I’m on the fence a little about this one, but want to throw it out there.
“How Data Mining Facebook Messages Can Reveal Substance Abusers.
Substance abuse changes people’s patterns of behavior, and this is detectable in their social media messages, say researchers.”
I have mentioned in the past about how machine learning roughly works.
You feed it a ton of data, tell it what is good and bad (roughly) and then it goes to work on said ton, learns and spits out some results on the data.
You tweak it and feed it back in. In short order, the machine knows better than we (humans) do.
This is exactly what they did.
Got people to volunteer if they used drugs, alcohol or smoked and posted on Facebook.
The machine learning used those people’s habits to learn the differences in their posts between when they were clean and when they were on drugs.
Once that baseline was set, they then released the AI on Facebook at large and….. well, according to them….
The results make for interesting reading. The team says its technique was hugely successful. “Our best models achieved 86% for predicting tobacco use, 81% for alcohol use and 84% for drug use, all of which significantly outperformed existing methods,” say Bickel and co.
Whoa. So the computer can pick with ~80% accuracy if someone is posting on Facebook under the influence.
Scary. Super scary.
I will leave it to you to think about that for a moment.
(And Dan, just because you don’t post to Facebook does not let you off the hook, you have a blog right… it can be scraped…. Gary. Mate. No hope for you :-).
So why was I on the fence about blogging about it? Because it appears that the study has not been peer reviewed. The whole ‘study’ was done by a couple of guys that had an interest in machine learning.