Abstract
The rapid expansion of artificial intelligence, digital trace data, and computational modeling is reshaping how well-being can be measured, understood, and supported. Recent work has already proposed computational positive psychology (CPP) as an emerging direction for advancing positive psychology in the digital era. Building on this development, this article proposes "Computational Well-being Science" (CWS): a broader, theory-guided research direction that uses multimodal digital traces, computational modeling, and multi-source data to understand, measure, predict, and support human well-being across individual, relational, and population levels. I situate CWS within a "three turns" account of positive psychology's evolution--from institutionalization (1999-2010) through reflexive critique (2011-2020) to a computational turn (2021-present)--and define the concept through three foundations: conceptual, methodological, and infrastructural. I then distinguish CWS from eight adjacent fields, including CPP, and outline a four-stage research pipeline: datafication, modeling, interpretation, and intervention. Key challenges include privacy ethics, algorithmic bias, measurement validity, and the risk of theoretical hollowing. The central argument is that CWS should not be understood as a replacement for positive psychology or CPP, but as an extension of the study of human flourishing into a broader computational, data-intensive, and AI-mediated well-being science paradigm.