I took a brief look at that article. It's based on a self-report survey of 97K people in Japan.
So the first things to note are:
- It's a self-report survey, not an experiment. So despite the headline, no cause and effect statements can be made.
- Presumably, most of these people are Japanese, and they play mostly Japanese games. Results might not generalize to other countries.
They found the people who had a console, and played games on them, had improved mental health over their peers, however those people who played for over three hours a day saw diminishing additional benefits. They also found the simple act of owning a console was linked to improved life satisfaction, and helped reduce their psychological distress.
Correlation doesn't equal causation. There are many potential confounds here that haven't been mentioned. To name an obvious one, maybe the people who own a console (as opposed to those who don't) are better off financially, and it is their higher income that contributes to their improved life satisfaction, rather than merely owning a console.
It makes sense that any positive correlation would drop off for people who played 3+ hours a day.
They also note that the survey was completed during covid, so that could've impacted the results.
I'd like to believe that playing games improves mental health, but I doubt that's true - as a general statement, anyhow. It may be true for some people, at some times, with some games. But overall, on average, for most people? Doubt it. In fact, I believe there is data suggesting the opposite (although I'm not sure how strong that data is).
p.s. "Playing Switch within the survey month improved mental health by 0.81 s.d. (95% CI 0.53–1.10,
P < 0.001), while playing PS5 resulted in a 0.20 s.d. (95% CI 0.05–0.36,
P = 0.012) improvement." So there you go. Playing Switch is 4 x better for you than playing PS5. ;p
I'm very suspicious of their statistical methods. They are using something called causal modeling, which is a statistical method to try to wring causal relationships out of purely correlational data. That sounds quite fishy to me. I don't know enough about causal modeling and don't have the time/patience to dig in depth, but it's not something I have much confidence in. Researchers like to do this a lot - try to derive causal inferences from correlational data. I can see why it's popular - correlations are very easy to obtain, and it's much harder and more expensive (sometimes impractical) to run experiments. But everything I know about experimental design and statistics indicates that you cannot derive causation from correlation, no matter what you call your statistical manipulation.
That's probably more than you wanted to know, lol.
