[halhen]

I took a job in a data analytics firm, which had the national Swedish gambling operator as customer. The idea was to predict which players were at risk of developing gambling problems using ML. Backing this was also a team of psychologists. (I left about a year ago, so I'm writing in the past tense below.)

As it turns out, the task became to identify those who are at risk today from gambling data, which is relatively easy to do probabilistically with actionable precision. From here, the next step was to figure out if the person wanted help to change, which also is surprisingly easy to do by just asking.

This is by the way also to the industry's excuse that "we shouldn't tell adults what to do!". When asked, on a gambling site, a majority of those with gambling problems SAY that they would like to change, but aren't able to. Then it's no longer about telling people how to behave, but instead helping those who ask for help. But by not asking, the industry can get away with "we shouldn't tell..."

The really hard nut to crack was to give those who want but aren't able to. It's definitely not just about telling them -- most know, _and take responsibility for_, that they do destructive things. Normative feedback (this is how much others play) seemed to work on a mid-term. Calling people also did some positives; many were relieved that somebody saw them and that they finally could talk to somebody, very few got annoyed. But as a first approximation, fighting problem gambling is still an unsolved problem.