~No, no, no. That's not what ancestor post was saying at all.~
~Watching Game of Thrones leads to playing Grand Theft Auto, and doing that will make you a murderer. When you're sliding down the slippery slope, you have to hit every rock on the way down.~
The theory reverses the arrow of causality.
If the Army uses simulation and gaming as part of its psychological conditioning program to not only get its soldiers to fire their weapons, but also aim at and kill the enemy, that does not imply that those games or simulations, used in isolation, will necessarily result in desensitization to violence.
I seriously want to design a contagion that imparts to individuals an intuitive understanding of conditional probabilities, and how to reverse one.
If you know that 99% of child abusers viewed images of child abuse prior to engaging in that activity themselves, you only know P(A|B), where A is viewing photos, and B is taking action. You do not know P(B|A), which is the likelihood that someone that someone who has viewed photos has done abuse, unless you also know the unconditional probabilities for P(A) and P(B). You can probably infer those two from crime statistics and investigative activities like those described in the article. But given the existing legal prohibitions and resultant secretive behaviors, I doubt you could determine either with a great deal of precision.
If (hypothetically) P(A|B)=1.0, P(A)=0.001 and P(B)=0.0001, then P(B|A)=0.1 . That means that even if you know that everyone who did A also did B, you can't assume that everyone who has done B has necessarily done A.
We could more easily calculate for the Game of Thrones and murder hypothesis.
Let's say A is "committed murder in 2014" and B is "watched GoT in 2014". From crime statistics, I estimate that P(A)=0.000045 . From ratings estimates, P(B)=0.063 . Now, I'm not entirely certain how many new murderers in 2014 had watched GoT, but I'll guess that P(B|A) = 0.063 . That makes P(A|B)=0.000045 . Huh. It looks like I assumed they were independent events....
If you went and did an actual survey of 2014's murderers, and discovered that P(B|A) was actually 0.12, that's a great find, but it still only means that P(A|B)=0.000086 . That's not a convincing argument that watching GoT causes actual murder. It just means that 0.0086% of GoT watchers committed murder. That might be explained by GoT inducing violent behavior in its viewers, but from the numbers, it's not bloody likely.