Designing Bias-Aware Smart-Home Systems: System-Level Integration of Ethics and Automation Bias Modeling in AI for Early CKD

Authors: Meacham, S. and et, A.

Journal: AI and Society

Publisher: Springer Nature

eISSN: 1435-5655

ISSN: 0951-5666

Abstract:

As artificial intelligence becomes increasingly embedded in smart-home healthcare systems, its potential to support chronic disease management raises pressing questions about automation bias, trust, and the responsible governance of intelligent technologies. This paper presents a bias-aware smart-home system for early monitoring of chronic kidney disease (CKD), integrating wearable IoT sensors with large language model (LLM)-driven clinical decision support. Crucially, we move beyond individual algorithmic fairness to model social and ethical concerns at the system level. The architecture incorporates human-in-the-loop (HITL) oversight, confidence scoring, and demographic fairness evaluators—not as afterthoughts, but as core design elements. Using SYSML-based requirement and behavioral modeling, we illustrate how AI ethics can be embedded by design across workflows and components. This approach demonstrates a scalable pathway toward socially aligned AI systems in healthcare, offering a model of intelligent infrastructure that centers clinical accountability, transparency, and equity from the outset.

Source: Manual