understanding? This strikes me as having a lot of shared ground with the idea of design patterns, used originally in architecture by Christopher Alexander, and then extended (analogously) to software development. It’s the idea of approaching a situation in terms of a working design, not starting from scratch every time. The level of abstraction is approximately the same as you’re striving for.
Yeah I thought about "deep structure" (you know from which field. :-D). But that's exactly it: the architecture of the problem space is isomorphic or analogous, despite really different areas.
This conversation is a fair way above my head, but I'm wondering if the situation where a doctor was watching a NASCAR race and wondered how it was that the team managed minor repairs so quickly. The doctor learned that the team keeps complete kits of common repairs within easy reach at all times, cutting down on response time. Said doctor then applied this thinking to surgeries, and it's now common practice for kits to address many common medical emergencies to be easily accessible within ER units (or maybe OR rooms, I can't remember), which leads to lower mortality rates because common emergencies can be addressed more quickly. Because of car racing.
Taking a look at Wikipedia's disambiguation page for 'homology' (https://en.wikipedia.org/wiki/Homology), I see a group that includes Homology (anthropology), Homology (psychology), and Homology (sociology) that all revolve around the 'deeply-rooted similarity' notion.
I guess my speculation is that a Homology (technosociology) could someday capture a part of the idea you're developing about the way we do or fail to make and act on the same learnings across different subject matter domains (different methods for rare events being a perfect example). The hypothesized 'deep roots' would be in our human nature, existing cultures, and - especially, given (what I perceive to be) your favorite angles of attack - pervasive patterns of use and misuse of information at scale under the pressure of rapidly advancing technology.
Of course, even if you buy this, it's only one of many things you could choose to emphasize in your choice of terminology. If it doesn't resonate with you, that's good enough for me!
Yes. It provides a starting point, which may fall apart as you go deeper, or may bear fruit (or viruses), but at least it’s not a blank sheet of paper.
It was a little unclear from the original formulation whether the emphasis is problematic scenarios in which a conceptual toolkit is inappropriately brought over and misapplied (she spoke of "conflating" terrorism with structural ethnic tensions which would imply a fallacy), or the more neutral observation that these meta-principles can be usefully applied in other situations that look different on the surface. That might have implications for the terminology.
Putting that aside: if the point of interest is behavior and events that originally led to the outcome, I feel like the "functional" vs. "topographic/topological" distinction works best. If the focal point is the forensic task of inferring post-hoc what might have happened (reasoning and methodology, etc.), then maybe a more figurative, analogy-adjacent term that evokes the idea of "carrying over" is appropriate. How about the term "resonances" or "tropes"? Or "epistemic affordances?"
Meta-principles! There is a structural similarity: rare event, background event not uncommon but requires a chain of unlucky incidents to come together, can have catastrophic/terrible consequences. Different than anything, say, on a normal distribution of occurrence frequency .Tail risk, but the process leading to the tail risk is also specific and shared.
Oh I see. That's helpful, thanks. I had a couple other thoughts on this subject, that hopefully are not too late to share here (totally fine if no one jumps on them). One is that I wonder if there's a difference between drawing upon adjacent or analogous kinds of situations (like the plane crash example) that bear hidden structural similarities, and drawing on historical knowledge and accumulated hindsight about related kinds of events in the past, which echo or rhyme in this new setting. In that sense it wouldn't be the horizontal or vertical dimensions of knowledge that we've discussed here before, but a more temporal-historical, geneological dimension. The reason I was attracted to the term "resonances" or "tropes," or the notion of "evoking," is that they speak to this more historical - even humanities - perspective.
Another stray thought is that while we have been relying on other big disasters for insight, perhaps there are more mundane examples of daily personal mishaps or micro-"tragedies" that point toward useful analogues as well, which would function at larger scale in disasters. Take the film 127 Hours, the James Franco one about the hiker in Utah who was unlucky enough to fall into a crevice in such a way that a boulder was dislodged that happened to pin his arm in exactly the wrong-sized space (forcing him to cut it off if he hoped to survive). This would appear on the surface to just be a case of bad luck. Yet it feels vaguely reminiscent of (hopefully less dramatic) "disasters" we've all found ourselves in where not just one thing goes wrong, but three things at the same time that somehow work to reinforce one another. Is this merely the inevitable statistical result of many, many repeated trials from which an eventual tail event was inevitable, which would have occurred some other way eventually? Or, does it reflect something more interesting, the interesting features of which make it ultimately preventable?
You’ve added value judgement needlessly to your long tail. Good results can come together as well in the same way. The business cliché in vogue a couple of years ago was “perfect storm,” or “planetary alignment.”
