World Happiness Report 2023 134 Through weighting, (c) selection biases are addressed. Lastly, through tracking within-user changes in Gen 3, (d) social media estimates can yield stable longitudinal estimates beyond cross-sectional analyses, and (e) provide more nuanced methodological design control (such as through difference-in-difference or instrumental variable designs). Taken together, social media-based measurement of well-being has come a long way. Around 2010, it started as technological demonstrations that applied simple dictionaries (designed for different applications) to noisy and unstabilized random feeds of Twitter data yielding unreliable time series estimates. With the evolution across generations of data aggregation and levels of language models, current state-of-the-art methods produce robust cross-sectional regional estimates of well-being.6 They are just maturing to the point of producing stable longitudinal estimates that allow for the detection of meaningful changes in well-being and mental health of countries, regions, and cities. A lot of the initial development of these methods has taken place in the U.S., mainly because most well-being survey data for training and benchmarking of the models have been collected there. However, with the maturation of the methods and reproduction of the findings by multiple labs, the approach is ready to be implemented in different countries around the world, as showcased by the Instituto Nacional de Estadística y Geografía (INEGI) of Mexico building a first such prototype.7 The Biggest Dataset in Human History The need for timely well-being measurement To achieve high-level policy goals, such as the promotion of well-being as proposed in the Sustainable Development Goals,8 policymakers need to be able to evaluate the effectiveness of different implementations across private and public sector institutions and organizations. For that, “everyone in the world should be represented in up-to-date and timely data that can be used to
RkJQdWJsaXNoZXIy NzQwMjQ=