Are collective emotions related to the social resilience of a society? We addressed that question using Twitter data from France in relation to the terrorist attacks of Paris in November 2015.
We analyzed more than 17 million original tweets produced by more than 60 thousand French Twitter users. The daily number of tweets in our dataset looks like this:
The red line marks the day of the attacks, when French Twitter users posted more than double the typical amount of tweets of a normal day.
We analyzed the content of tweets, measuring the frequency of words in various dictionaries. The daily scores of positive and negative affect terms look like this:
There was a strong reaction of more negative affect and a soft decrease in positive affect. The slower decay of negative affect shows the existence of a collective emotion. This negative emotion can be analyzed in terms of anxiety, sadness, and anger:
The fastest and strongest response was anxiety and sadness peaked the day after the attacks. They clearly show that the negative emotional response did not instantly disappear, but it lasted for days after the attacks.
We measured the same scores for three types of words related to social resilience: social process terms (talking about other people), prosocial behavior terms (talking about helping others), and shared value terms (liberté, egalité, fraternité):
All three of them show a strong increase in the day after the attacks that decreases very slowly towards the baseline. We can see that these signals of social resilience react a day later than collective emotions and last longer.
We analyzed the emotions of users during the two weeks after the attacks and divided them in the users involved in the collective emotion (high emotional synchronization) and the ones not involved (low emotional synchronization). Their monthly average frequency of social process terms looks like this:
The two groups are barely distinguishable before the attacks. For months after the attacks, the high emotional synchronization group shows clearly higher frequency of social process terms.
We can measure the difference between groups and plot it over time for social process terms, prosocial terms, and shared values terms:
Individuals involved in the collective emotion used all three classes of terms more often than the rest. This lasted for many months after the attack.
These results suggest that French Twitter users shared a negative collective emotion after the terrorist attacks, and that the experience of that emotion led to higher frequency of language related to social resilience.