Janush Holyst: "Emotions Are the Main Engines for Discussions in a Cyber Community."

Emotions are an integral part of our lives. Currently we spend more and more time on the world wide web. So how do we transfer our emotions from the offline to the online reality? How do collective emotions affect a user’s activity? Janush Holyst, professor at the Warsaw University of Technology and visiting professor at ITMO’s High-Performance Computing Department, talked about it in his lecture on collective emotions in social networks.

There is no general insight into the definition of emotions. The only thing we know is that they affect our lives. In different countries people describe this phenomenon in various ways. For example, the Japanese have an emotion called "tokimak," which appears when one falls in love or is excited by something. It is hard to explain it to people from other countries.

From the biological point of view, we need emotions to react quickly. For instance, fear is necessary to make us run away in case of danger as opposed to stay thinking what to do. Apart from personal emotions we are interested in collective emotions. The reason why people visit sport competitions is more about a collective feeling you get when you are together with other members of a social group rather than watching the actual sport game.

Unfortunately sociologists continue to argue about what basic discrete emotions are. This is a real problem, as a physicist, I got used to knowing the composition of a matter. But speaking of emotions, scientists cannot determine fundamental phenomena. There are several classifications of emotions, one of them was made by Robert Plutchik in 1980. The researcher divided emotions into opposite pairs like pleasure-sadness, trust-disgust, fear-anger, astonishment- anticipation.

One of the simplest ways of defining emotions was offered by James Russell. According to the "circumplex model", each emotion is a linear combination of valence and arousal. It means that each emotion is marked as positive or negative and can be strong or weak.

Furthermore, there is no common understanding of the duration of emotions. Most scientists think that it can last from one second to several minutes. If you experience an emotion much longer it means that it is caused by some physiological or social phenomena, for instance, depression.

Emotions influence not only our conscience, but also our bodies and brains. Thanks to its' peculiarity we can detect emotions. The easiest way is to analyse behavior of face muscles. By observing activity of forehead muscles and deformation of lips we can understand whether someone feels good or bad. The second way is to measure the rate of skin cover conductivity. Affected by various emotions a body produces a different amount of sweat. Furthermore the ECG rate (electrocardiogram) changes depending on emotions.

Currently we spend more and more time using computers. It makes us a part of a new social-technological system. So the question is, can we exchange emotions using information services? Is it possible to detect emotions in cyberspace? Are these collective emotions equivalent to emotions in real life?

One of the best tools that may help to answer these questions is the SentiStrenght program. It estimates one’s mood using the following marks: from -1 to -5 to grade negative emotions and from 1 to 5 to grade positive ones. But we face a problem with classifying. For instance, a phrase "I missed you" includes both positive and negative aspects. On the one hand, an addresser is kind to an addressee, but, on the other hand, it means that the addresser feels sad due to the fact that he didn’t communicate with the addressee for a while. Only one word "miss" expresses two emotions, let alone more complex examples like, for instance, "this movie was wonderful, its actors and actresses are famous, soundtracks were excellent. However I wish I had stayed home." This phrase is mostly positive but the last sentence shows how I feel about this and reverses the meaning completely.

We used this program to detect collective emotions in the internet using several platforms. We have analyzed more than 2.4 million messages at BBC forum, 1.6 million at the Digg and 242 thousand ones at the Blog06. We marked them with -1, 0 or 1. We found our emotional clusters that are chains of messages containing the same emotions. We also discovered that if a discussion includes several comments containing one common emotion, the next ones will also be the same. To get more precise information we randomized data and shortened the amount of clusters.

In real life, emotions help us to avoid dangerous situations. But why do we need them while being online? Using BBC forum we found out that the emotional beginning of a discussion (no matter whether it is positive or negative) affects a user’s activity and makes the discussion longer. It lets us think that emotions are the main engines for discussions in a cyber community.

In his lecture Mr. Holyst presented the results of the research made by the Cyberemotions consortium. The project includes social technological apps, emotional avatars and analyses of comments on a website of online stores.

Archive by year: