Could you elaborate on the distribution of opinions you observe at these exercises? Do the opinions get written up? Is MD’s opinion within the 90th percentile of pessimism?
From our internal report (IM me for additional details)
Risk of a severe natural pandemic:
Likelihood estimate ~ 1⁄55 years [1]
Impact estimate ~ > 10 000 domestic fatalities, severe disruption of social services
Confidence grade—D (Low confidence in the judgement based on a relatively small body of knowledge, relatively small quantity and quality of relevant data and somewhat consistent relevant assessments).
[1] There were several things I thought were sub-optimal about this process, not least of which was the confusion between frequentist and Bayesian probabilities; the latter were what was really being elicited, but expressed as the former. This led to confusion when we were asked to estimate the probability of events with no historical precedent.
If you don’t believe that the world is rapidly changing, then 1⁄55 years seems fairly summarized by “when it happens.”
“10 000 domestic fatalities” and “severe disruption of social services” seem like a weird pair. The latter sounds much more severe than the former. Of course, a disease that infects many but kills few could accomplish both.
If this is the most severe thing you put odds on, it’s quite far from the 1% fatality rate, let alone the 20-30% MD talks about; he is probably way beyond the 90th percentile of pessimism. Is it possible that he is mixing up infection with mortality?
“10 000 domestic fatalities” and “severe disruption of social services” seem like a weird pair. The latter sounds much more severe than the former.
See “9/11, immediate fatalities” and “9/11, consequences” for comparison. I can easily see huge indirect impact due to panic and the like.
[edited to add:] Also: I wouldn’t be surprised to see a graph for the distribution of “disease fatalities” that has most of the mass around 10k, perhaps with a long, low tail, but the graph of “social disruption vs. fatalities” rising very sharply before 10k, but then growing only slowly.
Yes, in a catastrophe localized to a city 10k fatalities pairs sensibly with “severe disruption of social services,” but we’re talking about a pandemic.
10k fatalities, 100k gravely ill but not dying, the media confuses the two, politicians try to push lower numbers, real numbers are discovered and as a result much higher numbers are extrapolated, folk without the disease but similar or imagined symptoms overwhelm the hospitals, large scale quarantine (appropriate or not) tie up qualified personnel, actual or imaginary paucity of vaccines causes a few riots...
There’s lots of stuff that can get out of proportion. And anyway, “severe disruption of social services” is kind of vague. I mean, it sounds bad, but that might be misleading. For instance, the phrase as given does not say “country-wide”.
Could you elaborate on the distribution of opinions you observe at these exercises? Do the opinions get written up? Is MD’s opinion within the 90th percentile of pessimism?
From our internal report (IM me for additional details)
Risk of a severe natural pandemic:
Likelihood estimate ~ 1⁄55 years [1]
Impact estimate ~ > 10 000 domestic fatalities, severe disruption of social services
Confidence grade—D (Low confidence in the judgement based on a relatively small body of knowledge, relatively small quantity and quality of relevant data and somewhat consistent relevant assessments).
[1] There were several things I thought were sub-optimal about this process, not least of which was the confusion between frequentist and Bayesian probabilities; the latter were what was really being elicited, but expressed as the former. This led to confusion when we were asked to estimate the probability of events with no historical precedent.
Thanks.
If you don’t believe that the world is rapidly changing, then 1⁄55 years seems fairly summarized by “when it happens.”
“10 000 domestic fatalities” and “severe disruption of social services” seem like a weird pair. The latter sounds much more severe than the former. Of course, a disease that infects many but kills few could accomplish both.
If this is the most severe thing you put odds on, it’s quite far from the 1% fatality rate, let alone the 20-30% MD talks about; he is probably way beyond the 90th percentile of pessimism. Is it possible that he is mixing up infection with mortality?
See “9/11, immediate fatalities” and “9/11, consequences” for comparison. I can easily see huge indirect impact due to panic and the like.
[edited to add:] Also: I wouldn’t be surprised to see a graph for the distribution of “disease fatalities” that has most of the mass around 10k, perhaps with a long, low tail, but the graph of “social disruption vs. fatalities” rising very sharply before 10k, but then growing only slowly.
Yes, in a catastrophe localized to a city 10k fatalities pairs sensibly with “severe disruption of social services,” but we’re talking about a pandemic.
10k fatalities, 100k gravely ill but not dying, the media confuses the two, politicians try to push lower numbers, real numbers are discovered and as a result much higher numbers are extrapolated, folk without the disease but similar or imagined symptoms overwhelm the hospitals, large scale quarantine (appropriate or not) tie up qualified personnel, actual or imaginary paucity of vaccines causes a few riots...
There’s lots of stuff that can get out of proportion. And anyway, “severe disruption of social services” is kind of vague. I mean, it sounds bad, but that might be misleading. For instance, the phrase as given does not say “country-wide”.
I’ll dig this up when I’m back in the office Thursday, I have at least one report handy.