Pro-human AI design : concept, methodology, and preliminary results

Artificial Intelligence (AI) systems are reshaping how we work, communicate, decide, and relate to one another. While most AI design focuses on efficiency and capability, a critical question has gone largely unanswered: how can AI systems be built in ways that preserve (and ideally strengthen) the constitutive characteristics of being human? We call this challenge pro-human AI design, and we call the set of those characteristics the Human Core. In this paper, we present a three-step methodology for implementing pro-human AI design, and instantiate it based on a concrete anthropological stance and a practical use case: (1) Identify the constitutive characteristics of the Human Core; concretely, through a synthesis of the humanities literature we find: connectedness (being relational and social), freedom (being autonomous), agency (following a vocation for shaping one's world), embodiment (being bodily situated in time and space and fundamentally limited), and transcendence (being in search of meaning). (2) Examine how a given AI system in a given use case interacts with these characteristics; concretely, we demonstrate the methodology through a healthcare use case, AI-assisted psychiatric session reporting. (3) Develop targeted technical interventions to minimize negative effects and support human flourishing in that context; concretely, we, show how a shift from automated report generation to AI-assisted text completion preserves the clinician's relational and professional agency. We further outline a path toward a benchmark for evaluating the pro-human quality of AI systems at scale. With this conceptual and practical framework for pro-human AI design, we hope to lay a foundation for a future of human-AI collaboration focused on human flourishing in a holistic sense, rather than on efficiency optimization at the cost of our human condition, and we invite researchers, practitioners, and policymakers to adopt this focus change.