The value of social media language for the assessment of wellbeing: A systematic review and meta-analysis

Abstract

ABSTRACT Wellbeing is predominantly measured through self-reports, which is time-consuming and costly. It can also be measured by automatically analysing language expressed on social media platforms, through social media text mining (SMTM). We present a systematic review based on 45 studies, and a meta-analysis of 32 convergent validities from 18 studies reporting correlations between SMTM and survey-based wellbeing. We find that (1) studies were mostly limited to the English language, (2) Twitter was predominantly used for data collection, (3) word-level and data-driven methods were similarly prominent, and (4) life satisfaction was the most common outcome studied. We found that SMTM-based estimates of wellbeing correlated with survey-reported scores across studies at a meta-analytic average of r = .33(95% CI [.25, .40]) for individual-level assessments of wellbeing, and at r = .54(95% CI [.37, .67]) for regional measures of well-being. We provide recommendations for future SMTM wellbeing studies.