News outlets often discuss scientific research investigating possible causal relationships between two factors. One such causal relationship that is frequently discussed in the media evolves around the question ‘Does playing violent video games make people aggressive?’ (for example, see these articles by CNN, The New York Timesand The Guardian). But when it comes to such topics, how can we actually know whether a causal relationship really exists? Social scientists often run an experiment to find out. However, news articles discussing the results of such experiments usually do not explain what social scientists actually do when they run an experiment.
The typical image of a scientist doing an experiment involves a nerdy-looking person with glasses and a white coat, holding bottles with bright coloured fluids and blowing things up. However, I have done several experiments and I have yet to create an explosion… Let’s go back to our current example: Does playing violent video games cause aggression? In an experiment, researchers compare situations. In this case, the first situation includes participant A playing a violent video game after which we measure her level of aggression. The second situation is identical to first one, except that here, participant A does not play a violent video game. If the participant is more aggressive in the first situation than in the second situation, we can conclude that the video game caused her aggression. We can say so because the playing of a violent video game was the only difference between the situations – everything else was the same. So, the only thing that could have caused the difference in aggression, is the video game. This is (a simplified version of) a typical experiment as it is done in the social sciences.
However, here we face an obvious problem: one and the same person cannot play a violent video game and at the same time not play a violent video game. As a solution to this problem referred to as the fundamental problem of causal inference, we could invite a second participant. One participant will be playing a violent video game, while the other participant does not. However, if we then find differences in the level of aggression between the two participants, we cannot know whether the difference in aggression was caused by the game or by the different personality traits of the participants – perhaps one of the participants was simply a more aggressive person. To factor out the possible influence of individual differences, scientists often invite large groups of people to participate in their experiments.
Of course, the examples given here are simplified experiments and there are additional problems that can arise. For example, there could be other factors that influenced participants’ aggression during the experiment that we overlooked*. Such problems create challenges for social scientists investigating causal relationships. Hence, most experiments do not provide clear cut answers to societal questions. Their results must therefore be interpreted critically and they are usually presented with a high dose of nuance. Although I have no scientific proof for the causal relationship, I dare say that this is one of the factors causing researchers’ answers to questions to often start with the well-known “Well, it depends…”.
*Not to mention the ethical concerns that arise when having minors play violent video games knowing that this may cause them to behave aggressive towards others. However, this topic is beyond the scope of this particular blogpost