I've been known to say that economists serve the same function in the modern world that soothsayers did in the ancient world, giving a veneer of divine sanction to the arbitrary choices of the powerful.
@knowuh hey now, meteorology involves multiple computer models and has a reasonable track record of being good about stating their confidence intervals. Economists now… @publius@_chris_real@jonny
There is no Nobel price for economy for a reason: It is a belief system, not a science.
(The price given out every year is by the Bank Of Sweden "In The Memory Of Alfred Nobel", not by the actual Nobel Foundation)
@martenson huh, that seems... stupid. The reason Alfred Nobel didn't create a price for economics was just that, it was not a science. But oh well, the point still stands. It is not a Nobel price, hence why it is named "National price in economics in memory of Alfred Nobel, it is not listed as an official Nobel price.
@WhyNotZoidberg@martenson
There are Nobel prices for Literature and Peace - two decidedly non-Science subjects. Alfred Nobel intended the prices for whoever conferred the "greatest benefit on mankind". During Nobel’s lifetime the academic landscape looked very different - economics is a (social) science, but the underlying statistics were still a young discipline when Nobel was alive. To claim that Nobel didn’t create a prize for economics because he didn’t consider it a science seems absurd.
@johannes_lehmann@martenson no, it's not absurd. Economy is a belief system not a social science, you can tell by the way it collapses as soon as rich people stop believing in it. Plus the weirdness of the theories awarded over the years, like it's normal to only pick your mate by if she's rich or not....
@WhyNotZoidberg
Slight disagree here in that, as physics is scientific study of the real world, economics is scientific study of an imaginary system. It's science in that they take it as seriously as real science. It just bears no semblance to reality... @jonny
@tdverstynen MVP of most hated discipline has got to be the overlap, the "grit" and "growth mindset" research that blames people for structural failings and tells u to just work harder lmao
@jonny Plot twist: "Jönköping University" is not an accredited university. But Swedish law only bars them from using the Swedish term "universitet", not the equivalent word in other languages.
If people think THIS kind of outrageous bullshit is reasonable, you can BET they're going to automatically treat the "AI" wordsalad generators as magical fountains of truth. Without even stopping to think twice.
@AubreyDeLosDestinos@jcolomb@violetmadder@mistersql@jonny if you read it, the person is asking in good faith, it's just that they think that interpolation or imputation might be better from a LLM than from Excel. It shows a deep confusion about several things. I'm imagining someone who needs llama pee for an experiment and can't get any, who just gets some sewer water instead.
I mean, It's one thing to do projection or interpolation by conscientiously building a mathematical model whose functions you understand and disclose.
It's quite another to ask an LLM, "What's a vaguely plausible piece of garbage to dump in this hole, based on random shit digested from millions of unverifiable sources that may or may not have already been getting digested and regurgitated by LLM wordsalad circlejerks in the boiling cauldrons of hell's unstoppable spam?"
The article says it was published in a peer-reviewed journal, doesn't that mean those peers should have checked this before publishing?! I thought that was the definition of the term?
@jospueyo yes ofc transparent reporting is the main reason why these problems are possible, but that would also be super bad to do in this case but at least it would have the pretense of "lost in the sauce of code" vs. "i went through and manually filled them in by clicking and dragging"
In some cases, uses the mean or random value is justified, as long as one notes it. However, there should also be a discussion about how this might make any assumptions difficult. I`d only ever use it for a process study, not published research. I.e. only use it to get an idea about parameters.
Just plugging it in and pretending is fraud. https://www.datacamp.com/tutorial/techniques-to-handle-missing-data-values
@nilsskirnir
I mean theres imputation as a statistical technique and then theres excel autocomplete based off the values of the two preceding cells, which is... not that
@jonny And imputation must be noted and discussed.
I used to do cognitive sci and there was always missing data. A good part of the statistical methods was spent justifying how one filled it in. And it was always the part that the mentors would go on about.........to no end.
@jonny "New Zealand’s data had been copied from the Netherlands"
Imagine a chemistry paper where they couldn't find data on a certain property of bismuth, so (without disclosing it in the paper) they used data about beryllium. Not because that's the closest element on the periodic table, or even anywhere near it, but because they're closest in alphabetical order.
@ids1024
RIGHT. its a perfect example of the affordances of a tool dictating outcomes. "Need a value here, and the thing at hand is to click and drag from the neighboring cell, so..."
@NicoleCRust
They seemed to genuinely not know it was a bad idea to do that. Like they openly volunteered that was the method, and even demoed how to do it when this person asked. Usually im like "if you need to do imputation on something that isnt a well-behaved, oversampled timeseries then what you really need to do is change your analysis strategy" but this is like 5 levels beyond that
Add comment