Designing Inclusive and Responsible AI Agents for Diverse User Groups
Ahdiyeh Alipour · Frontiers in artificial intelligence and applications · 2026
The widespread use of AI agents in healthcare, education, and government services underscores the need for responsible interactions between these systems and diverse users in digital societies. Responsible AI interaction goes beyond technical performance and requires human-centered design that promotes transparency, accountability, fairness, privacy, and calibrated trust. Yet many AI systems still rely on generalized user models and a one-size-fits-all approach, neglecting how individual differences such as age, gender, cultural background, personality, and especially digital literacy can shape users’ trust, usability, inclusion, and perceived fairness in human-AI interaction. In public digital services, for instance, where users must interpret and act upon agents’ recommendations, differences in digital literacy may increase vulnerability to exclusion and mistrust. This interdisciplinary project adopts a mixed-methods design across five studies. Study 1 systematically reviews how trust in AI agents is conceptualized and measured, with attention to digital literacy. Study 2 combines surveys and semi-structured interviews to identify interaction barriers among users with different levels of digital literacy. These insights inform Study 3, which experimentally tests how alternative interaction strategies influence trust, usability, inclusivity, and fairness perceptions. Study 4 examines the role of agent modality by comparing embodied and non-embodied AI agents in public-service tasks. Finally, Study 5 employs co-design workshops and sandbox pilots with users and stakeholders to iteratively refine AI agent designs. Together, these studies will generate empirically grounded design insights to advance inclusive AI and more equitable access to AI-mediated public services.