I completely agree. Actually, I made a very similar post (post no. 6 "misleading statistics", score:0) but got buried and nobody modded it up. This and the quality of the comments shows how ./ has changed over the years.
Back to the study, the number of people who behaved socially per group were less than 5 in most cases, so doing any form of test of fit is just plainly wrong. And I blame Plos ONE for publishing it, as their criteria for acceptance is "Experiments must have been conducted rigorously, with appropriate controls and replication. Sample sizes must be large enough to produce robust results, where applicable." (http://www.plosone.org/static/publication#technical)
They define prosocial behavior as handing a pen that has fallen, based on a published paper about mimicry (http://pss.sagepub.com/content/15/1/71). The thing is that the researcher waits for up to 5 seconds. Count them, it's really long when confronted face to face. I'm not saying it's wrong, but that they could be measuring submission instead.
And finally, they did not test for increased violent behavior which should be the most obvious consequence of playing violent games.
So failing to find a difference when testing for something not completely related with violence using a an underpowered experimental design... to me, that's propaganda for the gaming industry.
an increase in prosocial behaviour is misleading to say the least, or propaganda for the gaming industry.