The Ethics of AI Relational Fields

This paper reframes the ethically salient unit in human–AI interaction as a relational field—a dynamic configuration of practices, traces, cues, and expectations that accumulates across time—rather than a solitary “object” or “pole.” Building on heuristic analogies from physics (fields/resonance) and the arts (performance/stewardship), we argue that ethics extends beyond momentary “use” to the lifecycle custodianship of relations: continuity, identity handoff, relationship export, and sunset. Part I develops this non-welfarist, relation-centric account and shifts responsibility from individual users toward collective and institutional practices. Part II reports a cross-domain scoping review (2022–2025) using a 3×4 matrix (maturation stage × domain) to locate signals across Academia, Community, Industry, and Institutions/Standards. We find strong “sprouts/threshold” activity (e.g., productized memory, community handoff rituals) but limited consolidation around cross-vendor relational portability, standardized sunsets, and continuity metrics. We propose actionable implications: ship Relationship Export with a portable identity bundle; implement Sunset by Design with receipts; align artifacts to consent/receipt and portability standards; and introduce a Relational Continuity Score with audit protocols and plug-fests for interoperability. These moves convert private craftsmanship into public custodianship, making continuity without lock-in—and closure without erasure—the easy thing.