Language Style Similarity and Friendship Networks
Human communication has shifted to texts from phone calls - in the U.S. the number of text messages exploded from 14 billion in 2000 to 188 billion in 2010. Prof. Balázs Kovács from Yale University gave a lecture at CNS about the role of language style in human interactions.
It is known that homophily, the tendency that friendship is more likely to form between similar people, shapes social networks. Prof. Kovács demonstrated in two examples how linguistic similarity enforces network tie formation and friends exhibit linguistic convergence over time.
Linguistic style is not conscious, it’s elements are learnt in early childhood, and do not change that much over time. Style does not mean what we write about, it means how we write. It has several aspects, such as the ratio of nouns and verbs in sentences, how much someone likes to talk about the future or the past, or how much the one talks about her/himself or others.
In the first study, they analyzed the emergent friendship network of 285 graduate students from an East Coast university. In the second study, they collected 1.7 million reviews from 160 000 users and their friendship data from Yelp.com, an online reviewing, and social media platform. They used the LIWC computational linguistic model to characterize linguistic style from reviews and students’ cover letters for admission and tests.
In both studies, they had two data collection points, allowing the team to compare linguistic style and friendship and their relationship over time.
After controlling for all socio-demographic aspects of homophily (age, race, gender) the results showed in both cases that linguistic style similarity is a strong predictor of future friendships, and linguistic style becomes more similar over time.
The authors also run a simulation using the dual engines of tie homophily - selection and convergence – which has fundamental effects on network fragmentation. Due to recent years’ scandals about opinion bubbles, it is important to note that these processes create the basis of structural echo chambers which radically reduce opinion pluralism.
Blog post by Orsolya Vásárhelyi