That speaks to another question, of whether the dynamics of tail events we would assess as "positive" function any differently from ones we are inclined to experience as "negative" - or whether that's purely reflective of our value judgments and social context and nothing to do with underlying structure and form. Perhaps it depends on whether the "terrible" scenario is terrible due to entropy, or due to a really terrible negentropy?
epistemic affordances … I assume Gibson-type affordances … is interesting to try to visualize.
so, in essence, you (the speaker) notice a similarity between two situations and approach them in similar ways, like walking to a building and figuring out how to open its door based on your previous door experience.
Maybe I was mixing terminologies - I suppose affordances isn't really the same idea as resonances or tropes. But my point there was that you use whatever comes in handy from adjacent kinds of situations (disasters) in serving as a reference point and guiding your inquiry, even where this doesn't look topographically similar and you might not have predicted earlier how it would come in handy. And that this may show up as a resonance or suggestive parallel, rather than an outright structural homology. In this sense, the resonance (what you call a similarity though it needn't be formally similar, just similar enough) can serve *as* an (epistemic) affordance. For better or worse it helps structure and inform the way you go about piecing all the other evidence together post-hoc.
[I had included an example but deleted since it may not be a great example]
I think we’re on the same page. I took epistemic affordances to be what’s letting you seize the current situation by understanding one YOU see as similar. similarity can be homological or analogical, or both. affordance is a neutral word in that case because it claims neither structural nor functional isomorphism.
That makes sense. In terms of your specific image of opening the door, maybe realistically it's less about opening it than just finding your way to the right building, and being able to locate where the right door is or what would count as a door. To actually open the door takes more than analogies; you need up-to-date (or newly emerging) evidence, and the local contingencies of the event in question.
I've been struggling with the same thing you've mentioned about classifying the domain-independent structural similarities of how we should properly handle knowledge. Informally I've been using "information kinematics" to try to capture that spirit of "the science of things hitting other things in such a way where the composition of the things is tremendously important, but you can still draw general lessons if you're careful about it." But I'm still not totally sure if it's self-explanatory enough that I can just slip it in to my writing.
Anyway, I completely agree and think this is the kind of thing we all need to be broken records about. We're so conditioned to think of rare events like lottery tickets - especially when so many modern institutions only exist because of the human affinity for gambling - but the real world has tons of control processes and unexplained correlations. So a rare event is a lot more likely to be a correspondence break from the base state of the system in a way we don't understand yet, and our base assumptions should be cryptic factors over bad luck. Honestly, I think the modern crop of self-proclaimed Bayesian's have no small hand of blame in this, since the very act of setting a prior implies that you think your *available language with regard to the system is complete*. For all of the supposed humility in laying out and updating your thoughts, I think the real world often turns on distinctions between outcomes A and A' that are similar enough that we didn't even think to define a difference between them until the impact of that difference suddenly presented itself.
I think it's a form of Perrow's "Normal Accidents" but clearly nuclear power plants aren't the same as virus research, but there are these categories of events that present us with large, even catastrophic tail risk, even as the ocean they swim in are full of fairly common event. It's that coupling—to use Perrow's terminology—that kicks of the potential disaster.
I agree that modern Bayesian rhetoric is sometimes self-limiting, and yes, a lot of things are clearer in the rear-view window and there's many things for which we have no language for the difference—until we do because we have to. It reminds me of my piece on how Turks now how to distinguish between many types of coups—we don't need pedantic arguments over if something was exactly a coup if the chief of staff wasn't involved but his deputy was. Our (unfortunate) experience gives us fine-grained language.
Your point about an identification problem in Bayesian framing really resonates. But then I wonder if the pandemic reasoning problems have been about too little Bayes, via stuck priors? Rare events should yield weak priors, and new evidence (or lack thereof) should cause big updates, right? Instead, both zoonosis and large droplet transmission were protected by a super high burden of proof. But I admit that doesn't fully address the framing issue. Maybe the category congruence/isometry concept could help with framing by broadening the context.
I agree, and also the epistemic humility before an event like this should also require a spirit of inquiry—instead we've been met with "nothing to see here" amidst general hand-waving. I know we don't have control over this, but an NTSB-style "no blame" investigation would be the perfect approach to pandemics.
How much do we understand about the factors influencing the scientists who signed the Lancet letter indicating that Covid was not lab-related? I know that Peter Daszak had some pretty serious conflicts of interest (which I believe the Lancet has now admitted). I know Zeynep indicated in several different writings that a number of the original signers have now indicated that a lab-leak is a possibility. Why did the different "experts" sign-on in the first place? This also seems an area that would benefit from some rigorous examination. That letter and the tone behind it provided more fodder for people who want to dismiss science and expertise. I wonder if there are ways scientific communication can be more transparent than it currently is (signing a conflict of interest form when submitting a grant or publication is woefully insufficient).
"Looking at the post-molecular biology era, when we were more in a position to be the cause of a pandemic, the known incidence is that two out of three pandemics were completely zoonotic: the same odds that Nate Silver had assigned to Hillary Clinton winning the presidency in 2016. It’s actually not saying that much either—again, given a rare outcome—except that ruling out pathways without evidence isn’t warranted."
