The study claimed to be able to predict cannabis impairment using only the smartphone data with about 67% accuracy
As someone who studied data science and did a good bit of machine learning in that time: That's terrible and it means that the whole story is a big nothing burger. It's barely better than random guesses that match the distribution.
You don't even need sensors, you can tell from the websites/apps people use and especially from their typing behaviour. I can do 50WPM on my phone when sober, but less than half that when I'm not.
That i could see but not everyone who is stoned is typing on their phone all the time. Article makes it sound like it can do this automatically unobstructivly.
Seems far easier to detect the sound of a lighter, crunchers hitting eachother and detecting sentence like “i am going outside to smoke my joint” combined with daily/weekly patterns in time.
Thanks to this i did find a great nerdy app to play around with my phones electromagnetic field sensor though.