The world’s oldest living culture leads the way with interpretable generative AI: Aboriginal women shaping the future of perinatal mental health services

Guanjin Wang, Hachem Bennamoun, Wai Hang Kwok, Rehan Raza, Ferdous Sohel, Bridgette Kelly, Trish Ratajczak, Corinne Reid, Rhonda Marriott, Roz Walker, Jayne Kotz and Tracy Reibel

PLOS Digital Health, 2026, vol. 5, issue 10, 1-23

Abstract: Despite their strengths, Aboriginal mothers experience significant health disparities, including perinatal mental health challenges influenced by social, historical, and cultural determinants. Routine risk screening informs follow-up care, yet conventional approaches often fail to address cultural considerations. The Baby Coming You Ready (BCYR) program provides a strengths-based, culturally adapted digital assessment designed to enhance perinatal mental health care. However, some clinicians continue to rely primarily on conventional risk scores, which can neglect broader contextual circumstances influencing Aboriginal mothers, including cultural strengths and protective factors, thereby limiting the program’s full impact. This exploratory model-development study examined whether generative AI data augmentation combined with an intrinsically interpretable machine-learning model could support culturally contextualised prediction of perinatal psychological distress among Aboriginal mothers. We analysed deidentified BCYR pilot data from 293 Aboriginal mothers across six Western Australian sites. After excluding five records with incomplete outcome data, the analytic dataset included 288 participants. Models were evaluated using repeated internal train-test splits, with synthetic data used only for training augmentation and performance assessed on held-out real test data. The highest internal performance was observed for the TVAE-augmented Explainable Boosting Machine at an augmentation ratio of 1.0, with an accuracy of 0.867, an F1 score of 0.817, and an AUC of 0.934. Model explanations identified self-blame, loneliness, family pride, day-to-day management, sleep, and related contextual factors as influential contributors to model-estimated risk. These explanations are intended as prompts for culturally safe discussion and clinical reflection. Findings suggest that interpretable generative AI may have potential to support culturally responsive perinatal mental health assessment. However, broader community-governed external validation and prospective clinical workflow evaluation are required before clinical implementation.Author summary: Perinatal mental health care for Aboriginal mothers should recognise their strengths, cultural context, family relationships, and lived experiences. Standard screening tools may overlook these important factors and may not support culturally safe conversations between Aboriginal mothers and health professionals. The Baby Coming You Ready program was developed to provide a strengths-based, culturally adapted digital assessment for Aboriginal families. However, the contextual and strengths-based information captured through BCYR may not always be fully incorporated into clinical interpretation. This exploratory model-development study examined whether generative and interpretable AI could support culturally contextualised prediction of perinatal psychological distress among Aboriginal mothers. We tested whether carefully augmented training data could support model development while retaining real participant data as the analysis foundation. We also used an intrinsically interpretable model so that the contributing factors could be examined. The best-performing model identified factors such as self-blame, loneliness, family pride, day-to-day management, sleep, and related contextual factors as important contributors to model-estimated psychological distress. The model explanations are intended as prompts for culturally safe discussion and clinical reflection, rather than to replace clinical judgement. However, this study remains exploratory, and further Aboriginal-led co-design, external validation, fairness assessment, and real-world evaluation are needed before clinical implementation.

Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001709 (text/html)
https://journals.plos.org/digitalhealth/article/fi ... 01709&type=printable (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:plo:pdig00:0001709

DOI: 10.1371/journal.pdig.0001709

Access Statistics for this article

More articles in PLOS Digital Health from Public Library of Science
Bibliographic data for series maintained by digitalhealth ().