I'm trying to figure out which list of events is being discussed here. I'm thinking that HIV is one of the completely zoonotic pandemics, and the 1977 flu pandemic is the one that isn't obviously zoonotic. I guess SARS, MERS, Zika, and Ebola were never pandemics, and I had to go to the Wikipedia tab on pandemics and epidemics and look at their list of global ones before I remembered that 2009 swine flu was probably the third one you were mentioning. But it does seem worthwhile to have a canonical list of the events that count as "pandemic" so that people can have more systematic discussions of them.
Also, one thing that's bothered me for a while is that I haven't been able to figure out what ever happened to the Zika virus. I recall in 2016 that it was seeming like a major worry, first in South America, and then gradually in North America. But at a certain point, I stopped hearing about it. I haven't been able to figure out whether transmission suddenly stopped everywhere in early 2017 (and if so, why), or whether it turned out that Zika wasn't as relevant to microcephaly as had been thought so that we could mostly ignore it, or if I stopped hearing about it for some other reason.
I’m wondering what those of us reading along at home are supposed to do with this:
> First, going forward, we should consider all possibilities as potential pathways and think about what can be done. I people don’t understand the evidentiary basis on which people are assigning precise(ish) likelihoods to the paths leading to a rare, perhaps once-in-a-century rare, event when a lot of paths have been demonstrated as viable.
To continue your analogy, this sounds like good advice for the investigators of a plane crash, but we aren’t plane crash investigators. We aren’t going to figure this out ourselves; we lack expertise and access to the relevant evidence. At best we can be supportive of those who are in a position to investigate and wait for future developments.
Since it seems there isn’t going to be a real investigation, I’m not sure what we can do other than shrug and move on. “Sure, it could have happened that way. So what? Flying is very safe in general, so I’m still going to buy a plane ticket.”
As the meme goes, this seems like both the simple and adept way of looking at it?
"1-More about that furin cleavage site debate (it ended up being too wonky for the article but it’s actually an interesting intersection between science, evolution and academic incentives)."
I work as copy editor for a science magazine (not a journal but a sort of trade mag for biologists). I have been closely following the lab leak debate since April 2020 and was first struck by Yuri Deigin's breakdown of the virus's genome in Medium, which first pointed to the furin cleavage site. Deigin and Segreto then wrote a paper about it, which in turn pointed to this one
"'You cannot, in a normal cell culture, maintain the furin cleavage site,”'he told me. When the COVID-19 virus is replicated in a cell culture in a lab, he said, the furin cleavage tends to delete itself. A peer-reviewed paper, published in late April in Nature, noted that habit and identified seven other papers that found a similar deletion. [LINK: https://www.nature.com/articles/s41564-021-00908-w#author-information]
"So if researchers were using traditional methods and their preferred cell lines to try to force the virus to replicate, mutate, and change, the furin cleavage site would likely disappear.
"The gain-of-function proponents say this furin site is too well adapted for humans to be an accident. But Goldstein said the opposite is true. The cleavage site is imperfect, so odd, that it could have only been a freak of nature. 'No virologist would use that cleavage site,' he said."
LONG STORY SHORT, I proposed to my science magazine employers that they host a debate between, say, Goldstein and Deigin, Segreto, or another proponent of considering a gain-of-function origin. They declined: they seem to find Deigin and Segreto a little shady (and in any event, neither is a virologist).
Maybe you could host that debate. In the absence of hard evidence, it really does come down to scientists' informed opinions on the question of how SARS-CoV-2 sprang into the world so seemingly pre-armed to cause a human pandemic.
Zeynep, this is even above and beyond your normal (casual and profound) connection: "Looking at the post-molecular biology era, when we were more in a position to be the cause of a pandemic, the known incidence is that two out of three pandemics were completely zoonotic: the same odds that Nate Silver had assigned to Hillary Clinton winning the presidency in 2016." Bam.
I see some openings for further understanding, even lacking participation by China. For example, do people eat horseshoe bats? Do they house them nearby, perhaps to control insects? If not, that eliminates whole areas of inquiry and lets us focus on others.
Regarding Biden calling on intelligence agencies to produce a report: I took this to suggest a belief that we have more information than we have surfaced in the information stream. (Because finding stuff in information streams is a big part of what intelligence agencies do. And sometimes fail to do...) Recently we have seen news reports of the content of previously overlooked messages of various sorts, and the great sleuthing that noted the deletion of the sequences from the database and the finding of a number of them once researchers knew that they existed. It became like finding the car keys, not having to hypothesize the existence of the car.
