This is /., so no one RTFA, but itâ(TM)s the Finnish parliament that stopped it for political reasons in December, only one year into the two year experiment, not because it failed. We wonâ(TM)t know what happened in the study until 2019.

It's any social media use at all while at work. One of the dimensions of "good" behaviors was participating in an online work community. Presumably, most people would not think that using Facebook would help their job performance, so they would not report that as "good".

Some "good" example survey items linked in the article:
I request help from people on social media when I am having trouble solving a problem at work.
I communicate with existing customers or clients via social media.
When someone posts something negative about our organization or its employees on social media, I try to do something about it.

The people participating in the study thought these behaviors would help their job performance.... communicating with customers, reaching out to new customers, participating in an online work community, communicating with coworkers, gathering information from colleagues, asking friends/coworkers/family for help solving a work problem, and using social media as a technical solution (e.g. transferring a file from one computer to another). On the surface, it looks like these things would help in many jobs. But from the data, they were unassociated with better work performance.

Well, the study is just about "leading to aggression" and not "leading to homicide." The NES often made me, at least, want to throw my controller through the wall. That experience is probably a lot less common these days (in this age of easier, accessible gaming).

I believe they collect anonymous click data. But guess who disables/declines collection of anonymous click data? People who actually know how to use their computers.

At some point, sure. Programmers will all be out of work once computers can effectively program themselves. But right now, it's not so straightforward. The OP describes call centers, and that's a really good example - while a simulator can present call center tasks to a job candidate (with simulated customer voices, for example), a simulator responding to customer service calls would not be nearly so successful.

A video game simulation is a controlled environment (HR can create a set of scenarios to be tested in the simulation) but real life is more random. It's up to HR to create a reasonable sample of work scenarios for the simulation, and the quality of those scenarios is directly proportional to the quality of the information you get out of the simulation. And it's also important to realize that they not claiming that simulations are the only hiring tool you'll ever need. They're just better than interviews alone (and way better than unstructured interviews, which are almost worthless).

I will say that I don't know a single organization that uses a simulation and NOTHING else. It would be very difficult to assess characteristics like interpersonal skills and job experience. A simulation is usually just one step of a larger hurdle-based system (usually a late step, since they tend to be expensive per-applicant).