Beyond the Hype: Risks, Realism, and Responsibilities in Digital Twins for Alzheimer’s Disease

AI-driven digital twins (DTs) promise revolutionary advances in healthcare, yet their application to Alzheimer’s disease (AD) diagnostics and management remains fraught with unresolved technical and ethical dilemmas. The staggering complexity of neurodegenerative processes and intrinsic human variability challenges the very premise of creating clinically viable DTs for AD. Central to this analysis is the concept of “deep learnability”, a concept that questions whether the inherent complexity of AD can be efficiently modeled given biological constraints and data limitations. This theoretical contribution is operationalized through a practical, clinically-focused metric that requires new deep learning models to significantly outperform the best interpretable alternatives to justify their use. This work also advocates for a mandatory three-tier framework for any AD DT: (1) technical validation ensuring computational robustness; (2) biological validation confirming alignment with established biomarker and imaging data; and (3) clinical validation demonstrating tangible utility in improving patient diagnosis and outcomes. This structured approach, combined with a call for global standardization and ethical design, is essential to translate the promise of DTs into clinically viable tools.