Feasibility of a generative AI chatbot to support breastfeeding in the Brazilian Unified National Health System.

Abstract

This exploratory feasibility study aimed to evaluate the usability, user experience, and preliminary impacts of a generative language model-based chatbot (Lhia 2.0) on maternal self-efficacy and attitudes toward breastfeeding. A total of 13 mothers receiving care at the Human Milk Bank of the University Hospital of the Federal University of Maranhão interacted with the chatbot via WhatsApp over 15 days. The following instruments were applied before and after the use of Lhia 2.0: the Breastfeeding Self-Efficacy Scale - Short Form (BSES-SF) and the Iowa Infant Feeding Attitude Scale (IIFAS). The User Experience Questionnaire - Short Version (UEQ-S) was applied after the interaction. Results indicated excellent usability (pragmatic quality = +2.36; hedonic quality = +2.57) and high engagement (average interaction time = 92 minutes). Statistically significant improvements were observed in BSES-SF (mean difference = 3.00; p = 0.042; Cohen's d = 0.778) and in IIFAS (mean difference = 3.50; p = 0.021; Cohen's d = 0.778). The results indicated high acceptance among mothers, with notable user engagement, excellent perceived usability, and positive effects on confidence and attitudes toward breastfeeding. Based on these findings, Lhia 2.0 is considered a feasible and promising tool for integration into breastfeeding promotion strategies, particularly within the context of the Brazilian Unified National Health System.