Being more of an interested spectator than a stakeholder in identify with clarity the source of this global pandemic, I will follow your endeavors. Unfortunately, China's actions seem not to want or publicly identify your goal, and the desire of the world health community to solve this mystery with sufficient data. Assuming the origin is from the wet market, and China collected all the animals at this market, it is strange that they say they destroyed these animals rather than saving them from finding the source and bringing more clarity to this vector.
My concern and memory are that with the Spanish Flue, the second year caused a more significant loss of life, maybe through a more deadly variation of the flue. However, since the world will be the incubator for Covid in the next few years, and examples like the Delta variation, what are the rare odds that a new variation dominates but does not kill or kill millions?
In comparing airplane crashes and the efforts to identify the details around why this crash event happened. The goal being so that new actions are taken to either better train pilots, repair a broken part, create another safety break when the systemic system of safety has a rare flaw is a perfect analogy. I seem to recall a French plane crash in the South Atlantic that took years and millions of dollars to find the plane, concluding it was a pilot error. A few years ago, an Indonesian plane disappeared, and radar and satellite data showed it crashing somewhere in the Indian Ocean to the west of Australia. This plane and mystery may never be solved unless its actual location, recovery happens.
Today I read this article that points out an outsider's insider perspective on the professional standards at the viral lab and media rumors that would challenge her experiences.
Bloomberg article "The Last-And Only-Foreign Scientist in the Wuhan Lab Speaks Out."
"Virologist Danielle Anderson says half-truths have obscured an accurate accounting of Wuhan lab's functions, but she does think an investigation is needed to nail down virus's origin."
"Once a plane crash does occur, we don’t tell the investigators that planes are really really safe, so let’s not worry about it. In fact, the thoroughness of that investigation is exactly what makes planes really really safe, in the future: we figure it out and fix it."
I'm not sure the airplane safety comparison works here- airplanes are human-designed and human-built machines, under human control, and humans have the ability to fully reconstruct a failure sequence and directly address it. This is even down to the level of airliner CPUs being hardened against extremely rare but known failure modes- comic rays causing a digital bit somewhere to flip from 0 to 1.
In contrast, it can very difficult to understand and manipulate the biological world, including viruses, to the same level of precise causality. For the original SARS, the best we have found more than a decade later is a cave where bats have different viruses that contain all of the constituent parts of the virus.
Additionally, this argument almost sounds like great justification for pursuing gain-of-function research on coronaviruses after the 2003 SARS outbreak!
But labs have similar guards against failure modes—and may well need more. Equipment failure is not an uncommon cause, and it can be stochastic like you describe. The problem is, of course, every additional layer adds expense and burden and the scientific community — at least the many I talked to — were not fans of that. (As an academic, I understand! I feel the same way about my own research bureaucracy, and some of the rules do seem performative. On the other hand, we're all interested parties to what makes our life a little harder and our research slower).
I’ve been trying to find a word for this? Structural isomorphism? Methodological similarity? I’m very open to suggestions! How about CONCEPTUALLY ANALOGOUS?
Epistemology: The theory of knowledge, especially with regard to its methods, validity, and scope. Epistemology is the investigation of what distinguishes justified belief from opinion.
“there are a lot of meta-features of different kinds of events that create conceptual tool-kits that can be brought over, even though on the surface, the event is a different category. I’ve been trying to find a word for this? Structural isomorphism? Methodological similarity? I’m very open to suggestions!”
Structural isomorphism is too restrictive. Think of oat milk. It’s not cow milk (or goat or sheep) but many people want them to behave similarly. When they don’t behave in the same way, people try to find work arounds. That makes it functional or behavioral isomorphism, if isomorphism itself is the right word.
Methodological similarity is similarly too restrictive, since this is behavior that people exhibit under everyday circumstances, not just academic research. Again, oat milk.
It’s more a form of analogy: using an example you know well as an analog for what you don’t know. It can often get you into the new example because there are similarities, or can yield incredible insights because of how the two differ,
Well, it needs a name! (I tell the linguist. :-D)
$0.02... whether 'analogy' is a better fit than, say, 'homology' depends on whether you want to emphasize the disjointness of the subject matter domains, or the behavioral toolkits we execute over them, which have a lot in common and a familial relationship. I tend to think the latter is key here, but everyone's mileage may vary...
I see your bid and I raise you to $0.03. An avian wing is homologous to a human arm. A bee’s wing or an airplane wing are analogous to a bird’s wing. And then there are bats, whose front legs are homologous to human arms, and whose wings are analogous to birds’ wings.
Clearly the word “similarity” is too vague. We can solve that through obfuscation by using the homologous Greek word “homoiotis.” Sounds technical and fancy that way.
Imho "solve ... through obfuscation" is worth at least a buck :) But the arm-wing-leg thing is exactly on point. If we'd like to leverage what we know about plane crash analysis (or whatever) to optimize pandemic response (or whatever), and we're looking at some aspect, let's say stakeholder withholding of information (whatever), then drilling into whether the similarity is deeply rooted or merely convergent is a key part of what we're talking about. I think.
