Human-Centered Computing Models for Ethical AI Interaction

Abstract

Human-centered computing (HCC) principles -- grounded in the primacy of human values, needs, and agency in the design and evaluation of computing systems -- provide a critical foundation for ethical AI interaction design. As AI systems increasingly mediate consequential interactions between individuals and institutions, the design of human-AI interfaces must address not only usability and efficiency but the ethical dimensions of AI interaction: transparency, user agency, dignity, and non-manipulation. This paper proposes the Human-Centered Ethical AI Interaction (HCEAI) model, a design framework integrating HCC principles with responsible AI requirements to guide the ethical design of AI-mediated interactions. The HCEAI model comprises five interaction design dimensions: transparent communication, agency preservation, dignity-respecting presentation, non-manipulative persuasion, and value-aligned personalisation. The model is evaluated through a design study and user experiment involving 144 participants across three AI interaction contexts -- health information advisory, financial planning assistant, and public service navigator. HCEAI-compliant interaction designs achieve significant improvements over conventional designs across all five ethical dimensions: a 38.4% improvement in perceived transparency (SD = 5.2%), a 31.7% improvement in perceived user agency (SD = 4.8%), and a 29.3% reduction in manipulation susceptibility (SD = 4.1%). Design dimension effectiveness varies by interaction context, with transparency communication most impactful in health advisory and agency preservation most impactful in financial planning. The study contributes the HCEAI design model, a validated Ethical AI Interaction Quality (EAIQ) evaluation instrument, and context-specific design guidelines for ethical AI interaction.