Exactly.
The general approach is to use interpretable models where you can understand how the model works and what features it uses to discriminate, but that doesn’t work for all ML approaches (and even when it does our understanding is incomplete.)
Maybe not the hardest, but still challenging. Unknown biases in training data are a challenge in any experimental design. Opaque ML frequently makes them more challenging to discover.
I wonder if there will be an anti SLAPP action soon from Cohen’s team. (Not sure what the rules are in that jurisdiction.)
They also make NEMA plugs with threaded rings around the boot for applications in marine and other harsh environments.
The most serious thing I heard of was Sotomayor’s staff intimating to universities bringing the Justice in to speak hadn’t purchased enough copies of her book.
Can’t wait for the calls of “But her books!”
Figuring out the Parkinson’s linkage is challenging too, because glyphosate is just one of many chemicals used in agricultural settings. It wouldn’t be surprising for the correlation to be caused by another chemical with strong evidence of casual linkage to Parkinson’s that itself is correlated with glyphosate, like Parquat. (Since Parquat is a herbicide, places that used it may also use (or have switched to) glyphosate.) Totally worth continued scientific study.