I honestly don't understand why this headline is at all counterintuitive.

People are currently unhappy because of the pandemic, and people can point at that and understand that there's an exogenous cause right in all of our faces causing the problem and everyone is affected.

People are much more likely to commit suicide when they just believe that everyone else is doing fine and they're the only ones whose lives are shit.

That's a good point. There is evidence that suicide rates sometimes decline during times of external crises, e.g. war.

It’s not sloppy reasoning. We should update our beliefs in light of new evidence in proportion to the strength of the evidence and the strength of our prior belief. How much would an equivalent study on the existence of ESP move your belief on that subject? A good answer would be not much if at all because a priori we expect the study being wrong is far more likely than the existence of ESP. For me, it would take more than an observational study to update my belief at all. This is prioritizing anecdotes over data - moreover, my belief (ESP doesn’t exist) wasn’t scientific in the first place!

Epistemology is impossibly hard. Determining the strength of evidence is hard. In the places that a paper contradicts our mental model of the world, the only sane approach is to question the paper.

Little of what you've said is at odds with what I'm saying, it just is dressed up in the pop-Bayesian signifiers that seem to be popular right now.

To put it in that context, what I am contending is that people seem to have unreasonably high priors about the impact of this pandemic on suicides that don't seem to be based on much at all, and I believe underestimate the fundamental complexity of the world, that cause and effect are often difficult to trace.

The fact that you are comparing the idea that suicides have decreased during the pandemic with the existence of ESP is, I think, emblematic of exactly this sort of false overconfidence.

Anecdotes should have some, but marginal, impact on our posterior when coming head to head with data.

> Epistemology is impossibly hard. Determining the strength of evidence is hard. In the places that a paper contradicts our mental model of the world, the only sane approach is to question the paper.

References to the sanity of one approach over the other aside, I do not see how this follows from what you said before. I agree that determining the strength of evidence is hard. I would suggest that creating a causal model of the world from anecdata might be a bit harder.

e: As a sort of meta-point, I really do find it interesting the increasing currency of Bayesian concepts in pop imagination, often combined with arguments against good inference. It seems almost like pop-Bayesianism is the sophistry of the 21st century, both in the sort of argumentation that characterizes it and as a sociological phenomenon, who it appeals to.

If you don't deny a mental health crisis, then it's not sloppy reasoning to increase, in relative terms, your skepticism of studies which show a decrease in a logical consequence of a mental health crisis. This doesn't mean you reject the study. But counterintuitivity surfacing from contrary evidence in the form of both anecdotes and other data is a good as reason as any to be skeptical.

I believe that we had a mental health crisis that predated the pandemic.