Changes in Psycholinguistic Attributes of Social Media Users Before, During, and After Self-Reported Influenza Symptoms

Lucie Flekova, Vasileios Lampos, Ingemar J. CoxSMM4H @ EMNLP 2018 · OpenReview

TL;DR

This study explores how people's language on Twitter changes when they experience influenza-like illness (ILI). By analyzing tweets from before, during, and after reported symptoms, the researchers found shifts in social activity and cognitive processes, suggesting a link between language patterns and immune status.

What Problem Does This Paper Address?

There's a known connection between psychological and social factors and physical health, including how stress can impact our immune system. Simultaneously, researchers have demonstrated that language used on social media can reveal a lot about an individual's psychological state, personality, and even health. This paper bridges these two areas by investigating whether changes in an individual's language on Twitter can be observed around the time they report experiencing influenza-like illness (ILI). The core question is: does our online communication style change when we're sick, and if so, what does that tell us about our psychological and social state during illness and recovery?

Methods

The researchers employed a linguistic analysis approach using the Linguistic Inquiry and Word Count (LIWC) tool. LIWC categorizes words into various psychological and social dimensions, such as emotional processes, cognitive processes, personal concerns, and social activities. They analyzed tweets from users who self-reported ILI symptoms, comparing their language in three distinct periods: 7 days before the reported symptom, the period around the reported symptom (defined as one day before to two days after the self-report), and 14 days after the reported symptom. For each user and time period, they calculated the relative occurrence of words within specific LIWC categories. Statistical analysis was then used to determine the significance of differences in these linguistic features between the "before" and "during" periods, and between the "during" and "after" periods.

Datasets

The study utilized a dataset of approximately 14 million UK tweets collected between 2014 and 2016. From this, they identified around 2,600 tweets potentially indicating flu symptoms, which were then manually filtered to obtain 1,235 tweets from 285 unique users who self-reported ILI. These users were found to be prolific on Twitter, generating a substantial volume of tweets. The analysis focused on a specific time window around the self-reported illness event.

Key Insights and Ideas

  • Language as a Health Indicator: The study reinforces the idea that our online language is a sensitive indicator of our internal state, extending this to physical health events like ILI.
  • Post-Illness Social Boost: A notable finding is the increase in social activity and cognitive processes after the ILI peak. This suggests that recovery might involve a re-engagement with social networks and a more active cognitive state.
  • Coping Mechanisms in Language: The increased use of "causal" words (e.g., "reason," "why," "because") after ILI is interesting. This has been linked to improved coping skills, suggesting that individuals may be more actively trying to understand or explain their experience as they recover.
  • Reduced Anxiety Post-Illness: A decrease in the use of impersonal pronouns after ILI could indicate a reduction in anxiety, as higher usage of these pronouns has been previously associated with increased anxiety levels.
  • Focus on Others During Illness: Interestingly, during the illness itself, there was an increase in the use of second-person pronouns ("you"), suggesting a potential shift in focus towards others rather than oneself.

Key Results

The study observed statistically significant changes in several linguistic features around the time of self-reported ILI. Specifically:

  • Increased Social Activity: The relative frequency of words from "Friends" and "Family" LIWC categories increased after the ILI peak.
  • Increased Cognitive Processes: An increase in the usage of "causal" words was observed post-ILI.
  • Decreased Impersonal Pronouns: A significant decrease in impersonal pronouns was noted after the ILI peak.
  • Increased Second-Person Pronouns: A notable increase in the use of "you" was observed during the illness period.
  • No significant differences were found in overall emotion levels or the usage of the first-person pronoun "I."

Limitations and Future Work

The authors acknowledge that this is an initial exploratory study. One limitation is the definition of the "ILI peak period," which was based on an assumption and manual exploration. Future work includes extending the study by analyzing a larger dataset of ILI events (an additional 11,000 ILI events are planned) to further validate these findings and explore more nuanced linguistic patterns associated with illness and recovery. They also aim to investigate how these linguistic changes correlate with objective health data.

Summary updated

Abstract

Previous research has linked psychological and social variables to physical health. At the same time, psychological and social variables have been successfully predicted from the language used by individuals in social media. In this paper, we conduct an initial exploratory study linking these two areas. Using the social media platform of Twitter, we identify users self-reporting symptoms that are descriptive of influenza-like illness (ILI). We analyze the tweets of those users in the periods before, during, and after the reported symptoms, exploring emotional, cognitive, and structural components of language. We observe a post-ILI increase in social activity and cognitive processes, possibly supporting previous offline findings linking more active social activities and stronger cognitive coping skills to a better immune status.
BibTeX
@inproceedings{flekova2018changes,
  author = {Lucie Flekova and Vasileios Lampos and Ingemar J. Cox},
  title = {{Changes in Psycholinguistic Attributes of Social Media Users Before, During, and After Self-Reported Influenza Symptoms}},
  booktitle = {SMM4H @ EMNLP},
  year = {2018},
  url = {https://openreview.net/forum?id=WEPO8rnwfy},
}
Figure