Using AGI in Human Relationships Without Losing Human Agency

The use of AI-driven analytics in human relationships—romantic, familial, professional—has rapidly expanded. Compatibility scores, breakup predictions, emotional pattern recognition, and relationship ROI metrics are increasingly normalized. While these tools promise clarity and efficiency, they raise a fundamental concern: When AGI evaluates relationships, does it support human judgment, or does it silently replace it? Existing AI ethics literature focuses primarily on harm prevention, bias, and explainability. However, relational contexts introduce a deeper risk: the erosion of human responsibility through outsourced meaning and decision-making. This paper addresses that gap by introducing a PKF-based boundary model for AGI use in human relationships.