How do you distinguish deeply rooted from convergent? In evolution you could say genetic common ancestor, but here neither can share an ancestor.
I’ve run through a number of scenarios from the many disciplines I have any experience in and have found the pattern seems to be that “deeply rooted” correlates to “structural” and “convergent” to “functional” or “situational.”
Rereading the whole conversation, however, I now see something I hadn’t seen yesterday. You’re addressing issues about the architecture of the problem space: How can we look at this problem in a way that we can make some headway to understanding? This strikes me as having a lot of shared ground with the idea of design patterns, used originally in architecture by Christopher Alexander, and then extended (analogously) to software development. It’s the idea of approaching a situation in terms of a working design, not starting from scratch every time. The level of abstraction is approximately the same as you’re striving for.
Yeah I thought about "deep structure" (you know from which field. :-D). But that's exactly it: the architecture of the problem space is isomorphic or analogous, despite really different areas.
This conversation is a fair way above my head, but I'm wondering if the situation where a doctor was watching a NASCAR race and wondered how it was that the team managed minor repairs so quickly. The doctor learned that the team keeps complete kits of common repairs within easy reach at all times, cutting down on response time. Said doctor then applied this thinking to surgeries, and it's now common practice for kits to address many common medical emergencies to be easily accessible within ER units (or maybe OR rooms, I can't remember), which leads to lower mortality rates because common emergencies can be addressed more quickly. Because of car racing.
Taking a look at Wikipedia's disambiguation page for 'homology' (https://en.wikipedia.org/wiki/Homology), I see a group that includes Homology (anthropology), Homology (psychology), and Homology (sociology) that all revolve around the 'deeply-rooted similarity' notion.
I guess my speculation is that a Homology (technosociology) could someday capture a part of the idea you're developing about the way we do or fail to make and act on the same learnings across different subject matter domains (different methods for rare events being a perfect example). The hypothesized 'deep roots' would be in our human nature, existing cultures, and - especially, given (what I perceive to be) your favorite angles of attack - pervasive patterns of use and misuse of information at scale under the pressure of rapidly advancing technology.
Of course, even if you buy this, it's only one of many things you could choose to emphasize in your choice of terminology. If it doesn't resonate with you, that's good enough for me!
Yes. It provides a starting point, which may fall apart as you go deeper, or may bear fruit (or viruses), but at least it’s not a blank sheet of paper.
It was a little unclear from the original formulation whether the emphasis is problematic scenarios in which a conceptual toolkit is inappropriately brought over and misapplied (she spoke of "conflating" terrorism with structural ethnic tensions which would imply a fallacy), or the more neutral observation that these meta-principles can be usefully applied in other situations that look different on the surface. That might have implications for the terminology.
Putting that aside: if the point of interest is behavior and events that originally led to the outcome, I feel like the "functional" vs. "topographic/topological" distinction works best. If the focal point is the forensic task of inferring post-hoc what might have happened (reasoning and methodology, etc.), then maybe a more figurative, analogy-adjacent term that evokes the idea of "carrying over" is appropriate. How about the term "resonances" or "tropes"? Or "epistemic affordances?"
Meta-principles! There is a structural similarity: rare event, background event not uncommon but requires a chain of unlucky incidents to come together, can have catastrophic/terrible consequences. Different than anything, say, on a normal distribution of occurrence frequency .Tail risk, but the process leading to the tail risk is also specific and shared.
Oh I see. That's helpful, thanks. I had a couple other thoughts on this subject, that hopefully are not too late to share here (totally fine if no one jumps on them). One is that I wonder if there's a difference between drawing upon adjacent or analogous kinds of situations (like the plane crash example) that bear hidden structural similarities, and drawing on historical knowledge and accumulated hindsight about related kinds of events in the past, which echo or rhyme in this new setting. In that sense it wouldn't be the horizontal or vertical dimensions of knowledge that we've discussed here before, but a more temporal-historical, geneological dimension. The reason I was attracted to the term "resonances" or "tropes," or the notion of "evoking," is that they speak to this more historical - even humanities - perspective.
Another stray thought is that while we have been relying on other big disasters for insight, perhaps there are more mundane examples of daily personal mishaps or micro-"tragedies" that point toward useful analogues as well, which would function at larger scale in disasters. Take the film 127 Hours, the James Franco one about the hiker in Utah who was unlucky enough to fall into a crevice in such a way that a boulder was dislodged that happened to pin his arm in exactly the wrong-sized space (forcing him to cut it off if he hoped to survive). This would appear on the surface to just be a case of bad luck. Yet it feels vaguely reminiscent of (hopefully less dramatic) "disasters" we've all found ourselves in where not just one thing goes wrong, but three things at the same time that somehow work to reinforce one another. Is this merely the inevitable statistical result of many, many repeated trials from which an eventual tail event was inevitable, which would have occurred some other way eventually? Or, does it reflect something more interesting, the interesting features of which make it ultimately preventable?
