Oluwadara Adedeji

Ethical governance of AI-driven systems is often expressed through high-level principles and static documentation, creating a gap between regulatory requirements and system-level verification. This challenge is particularly acute in digital phenotyping, where continuous behavioural data raises concerns around consent, privacy, and fairness. In this paper, we propose a computational ethical framework for AI-driven digital phenotyping system in which ethical requirements are formalised as deontic temporal logic constraints, alongside a conceptual ethical agent that oversees the system and ensures that any supervised system satisfies the specified constraints. Using a case study involving financial data and mental health, we model key ethical properties and verify them using the Z3 Satisfiability Modulo Theories (SMT) solver. Our evaluation shows that the framework is logically consistent and that violations of the specified ethical properties are ruled out within the formal model through counterexample-based verification. This presents early research enabling continuous, machine-verifiable ethical checking, moving beyond retrospective compliance based on static documentation. We discuss limitations, including the need for real-world verification with data, the challenge with subjectivity and contextual sensitivity, the need for human oversight, and outline how such approaches can support the development of AI systems with continuous and auditable ethical guarantees.

The relationship between financial circumstances and mental health is well-established. New financial data sources such as bank transactions and digital payments offer new opportunities to better characterise this relationship. However, the prior financial data sources and analytical approaches have not been systematically reviewed. Understanding these is essential to guide future research and inform integrated interventions that address both financial and mental health outcomes. This scoping review systematically maps the research on money and mental health, examining: (1) the financial variables and data sources used, (2) modeling methods employed, and (3) methodological gaps that novel objective data might address. We systematically searched PubMed, PsycINFO, IEEE Xplore, ACM Digital Library, and Scopus. Papers were screened against predefined inclusion/exclusion criteria, and data extracted using a standardized spreadsheet. Analysis employed deductive coding guided by our research questions, refined iteratively through engagement with the data. PRISMA Extension for Scoping Reviews (PRISMA-ScR) was followed for reporting. Of the 43 included studies, most (n =34, 79%) examined mental health in connection with financial factors such as financial difficulty or financial strain, while a small number focused on predictive modeling with financial behavioral data (n=5, 12%), macroeconomic indicators (n=2, 5%), or intergenerational support between parent and child (n=2, 5%). Depression was the most common outcome (n=24, 56%), followed by anxiety, psychological distress, and bipolar disorder. Statistical methods dominated (77%), with 19% employing machine learning or deep learning. Ground truths relied predominantly on self-reported questionnaires—only four studies used objective financial data (three gambling records, one bank transaction). This review reinforces the complex, bidirectional relationship between financial circumstances and mental health. Most studies examined how financial difficulty affects mental health, while only a few explored how mental illness influences financial behavior, indicating a clear research gap. There is substantial opportunity to use objective financial data and more diverse analytical methods, particularly machine learning, to deepen understanding of the relationship and interactions between money and mental health and inform targeted interventions.

Mood and money are closely intertwined in bipolar disorder (BD), yet it remains unclear whether financial behaviour provides reliable mood-contingent signals, and it is largely absent from digital phenotyping research. Advances in open banking enable objective, passive measurement of financial behaviour at scale. We conducted semi-structured interviews with 19 (11 UK, 4 Ireland, 4 Nigeria) adults with BD to examine how mood manifests in financial behaviour, identify potential financial mood markers, and explore data-sharing preferences. Participants reported distinct mood-related changes in spending and reciprocal effects of money on mood. Three themes emerged: (i) symptomatic drivers shaped by credit access, seasonality, and social context; (ii) candidate markers such as transaction frequency, timing, category, and amount; and (iii) conditional willingness by most participants to share data influenced by privacy, anonymity, and granularity. High-mood spending reflected goal pursuit, generosity, and insomnia, while depression involved both reduced spending and retail therapy, challenging simple bidirectional assumptions. These findings provide empirical insight into financial data as a potential mood marker and inform future research and privacy-sensitive data collection.