A Guide to Constructing Indigenous Statistical Spaces for Prevention Science Research.

Abstract

Artificial intelligence (AI)-powered computational methods, such as machine learning and natural language processing, are increasingly applied in deaths of despair research among Indigenous populations. However, their application in Indigenous contexts is often constrained by epistemological misalignment, technical limitations, and ethical concerns. Integrating Indigenous Research Methodologies into AI-powered prevention science research is necessary to support Indigenous Data Sovereignty and address deaths of despair. The Indigenous Computational Approach (ICA) provides a structured reflexive protocol for constructing Indigenous Statistical Spaces that operationalize Indigenous Research Methodologies within computational workflows. ICA aligns four interdependent components: Researcher Standpoint, Indigenous Theoretical Frameworks, AI Data Analysis Technique, and Dissemination and Indigenous Governance. This protocol is supported by operational steps and an accompanying ICA Checklist. A previously published case study on the Indigenous Wholistic Factors Project illustrates the ICA in practice in the context of suicide risk modeling. The case study applied a lasso logistic regression model to structure feature selection on an Indigenous subsample of the 2019-2020 California Healthy Kids Survey (n = 2609). Ten of 17 candidate features were retained, and the model demonstrated strong discrimination (AUC = 0.87) and acceptable calibration (Brier score = 0.10). The ICA does not guarantee different empirical findings or superior model accuracy, but rather it restructures how AI models are designed, validated, and deployed for prevention science research. The ICA provides a replicable protocol for AI-powered prevention science research to support Indigenous self-determination and community-defined well-being.