You’ve added value judgement needlessly to your long tail. Good results can come together as well in the same way. The business cliché in vogue a couple of years ago was “perfect storm,” or “planetary alignment.”
That speaks to another question, of whether the dynamics of tail events we would assess as "positive" function any differently from ones we are inclined to experience as "negative" - or whether that's purely reflective of our value judgments and social context and nothing to do with underlying structure and form. Perhaps it depends on whether the "terrible" scenario is terrible due to entropy, or due to a really terrible negentropy?
I’m unclear what could be an inherently positive or negative result without a human observer to ascribe value.
epistemic affordances … I assume Gibson-type affordances … is interesting to try to visualize.
so, in essence, you (the speaker) notice a similarity between two situations and approach them in similar ways, like walking to a building and figuring out how to open its door based on your previous door experience.
Maybe I was mixing terminologies - I suppose affordances isn't really the same idea as resonances or tropes. But my point there was that you use whatever comes in handy from adjacent kinds of situations (disasters) in serving as a reference point and guiding your inquiry, even where this doesn't look topographically similar and you might not have predicted earlier how it would come in handy. And that this may show up as a resonance or suggestive parallel, rather than an outright structural homology. In this sense, the resonance (what you call a similarity though it needn't be formally similar, just similar enough) can serve *as* an (epistemic) affordance. For better or worse it helps structure and inform the way you go about piecing all the other evidence together post-hoc.
[I had included an example but deleted since it may not be a great example]
I think we’re on the same page. I took epistemic affordances to be what’s letting you seize the current situation by understanding one YOU see as similar. similarity can be homological or analogical, or both. affordance is a neutral word in that case because it claims neither structural nor functional isomorphism.
That makes sense. In terms of your specific image of opening the door, maybe realistically it's less about opening it than just finding your way to the right building, and being able to locate where the right door is or what would count as a door. To actually open the door takes more than analogies; you need up-to-date (or newly emerging) evidence, and the local contingencies of the event in question.
But this is all getting very....metaphorical!
I've been struggling with the same thing you've mentioned about classifying the domain-independent structural similarities of how we should properly handle knowledge. Informally I've been using "information kinematics" to try to capture that spirit of "the science of things hitting other things in such a way where the composition of the things is tremendously important, but you can still draw general lessons if you're careful about it." But I'm still not totally sure if it's self-explanatory enough that I can just slip it in to my writing.
Anyway, I completely agree and think this is the kind of thing we all need to be broken records about. We're so conditioned to think of rare events like lottery tickets - especially when so many modern institutions only exist because of the human affinity for gambling - but the real world has tons of control processes and unexplained correlations. So a rare event is a lot more likely to be a correspondence break from the base state of the system in a way we don't understand yet, and our base assumptions should be cryptic factors over bad luck. Honestly, I think the modern crop of self-proclaimed Bayesian's have no small hand of blame in this, since the very act of setting a prior implies that you think your *available language with regard to the system is complete*. For all of the supposed humility in laying out and updating your thoughts, I think the real world often turns on distinctions between outcomes A and A' that are similar enough that we didn't even think to define a difference between them until the impact of that difference suddenly presented itself.
I think it's a form of Perrow's "Normal Accidents" but clearly nuclear power plants aren't the same as virus research, but there are these categories of events that present us with large, even catastrophic tail risk, even as the ocean they swim in are full of fairly common event. It's that coupling—to use Perrow's terminology—that kicks of the potential disaster.
I agree that modern Bayesian rhetoric is sometimes self-limiting, and yes, a lot of things are clearer in the rear-view window and there's many things for which we have no language for the difference—until we do because we have to. It reminds me of my piece on how Turks now how to distinguish between many types of coups—we don't need pedantic arguments over if something was exactly a coup if the chief of staff wasn't involved but his deputy was. Our (unfortunate) experience gives us fine-grained language.
Your point about an identification problem in Bayesian framing really resonates. But then I wonder if the pandemic reasoning problems have been about too little Bayes, via stuck priors? Rare events should yield weak priors, and new evidence (or lack thereof) should cause big updates, right? Instead, both zoonosis and large droplet transmission were protected by a super high burden of proof. But I admit that doesn't fully address the framing issue. Maybe the category congruence/isometry concept could help with framing by broadening the context.
I agree, and also the epistemic humility before an event like this should also require a spirit of inquiry—instead we've been met with "nothing to see here" amidst general hand-waving. I know we don't have control over this, but an NTSB-style "no blame" investigation would be the perfect approach to pandemics.
