World Happiness Report 2023 135 measure progress and make decisions to improve people’s lives.”9 Specifically, ongoing data about people’s well-being can help to evaluate policy, provide accountability, and help close feedback loops about what works and what does not. For such ongoing evaluation, well-being estimates are needed at higher than annual and national levels of temporal and geographic aggregation. Particularly with an eye towards under-resourced contexts and developing economies, it would be ideal if such estimates could be derived unobtrusively and cost-effectively by analyzing digital traces that populations naturally produce on social media. The potential of social media data for population health and well-being As perhaps the most prominent of such data sources, social media data has become the largest cross-sectional and longitudinal dataset on human emotions, cognitions, behaviors, and health in human history.10 Social media platforms are widely used across the globe. In a survey conducted in 11 emerging economies and developing countries across a wide range of global regions (e.g., Venezuela, Kenya, India, Lebanon), social media platforms (such as Facebook) and messaging apps (such as WhatsApp) were found to be widely used. Across studied countries, a median of 64% of surveyed adults report currently using at least one social media platform or messaging app, ranging from 31 % (India) to 85% (Lebanon).11 Over the last decade, a body of research has developed – spanning computational linguistics, computer science, the social sciences, public health, and medicine – that mines social media to understand human health, progress, and well-being. For example, social media has been used to measure mental health, including depression,12 health behaviors, including excessive alcohol use,13 more general public health ailments (e.g., allergies and insomnia),14 communicable diseases, including the flu15 and H1N1 influenza,16 as well as the risk for non-communicable diseases,17 including heart disease mortality.18 The measurement of different well-being components Well-being is widely understood to have multiple components, including evaluative (life satisfaction), affective (positive and negative emotion), and eudaimonic components (purpose; OECD, 2013). Existing methods in the social sciences and in Natural Language Processing have been particularly well-suited to measuring the affective/ emotional component of well-being. Namely, in psychology, positive and negative emotion dictionaries are available, such as those provided by the widely-used Linguistic Inquiry and Word Count (LIWC) software.19 In Natural Language Processing, “sentiment analysis”, which aims to measure the overall affect/sentiment of texts, is widely studied by different research groups that routinely compare the performance of sentiment prediction systems on “shared tasks.”20 As a result, social media data has typically been analyzed with emotion dictionaries and sentiment analysis to derive estimates of well-being. In reviewing the early work of well-being estimates from social media, these affect-focused analyses in combination with simple random Twitter sampling techniques, led some scholars to conclude that well-being estimates “provide satisfactory accuracy for emotional experiences, but not yet for life satisfaction.”21 Other researchers recently reviewed studies using social media language to assess well-being.22 Of 45 studies, six used social media to estimate the aggregated well-being of geographies, and all of them relied on Twitter data and on emotional and sentiment dictionaries to derive their estimates. Over the last decade, a body of research has developed – spanning computational linguistics, computer science, the social sciences, public health, and medicine – that mines social media to understand human health, progress, and well-being.
RkJQdWJsaXNoZXIy NzQwMjQ=