HumAIne-Chatbot: Real-Time Personalized Conversational AI via Reinforcement Learning for Public-Interest Digital Services
Journal of Artificial Intelligence for Sustainable Development · 2026
DOI: 10.69828/4d4kwm ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Artificial Intelligence for Sustainable Development, 2026. Verfasst von Georgios Makridis und 6 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 41 Literaturangaben.
Abstract (Originalsprache)
As conversational AI systems become embedded in public services and civic infrastructures, ensuring their democratic alignment and public-interest orientation is critical [1, 2]. Current chatbots often provide generic responses that fail to adapt to diverse user needs, potentially reinforcing information inequalities and undermining trust in digital public services [2]. This paper introduces HumAIne-chatbot, an AI-driven conversational agent that personalizes interactions through adaptive user profiling while explicitly addressing concerns about fairness, accessibility, and democratic governance. The system combines implicit behavioral signals (typing speed, sentiment, language complexity) with explicit feedback to build user profiles via reinforcement learning, then dynamically adjusts response complexity and style [3]. To evaluate the system’s effectiveness and detect potential biases before real-world deployment, we employ a novel virtual-personamethodology that simulates 50 diverse user archetypes across demographics and expertise levels [4]. Controlled experiments across public-interest domains (health information, education, environmental sustainability) demonstrate significant improvements in satisfaction (+45.0%, p < 0.001), personalization accuracy, and task achievement compared to a non-personalized baseline. Critically, subgroup analysis reveals consistent performance across age and education levels, suggesting equitable treatment. We discuss governance implications, including how the virtual-persona framework can serve as a public-interest audit tool for detecting algorithmic discrimination, and proposemechanisms for transparency, contestability, and participatory design in public-sector conversational AI deployment [5, 6].
Zitieren
APA
Makridis, G., Fragiadakis, G., Oliveira, J., Saraiva, T., Mavrepis, P., Fatouros, G., & Kyriazis, D. (2026). HumAIne-Chatbot: Real-Time Personalized Conversational AI via Reinforcement Learning for Public-Interest Digital Services. Journal of Artificial Intelligence for Sustainable Development. https://doi.org/10.69828/4d4kwm
Vancouver
Makridis G, Fragiadakis G, Oliveira J, Saraiva T, Mavrepis P, Fatouros G, et al. HumAIne-Chatbot: Real-Time Personalized Conversational AI via Reinforcement Learning for Public-Interest Digital Services. Journal of Artificial Intelligence for Sustainable Development. 2026. doi:10.69828/4d4kwm
BibTeX
@article{Makridis2026,
author = {Makridis, Georgios and Fragiadakis, George and Oliveira, Jorge and Saraiva, Tomaz and Mavrepis, Philip and Fatouros, Georgios and Kyriazis, Dimosthenis},
title = {HumAIne-Chatbot: Real-Time Personalized Conversational AI via Reinforcement Learning for Public-Interest Digital Services},
journal = {Journal of Artificial Intelligence for Sustainable Development},
year = {2026},
doi = {10.69828/4d4kwm}
}
EUVIMED