The troll lives in everyone: the causes of aggressive behavior in online communities

The proportion of "bad" messages and users on CNN.com is increasing, that is, trolling can spread and grow.

While communication on the Internet has become an integral part of the lives of many people, scientists are increasingly interested in various antisocial behaviors that often manifest themselves on the Web: trolling, insults, and harassment (bullying). Online abuse victims experience emotional stress. In addition, sometimes aggressive behavior from the virtual space goes into the real world.

On many sites, trolling, insults, and bullying make up a significant proportion of comments. For example, on CNN.com, more than 20% of comments are deleted for violating community rules. On Geektimes, deleting a message is an extraordinary event, but it happens.

What is the cause of antisocial behavior of people on the Internet? Some experts have previously expressed the opinion that trolls are people with specific personality traits and special motivation. There is another point of view that aggressive behavior on the Internet is manifested under the influence of the external environment. If so, who provokes it? Special antisocial individuals or ordinary people? Is trolling around it an innate quality, or does such behavior arise spontaneously? What conditions are required to push a person to aggressive communication on the Internet? If trolling is transmitted from person to person, can it spread to the community as an infection?
')
These questions were investigated by scientists from Stanford and Cornell universities. Their work has not yet been published in the scientific press. She will be presented at the ACM conference on February 25, 2017. But the topic is so important and interesting that you can consider this scientific article even before the official publication and independent verification. After examining the causes of trolling and the laws of its distribution in society, we will be able to design more reliable social structures. Create online communities that are protected from aggression and insults.

In their study, the authors consider “trolling behavior” such behavior of a participant in an online discussion that goes beyond the acceptable framework adopted in this community. The scientific work itself describes the conditions of the experiment, as well as the analysis of more than 16 million posts from real discussions on CNN.com. As a platform for the experiment was chosen popular site with the discussion of news.

During the experiment, two possible characteristics of the trolling behavior were analyzed: the user's mood and the context of the discussion (are there any previous trolling messages). It turned out that bad mood and the presence of trolling messages doubled the likelihood of aggression on the part of the average user.

Then there was an analysis of 16 million posts on CNN.com, including millions of those labeled as trolling (violating community rules). It turned out that 25% of the trolling posts were written by people who had not had a single malicious message before. In other words, a very large amount of aggression on the Internet comes from ordinary people or bots.

Scientists tested on real data a hypothesis that a bad mood increases the likelihood of aggression on the Internet and insults other participants in the discussion. Indeed, it turned out that the share of offensive messages on CNN.com correlates with well-known schedules of changes in people's mood during the day . Like a bad mood, trolling indicators grow during the day and reach a peak late in the evening and at night, at the beginning of the work week (Monday) and just before its beginning (Sunday), which is clearly seen in the graphs.



In the right diagram on the y-axis, the probability is postponed that after the first trolling message, the repeated one will also be marked as aggressive. On the x axis, the time interval between the first and second messages is plotted. It is noticeable that the smaller the interval, the higher the likelihood of re-manifestation of aggression. This also indirectly indicates that a bad mood affects the degree of aggressiveness, increasing the likelihood of trolling by a particular user. Over time, the person "cools" and calms down.

Transfer of aggression from one topic to another

Next, scientists tested the hypothesis that a bad mood persists between different discussions, so that they ultimately influence each other. For this, messages from users who have not had a single aggressive post have been studied, but before that they took part in other discussions where there was not a single aggressive post (in blue on the graphs) or such was (orange).



The illustration shows that the likelihood of aggression by a regular user increases in all cases when he expressed his opinion in discussions on another topic, where there were aggressive messages (a), simply participated under the discussion thread on another topic without expressing his opinion (b).

Context influence

The surrounding messages are also very influential. The illustration above shows how the probability of a user’s trolling increases if the whole discussion starts with a “bad” post (s) or a separate thread of the discussion began with it (d).

The illustration below shows that in discussions with at least five posts the probability of a trolling message appears directly depends on the number of previous aggressive messages (a), on the distance to the previous aggressive message (b).



Of course, the topic of discussion itself greatly influences the degree of user aggression. Probably each site has its own “hot” controversial topics, it still depends on the audience. On CNN.com, the greatest aggression of readers is caused by the themes of the USA, show business and justice. The least controversial and aggressive topics are politics, technology and subjective authoring articles expressing personal opinion.

According to the authors, their study refutes the view that Internet trolls are some kind of antisocial individuals who sit in dark rooms and scribble malicious comments on forums and on social networks. Not at all. Aggressive behavior in the discussion is largely due to external factors. Moreover, this phenomenon can be dealt with with the help of competent mechanics in a social network.