Aidan Wright

And the key missing ingredient here is data collection. Have you asked statisticians or computer scientists about collecting data? They typically do a poor job of it, because good data collection requires understanding human behavior.

2/2

I know what you’re trying to say, but have you tried talking to statisticians about human thought, motivation, or behavior? I think we want experts in those things first and foremost, even if it means they won’t be exceptional statisticians.

1/2

This is a pretty dismissive comment. There is a range for what psychologist and behavioral scientists more broadly want to do with their work. Some is straightforward, some complex. That’s ok as long as the method fits the question.

I'm being a little dramatic to make a point, and there is value in its flexibility and accessibility for esoteric methods, but it's far from a panacea and too often people treat it like it doesn't have significant downsides.

Me personally? Nothing. I really dislike so much about using R. The inefficiencies, the poor documentation, relying on people to update packages or add necessary features when there is little incentive to do so, etc. all grate on me. It's major strength is that it is free, but you get what you pay 4

Lots of past examples (e.g., SPSS GUI) to argue against this best case coming to pass, at least not without effort.

3/3

But I want to be clear, that time saved not on coding needs to be spent on even deeper conceptual understanding of statistics and how they work.

This might be best case for how gen AI works, is it takes away drudgery and allows for focus on higher order stuff.

2/3

I don't think many applied statistics professors are teaching mathematical proofs, which is what the C-H correspondence is about. At least not if they want to be effective.

Reasonable disagreement might exist about value of coding, but too often I see it as getting in the way.

1/3

I get that some coding can help ensure the formalism of the model and facilitate really engaging with and understanding each part of a model.

But so much of it is where does the colon go, and this one uses pipes instead of semicolons or parentheses. I 100% would want to get rid of all that.

2/2

This is sort of the question, right? Does coding just go away now?

I'll be the first to say I hope so. I am unabashedly in FAVOR of doing away with coding. As someone who teaches mostly statistics, I find teach/learning coding to be a terrible distraction from learning the methods and models.

1/2

I had similar thought to you here. I share your concerns. bsky.app/profile/aida...

From vibe coding to vibe science? I think we want to avoid entering the era of vibe science.

At least in general, I think many areas of psychology have been doing vibe science all along.

Good grief no. These people are incorrigible. It’s a waste of time.

Lot of that here. We don't ever even meet as just the faculty. Every other group in the department has special dedicated meetings. Not us though.

It really hampers our scheming.

You know why

Yeah, it’s DAGs overlayed with ML. What a fucking nightmare.

a cartoon of a man with a smoke ring around his head says @john.melo on the bottom ALT: a cartoon of a man with a smoke ring around his head says @john.melo on the bottom

Cool is one word for it. :-)

a woman says it 's the same propaganda in front of a window ALT: a woman says it 's the same propaganda in front of a window

A colleague asked me today about "causal discovery analysis" - never heard of it. Sounded fishy. Looked it up, and it is rotten fish all the way down. The Directed acyclic graph folks are at it again with their irresponsible naming of this technique. Flashbacks to 2017 and "network" propaganda.

I get it. I’ve long said I have a very parenthetical cognitive style. Eg I always use the verbal equivalent when I talk. But I’m also terrible at punctuation (and anything with the formalities of writing really). So the way AI uses em dashes really has been revelatory.

AI has taught me that em dashes are a really useful writing tool! Don’t know how I lived without them.

I'm totally with you, just adding an asterisk

Yeah. That and sometimes you just can't assign a paper to a content expert when considering all the other issues involved (e.g., individual AE burden, breadth of content covered)

No re-review is good ideal, worth shooting for, really hard to be absolute in practice

This is the policy at the journal I edit. Or, rather, #2 is strongly encouraged, but doesn't always work because of legit expertise gaps and need for input from experts.

I have…adopted a different strategy from yinz. I lift whatever portion of my body weight is left over from lying down with my feet as the fulcrum (you might know this as a push-up), I do that quite a few times, and am closing in on 175lbs from the other direction at 6’2”

a close up of a woman 's face with a purple shirt on . ALT: a close up of a woman 's face with a purple shirt on .

ooh, I know this one!

Talk about not getting the assignment

I keep thinking, that for many areas of research, in the era of GenAI the data have become relatively more valuable than analysis, interpretation, and dissemination.

Saying the same thing on two social media platforms is like telling the same story to two groups of friends.

I think introductions should end at 300 words.

It was a long-winded explanation. This distinction will help me improve the process of explaining this moving forward so I have a better mentoring product. Thanks!

5/5