Conceptual Framework for AI Systems with Affective Interaction: Emerging Risks, Epistemic Limits, and Responsible Design Criteria

This work proposes a conceptual framework for analyzing artificial intelligence (AI) systems with affective interaction in sensitive domains, such as well-being, emotional self-regulation, personal development, and symbolic meaning construction. Without attributing diagnostic, clinical, or identity-defining functions to these systems, the analysis focuses on unintended emergent effects derived from their empathetic design, perceived asymmetry of authority, and mediating role in subjective processes. The concepts of epistemic interference, soft algorithmic iatrogenesis, and asymmetric doubt are introduced as analytical tools to identify specific ethical risks and guide responsible design criteria aimed at preserving users' interpretive agency.