Hacker News Reader

Comment by voldacar | Hacker News Reader

Given sane priors, shouldn't one always assume that they penetrated farther into the continent by some amount than the recovered artifacts indicate?

1) for any single artifact X, the chance that X is found at the innermost point that vikings reached is low, given the large number of artifacts they would have generated

2) the exact number of discoverable artifacts is unknown, and the majority of this number must be undiscovered so far

Archeologists' failure to think this way has always baffled me. Bayesian archeology should be a thing

View on HN

I think that they are being Bayesian, they're just holding it differently. Archaeologists usually look for a healthy body of evidence - entire settlements, for example, or evidence of campsites - before assuming that someone was there. A single out-of-place artifact with broken provenance, though, has little evidentiary value, because the broken provenance makes it difficult-to-impossible to establish when the artifact got to where it was. Even with good provenance, an isolated artifact is usually taken as evidence of trade rather than presence. (For example, see the discussion of the coin in TFA.)

So, you take a relatively lightweight piece of evidence that they made it even that far, stick it together with an enormous prior that they didn't, and, well, per Bayes' Rule, that artifact doesn't really shift the posterior probability by much. And, if Bayesian thinking can't even convincingly get you that far, it certainly can't get you any further.

Except there's another factor:

3) Artifacts can be (and often were!) traded between different places, so presence of an artifact associated with a culture may be indicative of trade and not physical presence of the artifact.

Yeah that is definitely true. In fact that makes a lot of these questions essentially impossible to answer

I'm pretty sure they do think this way. That's basically what this article is about.

I've learned you have to distinguish between what is known to be true with near certainty, what is considered plausible, and what is at the edge of possible. The history books focus on the first, but that doesn't mean archeologists aren't thinking about the 2nd and 3rd. Speculation on what is possible is what helps them decide what projects to attempt.

I don't know your background, but I first encountered this in either my first or second year of undergrad archaeology classes. It's definitely a well-appreciated principle.

As others have said though, it's not an appropriate evidentiary standard for declaring things publicly. That may be where the confusion is coming from.

Leave it to programmers to completely dismiss another field, in which they have no knowledge or experience, in two sentences.

I have heard it from lawyers too. I think it's very human.

Remember when HN wowed to disrupt Hollywood? Because surely it couldn't be that hard to make better movies and tv.

Other fields are rarely rigorous, and often misuse science. If the dismissive attitude is present, it has foundations.

And wishing that another field used legitimate statistics is hardly 'dismissive'. Its a reasonable question. The answer might be "because they don't know how".

Programmers accusing other fields of not being rigorous is like the pot calling the enamelware kettle black.

This is a field where favored practices in machine learning come and go at a furious pace, often based on the results of a single paper that didn't even attempt to isolate the new idea they're presenting as the primary cause of their result, as opposed to the 10 other variables they mention in the paper and that everyone knows can make a big difference, but that they didn't bother to control for. Other fields might not be uniformly meticulous about isolating confounding variables, but at least there's a common understanding that it's good to try.

So? How does that make them wrong?

You are confusing positive and negative claims.

1) Vikings were definitely at point X.

2) Vikings did not get further than point X.

Yes, due to the likelihood of undiscovered evidence, you can assume that things happened for which we don't have concrete evidence. You just can't determine exactly what those things are, so there isn't really anything interesting you can say about it.

I don't think archaeologists ever make claims like "People X never reached Z" when Z is an easy 50KM journey from Y, where People X were known to have been, without pretty convincing evidence to the contrary (e.g. a mass grave containing the entire estimated population, showing they all died pretty much on arrival)

Your claim that archaeologists don't think that way is bogus. How do you think they decide where to look for new evidence? A lot of archaeology may be driven by building works needing to prove they aren't going to destroy a Roman villa, but plenty is still driven by academics thinking that a particular meadow has a good chance of containing artefacts.

Depends on what you're trying to work out - if you want to know the furthest some lost drunk straggler wandered off and died in the woods - yes, but if you want to know the extent of relatively large exploration/camps then a single artifact doesn't actually prove that.

Given the incredibly small number of artifacts that are going to exist, I don't really see why it would matter.

The story about the sailboat trip to find the camp on the beach shows that they are plenty willing to investigate assumptions based on evidence they do have. With ~0 evidence to go on and a continent to search, what more would they do?

Reply on news.ycombinator.com