How much do we understand about the factors influencing the scientists who signed the Lancet letter indicating that Covid was not lab-related? I know that Peter Daszak had some pretty serious conflicts of interest (which I believe the Lancet has now admitted). I know Zeynep indicated in several different writings that a number of the original signers have now indicated that a lab-leak is a possibility. Why did the different "experts" sign-on in the first place? This also seems an area that would benefit from some rigorous examination. That letter and the tone behind it provided more fodder for people who want to dismiss science and expertise. I wonder if there are ways scientific communication can be more transparent than it currently is (signing a conflict of interest form when submitting a grant or publication is woefully insufficient).
"Looking at the post-molecular biology era, when we were more in a position to be the cause of a pandemic, the known incidence is that two out of three pandemics were completely zoonotic: the same odds that Nate Silver had assigned to Hillary Clinton winning the presidency in 2016. It’s actually not saying that much either—again, given a rare outcome—except that ruling out pathways without evidence isn’t warranted."
I'm trying to figure out which list of events is being discussed here. I'm thinking that HIV is one of the completely zoonotic pandemics, and the 1977 flu pandemic is the one that isn't obviously zoonotic. I guess SARS, MERS, Zika, and Ebola were never pandemics, and I had to go to the Wikipedia tab on pandemics and epidemics and look at their list of global ones before I remembered that 2009 swine flu was probably the third one you were mentioning. But it does seem worthwhile to have a canonical list of the events that count as "pandemic" so that people can have more systematic discussions of them.
Yeah, the 2009 H1N1 is the third one. And it's a fascinating case: it was indeed traced to central Mexico.
Also, one thing that's bothered me for a while is that I haven't been able to figure out what ever happened to the Zika virus. I recall in 2016 that it was seeming like a major worry, first in South America, and then gradually in North America. But at a certain point, I stopped hearing about it. I haven't been able to figure out whether transmission suddenly stopped everywhere in early 2017 (and if so, why), or whether it turned out that Zika wasn't as relevant to microcephaly as had been thought so that we could mostly ignore it, or if I stopped hearing about it for some other reason.
Good question!
I’m wondering what those of us reading along at home are supposed to do with this:
> First, going forward, we should consider all possibilities as potential pathways and think about what can be done. I people don’t understand the evidentiary basis on which people are assigning precise(ish) likelihoods to the paths leading to a rare, perhaps once-in-a-century rare, event when a lot of paths have been demonstrated as viable.
To continue your analogy, this sounds like good advice for the investigators of a plane crash, but we aren’t plane crash investigators. We aren’t going to figure this out ourselves; we lack expertise and access to the relevant evidence. At best we can be supportive of those who are in a position to investigate and wait for future developments.
Since it seems there isn’t going to be a real investigation, I’m not sure what we can do other than shrug and move on. “Sure, it could have happened that way. So what? Flying is very safe in general, so I’m still going to buy a plane ticket.”
As the meme goes, this seems like both the simple and adept way of looking at it?
THIS question. A suggestion follows.
"1-More about that furin cleavage site debate (it ended up being too wonky for the article but it’s actually an interesting intersection between science, evolution and academic incentives)."
I work as copy editor for a science magazine (not a journal but a sort of trade mag for biologists). I have been closely following the lab leak debate since April 2020 and was first struck by Yuri Deigin's breakdown of the virus's genome in Medium, which first pointed to the furin cleavage site. Deigin and Segreto then wrote a paper about it, which in turn pointed to this one
This article in Foreign Policy magazine
https://foreignpolicy.com/2021/06/15/lab-leak-theory-doesnt-hold-up-covid-china/?fbclid=IwAR0xkHA4DbSgZtGH9VrunZlrwFAYwuP9lsl1eSAGIQChGGm1LH6-E3GHELk
is mostly geopolitical pooh-poohing, but at the end cites some scientists and papers arguing that the furin cleavage site is more likely natural, acquired by recombination. Stephen Goldstein of the University of Utah, a human geneticist, is one, https://cellvolution.org/stephen-goldstein.htmlhttps://cellvolution.org/stephen-goldstein.html
"'You cannot, in a normal cell culture, maintain the furin cleavage site,”'he told me. When the COVID-19 virus is replicated in a cell culture in a lab, he said, the furin cleavage tends to delete itself. A peer-reviewed paper, published in late April in Nature, noted that habit and identified seven other papers that found a similar deletion. [LINK: https://www.nature.com/articles/s41564-021-00908-w#author-information]
"So if researchers were using traditional methods and their preferred cell lines to try to force the virus to replicate, mutate, and change, the furin cleavage site would likely disappear.
"The gain-of-function proponents say this furin site is too well adapted for humans to be an accident. But Goldstein said the opposite is true. The cleavage site is imperfect, so odd, that it could have only been a freak of nature. 'No virologist would use that cleavage site,' he said."
LONG STORY SHORT, I proposed to my science magazine employers that they host a debate between, say, Goldstein and Deigin, Segreto, or another proponent of considering a gain-of-function origin. They declined: they seem to find Deigin and Segreto a little shady (and in any event, neither is a virologist).
