Human-AI Social Roles in Mental Health — AROMA

Research · Human-AI Social Roles in Mental Health — AROMA

NYU Tandon COPHEE Lab · Jan 2026 – Present · with Issac Xian · advised by Prof. Vedant Das Swain

AROMA — Affective Role Ontology for Mental-health Assistance.

Intro

Mental-health AI is being deployed at scale and labeled indifferently as coach, friend, therapist, companion — with no shared definition of what those labels commit the system to. AROMA is a taxonomy of what the AI is being when it supports a user emotionally, grounded in a PRISMA literature corpus.

Research goals

• Give the field a shared vocabulary for AI care roles that's corpus-grounded, not vibes-based.
• Separate Role (who the AI is being) from Function (what it does on a turn) and Strategy (how it does it) — prior taxonomies collapse all three.
• Make the Authority–Agency Paradox of AI care roles explicit: users project authority onto AI that lacks the structural agency to honor it.

Research questions

• RQ1 — What set of care roles are corpus-supported, behaviorally distinct, and viable in current AI modalities?
• RQ2 — What are the activation conditions and invited human roles for each care role?
• RQ3 — Where is the Authority–Agency Paradox most dangerous, and how does it vary by role?

Plan

Two-phase: a PRISMA literature synthesis to mine the role-term vocabulary, then a structured taxonomy pass (inclusion criteria, cross-annotator agreement) to collapse the mined terms into a small set of care roles, each with a relational stance, primary support type, and authority–agency profile.

This is the plain-HTML mirror served to crawlers, LLMs, and curl. Humans with a JavaScript-enabled browser see the rich React/XP-themed SPA at the same URL.

All plain pages · Live site · sitemap.xml