Dataset on System Performance, Usability, and User Acceptance of Edge-Based Self-Reflective RAG Health Assistant
Year: 2026
Type: dataset
Abstract: This repository contains the empirical evaluation dataset and supplemental codebase for the research paper entitled "Dataset on System Performance, Usability, and User Acceptance of Emi: An Edge-Based Self-Reflective RAG Health Assistant" (submitted to Data in Brief, Elsevier). 1. Overview Emi is an edge-AI conversational agent tailored for Psychological First Aid (PFA) and Non-Communicable Disease (NCD) public health education in Semarang City. The system runs entirely on consumer-grade edge hardware (≤8 GB VRAM) utilizing quantized Gemma 4, Qdrant vector database, and LangGraph orchestration. 2. Dataset Contents 1_Raw_Data_N20.csv: Anonymized survey responses (N=20) evaluating Technology Acceptance Model (TAM) constructs and System Usability Scale (SUS) scores. 2_Outer_Stage1_Comparison.csv: Measurement model construct reliability, Cronbach's Alpha, Composite Reliability, and Average Variance Extracted (AVE). 3_Master_Outer_Loadings.csv: Cross-loadings and indicator factor matrix. 4_Stage2_HOC_Weights_Loadings.csv: Formative and reflective Higher-Order Construct (HOC) parameters. 5_Hypothesis_Bootstrapping.csv: Structural path coefficients, t-statistics, p-values, and mediation analysis results. 6_DSR_Technical_Logs_N115.csv: System latency logs, Time-To-First-Token (TTFT), token throughput (TPS), and RAGAS accuracy evaluations across 115 benchmark queries. 7_BlackBox_Safety_Test.csv: Boundary safety test cases, jailbreak resilience verification, and safety adherence evaluations. 3. Supplemental Codebase emi_core.py: Core LangGraph agent and prompt orchestration pipeline. emi_api.py: FastAPI backend service handling decoupled streaming and PostgreSQL logging. streamlit_emi.py: Frontend user interface with audio synthesis. mem_fill.ipynb: Ingestion pipeline utilizing GOT-OCR 2.0 and Qdrant embedding injection. docker-compose.yml: Container orchestration for PostgreSQL and Qdrant services.
Language: English
Subfield: Health Informatics
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7204007544