EVALUATION DATA ON SYSTEM PERFORMANCE, USABILITY, AND USER ACCEPTANCE OF EDGE-BASED SELF-REFLECTIVE RAG HEALTH ASSISTANT

Deskripsi

Public healthcare services in low-resource environments face significant data privacy risks and cost barriers in deploying cloud-based AI. To address this, Emi an offline, localized generative AI digital health assistant was developed for the Semarang City Health Office (Dinas Kesehatan Kota Semarang) to deliver secure Psychological First Aid (PFA) counseling and Non-Communicable Disease (PTM) health screening and education. This data article describes the technical performance and user usability evaluation datasets compiled during Emi’s development and deployment. The repository contains two primary datasets: technical interaction logs (N=115) mapping query categories (such as greeting, emergency, affective, general, and medical RAG) to system processing metrics (latency and accuracy), and user usability and acceptance survey logs (N=20) evaluating Semarang’s Gen-Z demographic across structural equation modeling (SEM) and System Usability Scale (SUS) constructs. These datasets provide empirical baselines for validating privacy-preserving, localized generative AI in municipal healthcare, serving as rare research data at the intersection of human-computer interaction, edge computing, and digital health informatics.