Erin Jonaitis

Applied statistician at UW-Madison, interested in longitudinal data analysis, reproducible workflows, and Alzheimer's disease. Views mine; all my errors are independent.

Last Notes

2025-03-27 13:31:22 UTC

- reply

That link isn’t working for me; not sure if it’s server nonsense — is there a news story linked?

2024-09-04 16:32:13 UTC

- reply

Teaching non-practitioners to understand what they're looking at is important, so thanks for making this a living course. My sense is that public understanding of what "AI" means is fairly weak. We're now seeing compliance folks pushing back on any use of our own data to train any AI model, which in the limit could result in absurdities like not being permitted to set internal cutpoints for normal vs abnormal on a blood test (which is necessary because not everything used in research is a clinically validated test that comes with package inserts, and because some assays have normal ranges which vary significantly by lab). I'm on board with a restriction on using generative AI specifically, but "no AI at all" pushes very far back into the past in a way I think laypeople don't really understand. People were doing AI work when I was in my PhD program 20 years ago! And it wasn't new then, either.

2024-08-22 14:30:36 UTC

- reply

I've had my mind blown by the Indigenous data sovereignty movement. Totally new way of thinking about consent for me. And for small genetic groups, particularly marginalized ones, it makes a great deal of sense.
For some things, a risk management dial pulled all the way to one side defeats the purpose of the thing, and we need to contend with some middle ground (think "safer sex"). I like the data sovereignty movement because it sets the dial somewhere in the middle: yes, we want our nation to be connected to the wider world, because we need some of what biomedicine has to offer and most of it is happening in that wider world, but we want to do it in a way that mitigates the risk to the nation.
(This is not a commentary on All of Us, I have no insight into that study and don't know whether they have any relationships with Native groups. One of the Alzheimer's cohorts I work with does, which is how I became aware of this movement.)

2024-08-22 14:19:31 UTC

- reply

I think this touches on a big question: under what circumstances should community norms prioritize risk management vs connection (to history, to other people)? Consumer genetic services are another example where this tradeoff comes into play, and an interesting one because "individual decisions" sort of aren't, really.

2024-08-22 13:54:53 UTC

- reply

*nods* Yeah, I see that, in terms of sensitive items (and of course all it takes is a shifting of political winds to change what's sensitive). It's a bigger threat with electronic records, I would expect, because it's easier to do stuff en masse, but of course if the set of books you want to snoop on people for is small, the analog version was plenty risky.
The thing I liked about these cards was that they made tangible something about the community function of a library, showing you how you and other people are related. It wouldn't surprise me if it's a general principle that the practices which can generate this kind of feeling also present risks.

2024-08-22 13:45:54 UTC

Found out where an item I recently ordered via ILL is shipping from, and I thought, hey, I bet I know a person or two who's used this very book. I kinda wish we still wrote names in cards in the back. That was a nice cultural practice.

2024-08-20 19:16:11 UTC

- reply

I thought about that one too, actually, but it sounds like a bigger time commitment than I'm looking for right now.

2024-08-14 02:41:24 UTC

- reply

Pleased to say that the voters of our state wisely saw through this ploy and rejected the referenda in question. I hope others were as disgusted by the disinformation as Mr. Me and I were.