However, isn't this how Bayesian thinking works? You give a higher prior to your believe depending of the source.

That's why we ask for sources. If one doesn't trust lamontcg to know what they're talking about, we shouldn't trust them to be honest about that fact. We're often more likely to have Bayesian priors about a primary source than the people who refer to them.

Yes, but in this case authority would be your prior, and a source would be the data to update that prior. If you are only relying on authority, you are solely going off of your personal biases, which while potentially informative, isn't as useful as actual data.

A perfectly reasonable approach; but then it is easier to ask people you trust rather than reading HN. Anonymous internet people are inherently unreliable.

Authority is constantly used as an effective source of trust in science.

As an effective shortcut but it's still the wrong thing, ultimately, to trust.

A scientist who doesn’t use sources effectively and accurately themselves is not to be trusted as either an authority or as a scientist. It is the greatest sin for a researcher.

What is the alternative? Everyone grabs postdoctorate degrees in everything that might come up?

Your question shows why it is a shortcut.

Heuristic might be a better term.

There is also a massive reproducibility crisis in science, and the two points are not not unrelated.

The “reproducibility crisis” is largely overblown. Individual scientific papers have always been close to junk on average. Look at a wide swath of papers from say the 1950’s or even 1850’s and you see not just poor methods but also outright fraud.

It absolutely isn't overblown, and "individual scientific papers have always been close to junk on average" is a pretty damning indictment.

Meta-studies have found that, depending on the field, up to 70% of published papers can't be reproduced. I'm not sure its possible for considering that a crisis to ever be "overblown".

Even 30% of papers being reproducible is orders of magnitude better than random chance. That’s the signal that pushes science forward.

As to old papers, go back to the 1950’s and plate tectonics was still being debated. What do you think the other geological papers where based on? When you read a specific older paper it’s generally because it stood the test fo time, but that’s just selection bias.

I completely agree... in principle. That being said, I actually think we tend to underestimate the value of trusting experts. Often times the sources out there that one could be quoting espouse conflicting and ambiguous conclusions and it ironically takes an expert to be able to parse through them all. If we could all trust experts a bit more and, rather than ask for sources, instead ask our experts to explain their reasoning to us, I think we'd all be better off in the long run in that we'd actually be learning from them rather than trying to constantly bootstrap becoming experts ourselves.

Unless you have a background in immunology the sources are probably not terrible useful to you.

Probably, but not definitely. The only thing preventing you from understanding the sources yourself is some effort. It's not as hard as it is sometimes imagined.

One problem with that is that any individual study should be read in the context of other published literature, if you want to form an informed opinion. The only alternative is authority - either looking at the journal where it was published, the authority of the authors, or the authority of someone who recommended the paper.

Otherwise, it's easy to find plausible-sounding papers that have failed to reproduce in subsequent studies, that use too low samples compared to similar studies with different conclusions, that were later retracted etc.

Oh, they will be, but in the opposite way to what GP seems to want. That is, I'd look at the sources as further sources of authority :). E.g. a bunch of papers from multiple teams, with some published by high-profile journals or authored by well-known researchers, indicates it's relatively safe to trust their conclusions. A bunch of papers from the same few people who seem to want to push their pet theory indicates low trustworthiness and need for independent verification.

Unless you've studied the domain in question in depth, it's not like you could use the citations in any other way. Trust, as a heuristic, is fundamental to advanced research as much as it is for everything else in society.

They could be useful to somebody else, and that's a good enough reason to have them.