Maarten van Smeden
maartenvsmeden.bsky.social
Maarten van Smeden
@maartenvsmeden.bsky.social
statistician • associate prof • team lead health data science and head methods research program at julius center • director ai methods lab, umc utrecht, netherlands • views and opinions my own
NEW PREPRINT

Explainable AI refers to an extremely popular group of approaches that aim to open "black box" AI models. But what can we see when we open the black AI box? We use Galit Shmueli's framework (to describe, predict or explain) to evaluate

arxiv.org/abs/2508.05753
August 11, 2025 at 6:53 AM
We should really ban all use of AI from all our education because the use of AI will make our students dumb, and banning innovation has worked really well before
May 19, 2025 at 2:32 PM
April 3, 2025 at 11:57 AM
PROBAST+AI is out! A risk of bias tool for prediction models developed using any kind of analytical approach (stats/ml/ai/ds)

Really proud of this collaboration with many prediction model experts around the world

www.bmj.com/content/388/...
March 24, 2025 at 11:16 AM
Five years ago @laurewynants.bsky.social and I started this living review. Within 1 year we reviewed >700 prediction models for COVID and we found many to be poor and decided to stop reviewing. Many external validation studies followed confirming many indeed were poor

www.bmj.com/content/369/...
March 7, 2025 at 8:22 AM
NEW PAPER

On quantifying uncertainty in your predictions. Excellently led by the group of @richarddriley.bsky.social

www.bmj.com/content/388/...
February 13, 2025 at 11:24 AM
Periodic reminder the world of data analysis cannot be meaningfully categorised into "machine learning" and "statistics". Two cultures with substantial overlap in the use of methods (e.g. logistic regression), analytical goals (e.g. causal inference) and history

jamanetwork.com/journals/jam...
February 4, 2025 at 8:59 AM
Scientific publishing cannot be free but an annual $2.5 billion seems excessive to me
January 27, 2025 at 1:13 PM
NEW PAPER

Really glad to see this one in print: the harm due to class imbalance corrections in prediction models developed using ML/AI

Excellently led by @alcarriero.bsky.social

onlinelibrary.wiley.com/doi/epdf/10....
January 27, 2025 at 8:01 AM
There is a pretty extensive description of the AUROC in the paper, including that it is "just" semi-proper. We also specifically state that model comparisons should use proper metrics and are not the focus of this paper. Hope that you agree with this :)
December 16, 2024 at 3:33 PM
Just had a laugh remembering this "original" research piece bmjopen.bmj.com/content/12/2...
December 16, 2024 at 12:46 PM
My favourite thing about the paper is this summary table
December 16, 2024 at 8:50 AM
NEW PREPRINT

A detailed overview of 32 popular predictive performance metrics for prediction models

arxiv.org/abs/2412.10288
December 16, 2024 at 8:44 AM
When someone confidently explains to you the relations between Statistics, ML, AI, Data Science etcetera, do a quick Google search and find many people confidently explaining the same relations very differently
December 15, 2024 at 5:43 PM
NEW PAPER

A scoping review of updating methods in AI/ML based prediction models

doi.org/10.1016/j.jc...

w/ @lottameijerink.bsky.social and others
December 12, 2024 at 11:11 AM
All the time
December 6, 2024 at 10:56 AM
Saving this for a future reply to peer review
December 6, 2024 at 7:53 AM
NEW PAPER

Paper excellently led by Kim Luijken on dealing with interventions/treatments when developing a clinical prediction model

Open access 👉 onlinelibrary.wiley.com/doi/10.1002/...
November 28, 2024 at 3:39 PM
Thank god for DALL-E, this will make things so much easier
November 27, 2024 at 3:57 PM
Academic publishing really is a billion dollar business largely based on public money
November 23, 2024 at 9:05 AM
With the prices of publishing "Gold Open Access" these days maybe we should rename it "Platinum Exclusive PDF on the internet"
November 23, 2024 at 8:38 AM
Reminds me of this beauty by @xkcd.com
November 22, 2024 at 12:38 PM
I made one for stats papers
November 18, 2024 at 4:02 AM
New paper

Wondering how to select the appropriate tool to help assess diagnostic and prognostic models in systematic reviews? This new paper aims to help

Open access 👉 www.jclinepi.com/article/S089...
November 15, 2024 at 10:03 AM
But let's not only focus on the bad things, good initiatives happen too
November 13, 2024 at 7:42 PM