FinHealth is my PhD investigation into whether everyday financial behaviour can serve as a passive signal for mood states in bipolar disorder. The schematic below is the conceptual frame the investigation is built around: three mood states — Depression, Balanced, and Hypomanic / manic — mapped to symptom-based features that are each grounded in observable financial behaviours.
The diagram shows a non-exhaustive slice of the feature set: Generosity (gifts, donations to charity, transfers), Goal pursuit (multiple purchases of the same item), Interest expression (purchases in the same category), Return pursuit (risky investments, gambling, cryptocurrencies), and Social performance (purchases in particular categories). More symptom-based features are in scope beyond the five shown here.
Alongside the modelling work, I am developing the data collection pipeline that feeds it — a Django application wired to the Plaid API for open banking transactions. Folding self-report and clinical scale instruments into the same system, so behavioural signal and ground truth land together, is planned as future work.