Assessing the applicability of existing design recommendations to AI companion design: A multi-method study

Key Findings

• — Nine design principle areas were synthesised in which existing recommendations from adjacent domains surface tensions when applied to AI companions.
• — Ethical and UX-oriented considerations were found to be deeply intertwined, so recommendations cannot be applied as separate checklists.
• — The four-phase process (literature review, practitioner co-analysis, internal heuristic evaluation, external expert assessment) is offered as a systematic problem analysis using the principle areas as an analytic artifact.
• — The paper's framing is that companion design is difficult as a design problem because relational, emotional and potentially long-term interaction changes what trustworthy-AI guidance means in practice.

Methodology Notes

Multi-method design analysis with practitioners and external experts; no user study of companion users. Published in IJHCS, September 2026 issue (Crossref gives month precision, OpenAlex 2026-09-01), closed access; an arXiv version (2609.14236) was submitted 2026-09-13. Verified via Crossref and OpenAlex metadata (title, journal, authors, University of Washington affiliations) and the arXiv abstract page; ScienceDirect is unreachable from this workspace and the nine principle areas are not listed here.

Sources