New Book Cuts Through Violent Video Game Myths

Terry Bosky suggests a recent interview from Game Couch with one of the authors of an upcoming book which fights the "myths and hysteria" surrounding violent video games. Dr. Cheryl K. Olson explains how many of the studies linking aggression with video games were flawed or misguided, and she discusses some of her own findings. Quoting: "Until now, the most-publicized studies came from a small group of experimental psychologists, studying college students playing nonviolent or violent games for 15 minutes. It's debatable whether those studies are relevant to real children, playing self-selected games for their own reasons (not for cash or extra credit!), in social settings, over many years. But media reports and political rhetoric often ignore that distinction. Also, the most-published researchers have built their careers around media violence. Their studies were designed under the assumption that violent video games are harmful, which dictated the questions they asked and how they framed their results. Media violence is just a small part of what we do, so we could look at the issue with fresh eyes and no agenda."

Re:Maybe I read that wrong

[Necreia]

"Why does that make their results any more trustworthy?" Because they don't get paid to come to a directed decision. They don't get money for saying that "Video Games don't cause violence" OR "Video Games do cause violence". They don't have an agenda, so what they say is based completely on the research and not instilled opinion. If you only trust sources with an agenda, then I pity you.

Re:An accurate sampling?

[Anonymous Coward]

Oh common that's statistics 101, kid. You could get a more accurate result with 120 participants, or a less accurate result with 12000 participants. That size of the sample doesn't make a difference as long as it is representative.

I am a researcher in this field

[Anonymous Coward]

Not violence in games specifically, but I have done original research on behavior and its relation to violent video games. My results show a temporary increase (lasting around 15-20 minutes) in violent behavior after playing a selection of video games with varying types of violence. Some of these are first person shooters, some are fighting games, etc.

Now, before you naysayers get your panties in a bunch, keep in mind, we are professionals. No one in my group had any agenda apart from doing good research. We had no stake in the outcome, and were not funded by anyone who would be able to influence our results. We controlled for all the variables you can think of and plenty you can't.

The results are good, and I trust our conclusions.

And if even one of you tries that "correlation is not causation" thing you love I'll scream, especially since it doesn't even apply to our study.

Re:An accurate sampling?

[Bigjeff5]

You really aren't versed in survey sampling standards, most surveys only involve a couple hundred people, if that. The way surveys work is you use a small number of people but you statistically balance the people involved based on catagories (they call these demographics). For example, if 40% of kids who play video games are between the ages of 6 and 10, white, and come from middle class families, then 40 out of 100 kids in the study need to be between the ages of 6 and 10, white, and come from middle class families. Depending on how accurate you wanted to go, if you have accurate demographics to start with you could get decent results using 100 kids in a single town, but that would be very very hard to do, and hard to verify your results. It's not 1200 that should worry you for accuracy, that's actually a pretty large number; it's the two states part. It seems to me they may not be taking region of the country into account for this, which might be a factor or might not.

Re:An accurate sampling?

[esme]

You're being to simplistic -- the sample size needed for good predictions isn't directly related to the total number of gamers. The size of the sample needed is related to error introduced by the measures used and the phenomenon measured. If you have a robust methodology, you may need only a few subjects. If there are huge errors introduced by your methodology (political polling is a good example of this), you may need thousands of subjects.

I didn't read the article (this is slashdot, after all...), but any good psychologist would include statistics indicating the probability that the results were caused by error or random chance, usually this number should be very low, 5% or lower. See the wikipedia article on P-values [wikipedia.org] for more on this.

But to answer your question: many psychological experiments are done with a much smaller number of subjects (50 or so), and get very low P-values. The effect being tested here may be harder to reliably measure, but the sample size is also pretty large. So there's no reason to think that 1,200 is too low, unless the stats say otherwise.

-Esme