A couple months ago, a research study looked at comments made in public discussion forums on the BBC website. It claims that negative emotions are what motivates people to share their opinion online.
The research team, a collaboration between Warsaw University of Technology and the University of Wolverhampton, analyzed the number and sentiment of posts written by individuals contributing to the BBC online forums. They collected almost 2.5 million posts from the Religion & Ethics, World News, or UK News message boards, spanning more than 4 years (2005-2009) and authored by over 18,000 people. The data set includes posts from almost 100,000 separate threads.
The first finding is not surprising: Scale-free distributions occurred both in each thread as well as overall activity. As the threads grew longer, though, the number of unique users decreased. In other words, longer threads were the result of a prolonged exchange between a few people, not an indication of a broader appeal. This makes sense when considering the “80-20 rule” that says most of the work is done by a few operators.
The second major insight from this study is more interesting. The sentiment analysis showed that most posts were emotionally negative. The same is true of the most active users in a given thread. Put together, threads are sustained primarily through negative discussions between a few participants.
A couple new questions arise:
How necessary is negativity to generate a lot of visible forum activity?—These forums are more likely to produce polarizing topics (e.g., religious beliefs) than, say, Flickr photos or social media conversations, so this shouldn’t be taken as a universal behavior of online discourse. If there is comparable activity in other forums, do they also reflect massive negativity? Even within the researchers’ graphs, there seems to be a point at around 100 comments where the emotion spreads. The fitted line extends the scale-free properties, but it would be interesting to look at 2-3 different trajectories that arise at that condition, like a triple point in phase transitions, and examine the differences between those states.
What is the relationship between bursts of activity and negativity?—The selected forums were populated by members who collectively with an average post count of 137. That translates to about two comments every three weeks. The reality is that active members come and go, entering and exiting periods of heavy activity. I would love to know the distribution of individual burstiness and how that relates to emotional content.