Hacker News Reader

Comment by renewiltord | Hacker News Reader

There was a lot of breathless nonsense from a bunch of people about this disease. I just Bayes low-weighted it because the prior probability is low. So far a whole bunch of these things are coming back non-replicable. It'll be an interesting test of each of our epistemological techniques.

My prior to start with is high that the contagiousness-increasing mutation (the novel spike protein) improves likelihood of the virus attacking more organs and improves its ability to massively infect the organ but that once you've got a massive infection of the organ, you should see effects after recovery that are similar to any other such infection of the organ.

I'm actually curious if anyone has neat software that tracks one's predictions and allows one to adjust their knowledge of the world systematically. Currently only using a shared spreadsheet with friends.

View on HN

I can't think of a past novel disease that didn't stir up fearful misinformation. Yet I also remind you that every disease has a different pathology. It is worthwhile to report anecdotal clinical presentations because that reporting spurs research. The research can then corroborate or controvert the reports.

A lot of what you suggested is kind of silly. Contagiousness is not the same thing as virulence or pathogenicity, and you can look at things like the common cold (contagious but mild) versus ebola (harder to transmit but severe) to disprove the proposition in your second paragraph.

The fact of the matter is that some people have not recovered after a long time ("long covid") and that clinical observation needs to be researched and explained before we brush off concerns about organ damage as "breathless nonsense." Some suggestions may be wrong but that doesn't mean their study is not useful. Every time we disprove a concern, we come closer to a full understanding of a disease that has affected people all over the globe.

> A lot of what you suggested is kind of silly. Contagiousness is not the same thing as virulence or pathogenicity, and you can look at things like the common cold (contagious but mild) versus ebola (harder to transmit but severe) to disprove the proposition in your second paragraph.

This is actually related to the stuff I'm referring to in my cousin comment. Existence proofs are not evidence of certain behaviour. An analogous situation might be "If you see a car slow down approaching a pedestrian crossing with a Stop sign it is likely to stop". Most pedestrians will consider this statement true (as evinced by how they behave - placing themselves in range of bodily harm). However, many pedestrians know that at least one car has not behaved in this manner. Responding to someone who agrees with that statement with "Cars certainly do blow through stop signs even if they're slowing down. Here are some news articles of that happening" only marginally moves the evidence-o-meter.

Some things need empirical understanding. "Related counter-examples" do not set the likelihood of the statement to zero, but only downgrade it somewhat.

Personally, I've found my performance on COVID predictions pretty decent. My biggest failure was that I believed outdoor masking was necessary, and it currently looks like that is not the case.

We shall see if Long COVID is any sort of true long-term disease that is particularly unique or merely the identification of other symptoms that previously went untested simply because most illnesses go untested.

And of course I have no problem with anyone sounding small alarms on small studies. A sufficient number of independent such instances is how we identify common patterns. I, obviously, weight each individual such instance low but corroboration increases likelihood. Still, we'll see. Currently, my method of acquiring knowledge outperforms public CDC statements about the disease¹, so I'm at least better at this than they are.

¹ Assuming, perhaps naïvely, that they prioritize dissemination of truth over public health objectives that may temporarily be better met through carefully told untruths.

Humans aren't ideal Bayesian rational agents, and even they are susceptible to selective reporting. I don't think this comment is useful.

Oh we most certainly aren't. And I frequently make errors as well, but I aspire to be better at this than I am and I think I'd enjoy the company of others like this.

You could try the https://lesswrong.com/ community. Most of them aren't like that at all, but enough are that you basically embody the stereotype.

LW is definitely along that path. Thank you for sharing.

Bayesian reasoning has its limits.

> "The man who has fed the chicken every day throughout its life at last wrings its neck instead, showing that more refined views as to the uniformity of nature would have been useful to the chicken" - Bertrand Russel

For sure. The media tends to report science press releases fairly uncritically. The result is that science reporting is reporting of claims of scientists and not of "science" itself. A better media ecosystem would incorporate the fact that many preliminary findings never pan out.

I'm quite curious if there is a group of like-minded individuals out there who would participate in a wiki that attempts to summarize (with degree of certainty) the current known knowledge.

The general Internet has a tendency to consider individual papers authoritative summaries of the general state of a field. And single studies rarely have that predictive power. For instance, one might read a single paper about a single study that says "Doubling of rectal cancer risk occurs if red meat is consumed right after sugar intake", and most Internet commentary will attempt to show this as incontrovertible truth. However, few such studies actually last or survive replication.

i.e. most Internet commentary over-weights single study results, with near binary probability of truth measures: either something is almost certain or almost never.

