May 31, 2019

expert reaction to study on violent video games and behaviour with real guns

Research, published in JAMA Open Network, reports adolescents playing 20 minutes of violent video games are more likely to practice dangerous behaviour with a gun than those who had played a non violent game.

Prof Andrew Przybylski, Associate Professor and Director of Research at the Oxford Internet Institute, University of Oxford, said:

“This is a very curious study which I believe should be viewed with high skepticism given the combination of its attention grabbing claims and actual details on how the study was conducted.

“Quite worrying, the study claims that it is a clinical trial but it does not meet this standard. There is an entry on clinicaltrials.gov but discrepancies between the information here and what is presented in the paper lead me to worry about the peer review process. Many measures present in the paper are absent in the trial plan, the numbers of participants vary between the plan and the paper and many essential details, such as the statistical analysis plan are entirely missing. Taking these together I’m quite concerned that the public or journalists might conclude this study was rigorously done… it was not.

“Further, the authors are non-commital on whether it complied with best research practices. Many indicators relating to well-done science such as data sharing are hinted at but are not followed through on. The registry shows the study is ongoing and the researchers have repeatedly, on three occasions, once recently as last month, delayed submitting their results for quality control checks on clinicaltrials.gov.

“The paper has numerous other obvious problems such as low statistical power, an unrealistic scenario involving how a 12 year old would view a decoy weapon in a university psychology laboratory, and its casting of Minecraft as a violent video game but the two statistical anomalies stick out. First, the results and outcomes mentioned in the plan registered on the trials page are either not statistically significant or are just barely, because there is not enough detail in the analysis plan this patter is suspect given the researchers stopped data collection early. Second, nearly 30 additional models are tested giving the hypothesis a large number of times to ‘work’ but the non violent version of Minecraft only looks like it has a smaller effect on handling a gun in one of these cases. We would expect at least one or two of these to be different on the basis of chance.

“All in all, the concerns I highlight should underline any doubts readers might have given the authors are essentially claiming that playing Minecraft for 20 minutes could convince an adolescent to play with a mysterious gun in a university laboratory is a good idea. It’s far more likely that these findings and the resulting press release is tapping into something less news worthy, statistical noise.”

* ‘Effect of Exposure to Gun Violence in Video Games on Children’s Dangerous Behavior With Real Guns’ by Justin Chang et al. was published in JAMA Open Network on Friday 31th May.

Declared interests

Prof Andrew Przybylski: I have published research on the topic of violent video game effects (https://royalsocietypublishing.org/doi/10.1098/rsos.171474) and I have no conflicts of interest to declare.