Let the data talk

Exploratory research is a preliminary investigation aimed at clarifying the nature of the problem to be solved. By the end of such research, the experts formulate hypotheses based on the data collected, which can be proved or disproved in subsequent confirmatory research. In his article (Let the data talk: embrace exploratory research), Balázs Aczél stresses the importance of creating a research tradition in the social sciences where exploratory and confirmatory studies are balanced. 

The author uses an example to demonstrate the difference between the two research methods. Imagine a research group that wants to study the relationship between aggressive behaviour and the consumption of violent films among children. If this research group were to conduct an exploratory study, they would ask parents questions about their children’s individual characteristics, the family’s home habits, and the kinds of film they watch. Once they have spoken to enough parents, the researchers would inspect the data collected to try to discover patterns. Taking the theoretical background into account, they would then make concrete predictions: for example, if parents talk to their children about the violence they see in films, young people will show less aggressive behaviour. Confirmatory research seeks to support or challenge such predictions with fresh data.

A common – and erroneous – assumption in the social sciences is that exploratory research is “less scientific”. As Balázs Aczél writes in his article, this was also suggested to him during his studies, namely that the main task of researchers is to test, and not generate, hypotheses. However, in the course of his work, he often found that he had not done enough exploratory work to be able to carry out a confirmatory study with complete certainty. Indeed, before designing a study, it is essential to know which of the many possible variables should be included in the analysis. For example, if we are looking at childhood aggression, should we measure its relationship with age, gender, or social status?

“When we are unsure about what we are after, we should stop and explore,” the author suggests. “Instead of committing ourselves to arbitrary specifications, we could let the data talk first.”