Abstract—AI-powered public health kiosks have the potential to transform healthcare delivery by providing personalized Over-The-Counter (OTC) medication guidance. This study focuses on implementing a context-aware kiosk that offers tailored recommendations based on user-reported symptoms, allergies, and medication history, without the need for full clinical records. The system aims to enhance accessibility and provide practical self-care advice in public settings. The AI-powered kiosk integrates patient-specific factors such as allergies, age, and Drug-Drug Interactions (DDIs) to offer personalized OTC medication guidance. Built on the enhanced GAMENet architecture, incorporating Graph Attention Networks (GAT) and Multi-Head Cross-Attention (MHCA), the system surpasses traditional symptom checkers. The system was trained using the MIMIC-III dataset (6350 patients) and DDI data from the TWOSIDES database, with external validation performed on the MIMIC-IV dataset (9036 patients). Both datasets were divided into training (75%), validation (8.3%), and testing (16.7%) subsets. Real-time safety checks used Anatomical Therapeutic Chemical (ATC) codes. The kiosk interface includes multilingual support, large fonts, voice commands, and Braille. Usability and acceptability were assessed through a survey of 33 healthcare professionals. Preliminary results from the MIMIC-III dataset show that the enhanced GAMENet model achieved a Precision-Recall Area Under the Curve (PRAUC) of approximately 0.74. Validation with the MIMIC-IV dataset further demonstrated the model’s improved performance, highlighting its ability to deliver accurate and secure healthcare recommendations. This study demonstrates the potential of AI-powered kiosks to provide safe and personalized OTC medication guidance. Future work will focus on deployment, usability studies, and scalability for broader adoption.
 
Keywords—public health kiosk, AI and federated learning, personalized healthcare, Over-The-Counter (OTC) medication recommendation, graph neural networks


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).