The internet has brought change to almost everything in our lives. In particular, the ways we acquire knowledge have significantly changed, partly due to online knowledge repositories such as Wikipedia. In fact, it has even changed the way science is being done. Social scientists increasingly are using online data to study our individual or collective behaviour on a scale and with an accuracy normally only seen in the natural sciences.
Sure, we are still far from having large experimental social science data sets similar to the ones produced in CERN, but at least we have digital observational data like that collected and analysed in observational astrophysics. Millions of people use online tools on a daily basis – for instance Wikipedia is read about 500,000 times per day.
One of the key topics in understanding social behaviour is what scientists call “collective memory”: how members of a social group will remember an event in the past collectively. Even though collective memory is a fundamental concept in sociology, there have been very few empirical studies on the subject, mostly because of a lack of data. Traditionally, scientists who research how the public recalls past events had to spend a lot of time and effort collecting data through interviews and surveys.
In a recent study, published in Science Advances, our team consisting of a sociologist, a computer engineer and two physicists made use of data from Wikipedia, through its publicly available daily statistics of page views of all the articles in the encyclopedia, to study collective memory.
We specifically looked at aircraft incidents in the whole history of aviation (as long as Wikipedia covers). This was because such events are well documented, but also because, unfortunately, there has been a rather large number of such crashes – making the statistical analysis robust.
We divided the events into recent (2008-2016) and past (anything before 2008). Examples of the recent flights are Malaysia Airlines flight 370, Malaysia Airlines flight 17, Air France flight 447 and Germanwings flight 9525. Past crashes include American Airlines Flight 587 and Iran Air flight 655.
We then used statistical methods to measure increased page views for articles on past events a week after a recent event had happened. We called this increase the “attention flow”. We were interested to know if the increase in the attention to the past event has any relationship with the similarity between or the timings of the recent and the past events. We also wanted to know if we can predict the amount of flow of attention to each past event when a new event occurs.
We found that when the Germanwings flight crashed in 2015, people acquired information from Wikipedia about the crash of an American Airlines flight outside New York City in November 2001. In fact, there was a three-fold increase in views on this page in the week after the Germanwings crash.
This seemed to be a pattern. We consistently observed a significant increase in the views of past events as a result of the recent events. On average, past events were viewed 1.4 times more than recent events in the week after they happened. This suggests that the memory of an event can become larger with time – receiving more attention than it originally got. We then tried to model this pattern – taking into account factors such as the impact of the recent and past events, the similarity between the events, and whether there was a hyperlink connecting the two articles directly to each other on Wikipedia.
For instance, in the case of the Germanwings and American Airlines flights, both incidents were related to the role of the pilot, which could be an important coupling factor. The American Airlines plane crashed due to pilot error while the Germanwings pilot intentionally crashed the aircraft. It became more interesting when we observed that there were no hyperlinks connecting these two articles on Wikipedia. Indeed, our general results were robust even when we removed all the pairs that were directly connected to one another by hyperlinks.
The most important factor in memory-triggering patterns was the original impact of the past event, which was measured by its average daily page views before the more recent event occurred. That means that some past events are intrinsically more memorable and our memory of them is more easily triggered. Examples of such events are the crashes related to the 9/11 terrorist attacks.
Time separation between the two events also plays an important role. The closer in time the two events are, the stronger the coupling between them. And when the time separation exceeds 45 years, it becomes very unlikely that the recent event triggers any memory of the past event.
The similarity between the two events turned out to be another important factor. This is illustrated by the memory of Iran Air flight 655, which was shot down by a US navy guided missile in 1988. This was actually not something that people remembered well at all. However, it suddenly got a lot of attention when the Malaysia Airlines 17 flight was hit by a missile over Ukraine in 2014. The Iran Air accident got on average about 500 daily views before the Malaysia Airlines event, but this increased to 120,000 views per day right afterwards
It’s important to note that we don’t really understand the underlying mechanisms behind these observations. The role of the media, individual memory or the structure and categorisation of articles on Wikipedia can all can play a part and will be subject to future research.
More traditional theories suggest that the media plays the central role in shaping our collective memory. However, a big question to ask now is whether the transition to online media and in particular social media will change this mechanism. These days, we often receive news through our Facebook friends, so can this explain why events that have not been in the news for years suddenly become so visible?
Knowing the answers to these questions and understanding how collective memory is being shaped not only is interesting from a scientific perspective, but also could have applications in journalism, media development, policy making, and even advertisement.