Maybe you could host that debate. In the absence of hard evidence, it really does come down to scientists' informed opinions on the question of how SARS-CoV-2 sprang into the world so seemingly pre-armed to cause a human pandemic.
Sorry, I failed to put in the link to Deigin and Segreto's article https://onlinelibrary.wiley.com/doi/full/10.1002/bies.202000240
and to this one that they point to. https://onlinelibrary.wiley.com/doi/10.1002/jmv.26478
Zeynep, this is even above and beyond your normal (casual and profound) connection: "Looking at the post-molecular biology era, when we were more in a position to be the cause of a pandemic, the known incidence is that two out of three pandemics were completely zoonotic: the same odds that Nate Silver had assigned to Hillary Clinton winning the presidency in 2016." Bam.
More thoughts on this piece: "Virologists still largely lean toward the theory that infected animals — perhaps a bat, or another animal raised for food — spread the virus to humans outside of a lab." (https://www.nytimes.com/2021/05/27/health/wuhan-coronavirus-lab-leak.html)
I see some openings for further understanding, even lacking participation by China. For example, do people eat horseshoe bats? Do they house them nearby, perhaps to control insects? If not, that eliminates whole areas of inquiry and lets us focus on others.
Regarding Biden calling on intelligence agencies to produce a report: I took this to suggest a belief that we have more information than we have surfaced in the information stream. (Because finding stuff in information streams is a big part of what intelligence agencies do. And sometimes fail to do...) Recently we have seen news reports of the content of previously overlooked messages of various sorts, and the great sleuthing that noted the deletion of the sequences from the database and the finding of a number of them once researchers knew that they existed. It became like finding the car keys, not having to hypothesize the existence of the car.
Forensic homology?
Analytic, epistemic, or [the M word]-ical homology? (Not sure how publicly you're using the M word at this point :)
I am! I like the M word a lot. :-D
Being more of an interested spectator than a stakeholder in identify with clarity the source of this global pandemic, I will follow your endeavors. Unfortunately, China's actions seem not to want or publicly identify your goal, and the desire of the world health community to solve this mystery with sufficient data. Assuming the origin is from the wet market, and China collected all the animals at this market, it is strange that they say they destroyed these animals rather than saving them from finding the source and bringing more clarity to this vector.
My concern and memory are that with the Spanish Flue, the second year caused a more significant loss of life, maybe through a more deadly variation of the flue. However, since the world will be the incubator for Covid in the next few years, and examples like the Delta variation, what are the rare odds that a new variation dominates but does not kill or kill millions?
In comparing airplane crashes and the efforts to identify the details around why this crash event happened. The goal being so that new actions are taken to either better train pilots, repair a broken part, create another safety break when the systemic system of safety has a rare flaw is a perfect analogy. I seem to recall a French plane crash in the South Atlantic that took years and millions of dollars to find the plane, concluding it was a pilot error. A few years ago, an Indonesian plane disappeared, and radar and satellite data showed it crashing somewhere in the Indian Ocean to the west of Australia. This plane and mystery may never be solved unless its actual location, recovery happens.
Today I read this article that points out an outsider's insider perspective on the professional standards at the viral lab and media rumors that would challenge her experiences.
Bloomberg article "The Last-And Only-Foreign Scientist in the Wuhan Lab Speaks Out."
"Virologist Danielle Anderson says half-truths have obscured an accurate accounting of Wuhan lab's functions, but she does think an investigation is needed to nail down virus's origin."
"Once a plane crash does occur, we don’t tell the investigators that planes are really really safe, so let’s not worry about it. In fact, the thoroughness of that investigation is exactly what makes planes really really safe, in the future: we figure it out and fix it."
I'm not sure the airplane safety comparison works here- airplanes are human-designed and human-built machines, under human control, and humans have the ability to fully reconstruct a failure sequence and directly address it. This is even down to the level of airliner CPUs being hardened against extremely rare but known failure modes- comic rays causing a digital bit somewhere to flip from 0 to 1.
In contrast, it can very difficult to understand and manipulate the biological world, including viruses, to the same level of precise causality. For the original SARS, the best we have found more than a decade later is a cave where bats have different viruses that contain all of the constituent parts of the virus.
Additionally, this argument almost sounds like great justification for pursuing gain-of-function research on coronaviruses after the 2003 SARS outbreak!
But labs have similar guards against failure modes—and may well need more. Equipment failure is not an uncommon cause, and it can be stochastic like you describe. The problem is, of course, every additional layer adds expense and burden and the scientific community — at least the many I talked to — were not fans of that. (As an academic, I understand! I feel the same way about my own research bureaucracy, and some of the rules do seem performative. On the other hand, we're all interested parties to what makes our life a little harder and our research slower).