Quantifying the Digital Phenotype of Loneliness on Twitter
Skip to main content
eScholarship
Open Access Publications from the University of California

Quantifying the Digital Phenotype of Loneliness on Twitter

Abstract

Social media promotes social connectedness, but social media users can still be lonely which is an important preceding condition to various mental health disorders such as anxiety and depression. Here we aim to describe online loneliness in individuals from the linguistic and social features of their platform use. We define a sample of Twitter users who explicitly report being lonely and compare their language to a matching random control sample. For each user, we create a text embedding - a numerical representation of the content of their online posts, excluding terms and expressions related to loneliness. We utilize principal component analysis on the resulting embeddings to condense the data into a smaller number of variables, while still retaining the majority of the variance. By doing so, we are able to position each user within a two-dimensional space, defined by the first two principal components, which capture the most significant amount of variation in the data. Lonely individuals are spatially separated from the control sample, indicating that lonely individuals exhibit distinct language patterns that is often self-referential, e.g. “I should” and “but I”. Indicators of online social relations, such as the number of online friends, favorites, mentions, show that lonely individuals have fewer social relations, while a sentiment analysis demonstrates that their posts have lower valence. Our results provide insights into the lexical, social, and affective markers that characterize loneliness online, providing a starting point for the development of diagnostics and prevention.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View