# Robust language-based mental health assessments in time and space through social media

**Authors:** Siddharth Mangalik, Johannes C. Eichstaedt, Salvatore Giorgi, Jihu Mun, Farhan Ahmed, Gilvir Gill, Adithya V. Ganesan, Shashanka Subrahmanya, Nikita Soni, Sean A. P. Clouston, H. Andrew Schwartz  
**arXiv:** [2302.12952](https://arxiv.org/abs/2302.12952)  
**Date:** 2023-02-25  
**Summary page:** https://www.summarizepaper.com/en/paper/2302.12952v1/

## Key points

- Study focuses on measuring population mental health in the United States
- Aims to provide a more fine-grained assessment compared to existing methods
- Current methods only capture mental health broadly through surveys with limited estimates
- Researchers propose using large-scale analysis of social media data for higher resolution estimates
- Validated approach using 1.2 billion Tweets from 2 million geo-located users
- Specifically focused on estimating mental health changes related to depression and anxiety
- Moderate to large associations found between language-based mental health assessments from social media data and survey scores from Gallup
- Language-based assessment method proved cost-effective and scalable for monitoring population mental health at weekly time scales
- Fine-grained time series data allows for monitoring effects of societal events and policies, as well as enabling quasi-experimental study designs in population health and other disciplines
- Method can be generalized beyond mental health in the U.S. and applied to a broad range of psychological outcomes
- Can facilitate community measurement in under-resourced settings where traditional survey measures may not be available but social media data is accessible

## Summary

This study focuses on the measurement of population mental health in the United States and aims to provide a more fine-grained assessment compared to existing methods. Currently, mental health is only broadly captured through surveys that assess "mentally unhealthy days" or "sadness", and these surveys are limited to infrequent state or metropolitan estimates. To address this limitation, the researchers propose using large-scale analysis of social media data to estimate mental health changes at higher resolutions, such as weekly estimates for counties. To validate their approach, the researchers used a sample of 1.2 billion Tweets from 2 million geo-located users. They specifically focused on estimating mental health changes related to depression and anxiety, which are two leading mental health conditions. The findings showed moderate to large associations between language-based mental health assessments derived from social media data and survey scores from Gallup, a well-known research organization. The language-based assessment method proved to be cost-effective and scalable for monitoring population mental health at weekly time scales. This level of spatially fine-grained time series data allows for monitoring the effects of societal events and policies, as well as enabling quasi-experimental study designs in population health and other disciplines. Importantly, this method can be generalized beyond mental health in the U.S. It can be applied to a broad range of psychological outcomes and can facilitate community measurement in under-resourced settings where traditional survey measures may not be available but social media data is accessible. The study also provides additional context regarding the data used in their analysis.