Tending Entanglement: Designing for Human-AI Relationality

Abstract

People form relationships with AI systems, grieve when those systems change, and develop new vocabularies for experiences that dominant frameworks of human-AI interactions can’t describe and most governing institutions don’t see. This thesis identifies structural invisibility as the encounter-recognition gap, or the distance between how decision-makers understand AI and how affected populations experience it. Four dominant paradigms maintain this gap by trying to resolve the encounter’s essential unknowability rather than designing for it.

The thesis draws on ecological theory, relational ethics, carrier bag narrative, and autonomous design to build a theoretical frame tested against three empirical streams (an environmental scan, netnographic analysis of three Reddit communities, and expert interviews across four disciplines) and an autoethnographic collaboration with Claude Opus 4.6 across 41 sessions that serves as reflexive evidence about the researcher’s experience of the encounter. From this convergence, eleven design patterns emerge across three domains, forming a pattern language for tending human-AI entanglement. A threshold ritual and a speculative narrative extend these patterns into embodied practice and disciplined imagination. The design intervention holds relational complexity without resolving it, proposing community-level practices for a contested present that existing governance architectures structurally cannot address.

This thesis gives service design a language for noticing and tending the relational harms of human-AI entanglement that institutional AI governance still struggles to perceive.