|
Rethinking Virtual Violence:
In essence, they carried out a correlational study and then followed up with an experimental study. In the correlational study, Anderson and Dill surveyed undergraduate college students and found positive correlations between self-report on violent video game play frequency and delinquency, and that game play frequency is negatively correlated with academic achievement. They then conclude that this is evidence that playing violent video games leads to delinquency in the long-term. This conclusion is problematic primarily because correlations do not imply causality. Imagine a researcher who measures the frequency of kindergarten attendance and age. He would find a negative correlation, and he might conclude that going to kindergarten causes individuals to become younger. The problem with his conclusion is that correlations never imply causality. Just because two behaviors or attributes co-occur doesn't mean that one causes the other. But one might also imagine a researcher who measures obesity and the amount of junk food eaten on a weekly basis, finds a positive correlation, and concludes that being obese causes people to eat more junk food. The problem is that even though there is a causal effect, a correlational study doesn't show which way the causality runs. In the case with the violent video game study, it could very well be that delinquency causes increased video game play and not the other way around. After all, if a delinquent decides to skip classes and plays video games, it seems odd to turn it around and say it is the video game that caused the delinquency. That would be like blaming the fire on the smoke. And finally, imagine a researcher who measures foot size and IQ, finds a positive correlation, and concludes that bigger feet increase IQ. The problem is that he neglected to consider a possible unmeasured third factor that influences both feet size and IQ - in this case, age. It might be the case with violent video games that a personality factor influences both delinquency and attraction to competitive video games, perhaps low Conscientiousness combined with High Extraversion, in which case delinquency and video game play frequency are innocent bystanders. In other words, the correlational study says nothing about the cause and effects of video game play frequency and delinquency. The problems with the experimental study are slightly more complex. Imagine that some researchers are interested in finding out whether people who purchase laundry detergent as opposed to dish detergent wash their clothes more often. These researchers set up an experiment where recruited subjects are randomly assigned to one of two conditions - laundry detergent or dish detergent. The subjects are mostly white, undergraduate college students taking introductory psychology courses and need to participate in experiments for course credit, or are asked by their friends in the psychology department conducting research for their classes, or are interested in the lottery offered by the researchers. The subjects first sign a consent form stating their understanding that they are taking part in a psychology research experiment. Then they are led into a room with a white shelf resembling those seen in supermarkets, and on this shelf is either a bottle of laundry detergent or dish detergent depending on their assigned condition. The subjects are told they should purchase the item on the shelf through the mock cashier using monopoly money given to them by the researchers. The researchers are interested in knowing whether the purchase of laundry detergent causes people to wash their clothes more often, but because they only have the subject for a short period of time, they devise some indirect methods of inferring whether the subjects who purchased laundry detergent will wash their clothes more often than the subjects who purchased dish detergent. The researchers cannot agree on a standardized method, so the following are some of the indirect measures they choose:
If these researchers find that subjects who purchase laundry detergent score higher on their indirect measures, what does it mean? And what can we say about researchers investigating whether violent video games cause people to be more aggressive and violent using this methodology. First of all, the devised measures have low construct validity. We shouldn't be surprised that exposure to stimulus of a certain theme/genre conditions people to respond quicker to related stimulus immediately after the exposure. Research in automatic-processing and priming in psychology have shown us these effects, and this is nothing new. For example, it has been shown that exposing people to words related to old age causes those people to walk slower and less upright in the short-term than people exposed to other words. And we know from experience that exposing people to horror movies makes them more anxious about being alone and they respond quicker to screams and creaking doors. This is nothing novel. The question is whether the exposure causes long-term effects, and this is something the research in violent video games does not address with its short-term measures of aggression and violence. But there are more serious flaws. People in the real world do not play video games after signing consent forms, under highly controlled and contrived settings such as a psychology lab room, while being observed by researchers. If the researchers are merely generalizing their findings to people under similarly unnatural and contrived settings, then this would not be a problem. However, they are trying to generalize to all humanity, and this generalization might not be valid. Highly controlled experiments tell us a lot about how people react under those conditions, but they may not tell us much about how people act in natural settings. For example, the transparency of their experiment makes it easy for subjects to exhibit the compliance effect, where the subjects try to produce the expectations of the researchers in an unconscious attempt to gain approval from an authority figure. This is especially true when the subjects are young children. A more serious flaw is that both "purchase" and "play" involve choice, and this motivation is ignored in experimental methodologies. This is a flaw because of two different issues. First of all, being forced to purchase a specific item or play a specific game deprives the subject of choice. If the researchers were interested in the effects of forcing individuals to play certain kinds of video games, then their methodology is sound, but they are trying to generalize their findings to people who freely choose what games to play. And this leads to the problem of endogeneity. People in the real world are not randomly assigned to play violent video games. In the real world, there are age, gender, personality and contextual factors that influence different people to prefer different kinds of games. There are reasons why some people choose to play Solitaire while other choose to play Quake or SimCity. If people in the real world were equally likely to enjoy all genres of video games, then this would not be an issue, but that is an absurd assumption. People who play different genres of computer games are probably very different kinds of people to begin with. By ignoring this element of reality, the experimental studies are answering a question that has no bearing in real life. They tell us what happens when you take a group of people some of who would never in their freewill choose to play a violent video game, and force them to do so in a lab setting. Imagine if we forced timid, reserved people to try sky-diving or bungee-jumping, would they really react in the same way as the spontaneous, thrill-seeking people who choose to do those activities of their own freewill. If the timid people express morbid fear and trauma after bungee-jumping, can we then conclude that people who repeatedly choose to perform that activity are accumulating unhealthy amounts of fear and trauma? Is that really a realistic conclusion we can draw? To understand video game violence, a far better question is to differentiate people who are highly likely to choose violent forms of entertainment as opposed to those that never would. In summary, Anderson and Dill use an invalid causality inference from their correlational study to extend the short-term effect of the highly-unnatural lab study. The problem is that the correlational study doesn't support any causal effect, and the lab study doesn't support any claims to long-term effects. The only real conclusion we can draw from research in violent video games is that when you force or bribe white undergraduate college students to interact with projected images of violence in highly controlled, unnatural and contrived laboratory settings, they are more likely to perceive ambiguous actions as aggressive, or have aggressive thoughts and impulses in the short time-period after the exposure to violent interactions. This is a far cry from being able to claim that violent video games cause a long-term increase in aggression and violence in the real world.
|