Designing and Evaluating Human-Centred AI Systems: Best-Practices from a Multidisciplinary View
Giuseppe Desolda, Andrea Esposito, Rosa Lanzilotti, Antonio PiccinnoAbstract
The rise of Human-Centred Artificial Intelligence (HCAI) highlights the growing need for AI systems that are not only functional but also ethical, transparent, and user-focused. Leveraging the results of a literature review, this paper briefly explores best practices and multidisciplinary methodologies for designing and evaluating HCAI systems. Drawing from fields such as Human-Computer Interaction (HCI), Software Engineering (SE), Machine Learning (ML), and Ethics, we provide a brief synthesis of design techniques, evaluation strategies, and fairness frameworks. We emphasize the importance of iterative, user-centered design processes and propose actionable insights for integrating interpretability, bias mitigation, and ethical considerations into HCAI systems. Despite these advancements, significant challenges persist, including the lack of tailored heuristics for evaluation and the complexity of ethical impact assessments. Addressing these gaps will require scalable evaluation methodologies and enhanced interdisciplinary collaboration.