Would prefer a place where I can discuss with people who are less stridently enthusiastic about these papers, and where long membership retains historical prediction performance (I don't mind people disregarding my future predictions if my current predictions are poor) with appropriate windowing to permit learning.

  > I'm quite curious if there is a group of like-minded individuals out there who would participate in a wiki that attempts to summarize (with degree of certainty) the current known knowledge.

Yes, it can be found at a http://wikipedia.org

I'm actually a long-term contributor to Wikipedia (in my 14th year) and, as you can imagine, therefore a huge fan. However, Wikipedia does not attempt a synthesis of knowledge, preferring to be a repeater of what is publicly currently believed to be knowledge.

i.e. if CNN, the WSJ, and the NYT report X is Y, that will count a lot more about X being Y than X actually being Y.

Wikipedia is a wonderful resource and it definitely partly scratches this itch, and I appreciate your sharing it, but I am looking for something that occupies a slightly different space.

I have never measured the speed of light myself. But Google and Wolfram Alpha agree with my memory of physics lectures and textbooks.

All of our epistemology is based on testimony. And in the degenerate but common case where the credibility of the source isn't considered, our epistemology devolves into memes.

If Wikipedians apply consistent effort towards reconciling contradictory information in favor of seeking out better* sources, then that makes it as robust a system as we know how to make.

*I don't know how to define "better" or "credible"

It may be based on testimony but this does not mean there is no metric for quality. There are tools which can help hint to us about the quality of some information: double blind controlled trials, reproduction, systemic reviews, longitudinal studies, published pre-study methodology, multiple lines of evidence. And some are considered as better than others. For example, a double blind controlled experiment is considered more indicitive of causal effect than an observational study. Which is still better than a self reported survey.

I don't think it's entirely a bad idea. It would be hard because it would be an elitist endeavor. As in, criteria and protocol would have to be established by a small group ahead of time rather than democratically. Such institutions are vulnerable to capture or stagnation. But they can be really useful while they are true to their mission.

In my experience, Wikipedia covers emerging phenomena incompletely.

It’s also bad at a question like “Is coffee bad for me?”. I’m not saying Wikipedia is good. It’s just not what I’m looking for here.

Probably not quite what you're looking for but there was an interesting social network where the posts were forced into a kind of logical structure. For example, if you posted a conjecture you had to list supporting points that had to be individual. Then if someone wanted to reply to any point, they had to address that point specifically. I can't remember it's name and am totally unsure of how successful it was at staying on principle. But it was an interesting idea.

Is it Kialo: https://www.kialo.com/

As I recall, the media applied plenty of scrutiny to any claim that Covid might be less deadly than the media coverage suggested, the breathless uncritical reporting lent almost entirely in one direction.

Important to clarify that much of scrutiny went in the other direction, cutting down concerns about how bad things could get in January and February of 2020.

After that I still saw a pattern where the media kept proffering worst case scenarios at every step, with no studies to back them up, even contradicting known science.

The most recent example of this is fear-mongering headlines indicating that the vaccines would not be effective against variants. This might be expected behaviour for the flu, but not a coronavirus.

You misunderstand me. I'm not referring to the virulence nor the deadliness of Covid-19. More to the tangential "scientific discoveries" that were credulously reported as they were published by doctors with little review or time to coroborate. Some examples

-claims about hydroxychloroquine

-claims about various other drug cocktails

-claims of a new covid related inflammatory disease in children

-claims about transmissibility to animals from humans

Just to name a few.

> claims about transmissibility to animals from humans

I think that ones true, no? Ferrets, big cats, house cats , dogs, bats and mink (with hundreds of mink to human spread in Denmark) seem to be able to get it. Probably a bunch of primates too beyond gorillas.

But not chickens and pigs (in the small study I dug up).

https://www.thelancet.com/journals/lanmic/article/PIIS2666-5...

https://www.cdc.gov/coronavirus/2019-ncov/daily-life-coping/...

This is all a big deal since it means that, even if we vaccinate every human, it might stick around in wildlife and re-infect humans again after it mutates or as immunity decays.

It's not about whether it's true or not. It's about whether the media could know it was true in the time they reported it. Which they could not because they reported it before any study could be done.

Reply on news.ycombinator.com