AI-Guided XR Meditation Experience · Think Round

Breathe.

An AI-guided XR meditation experience designed for Think Round's immersive healing spaces — removing decision-making from users at their most emotionally depleted.

Role

UX Researcher & Designer

Company

Think Round

Platform

XR / VR Immersive

Type

AI-Guided Experience

Overview

What is this project?

Think Round is building a multi-floor cultural and healing centre in the San Francisco Bay Area. Their Healing Rooms are designed to support visitors processing trauma, anxiety, and emotional distress. I was brought in as UX Researcher and Designer to design an AI-guided XR meditation experience that lives within this physical space — one that reads a visitor's emotional state through four questions and generates a complete, personalized session with zero further decisions required from the user.

"The standard meditation app UI asks the user to do too much work at exactly the moment they have the least capacity to do it."

Core design insight — the case for AI-first

Survey questions

Emotional states

Focus areas

Breathing profiles

Demographic questions

Hard AI rules

The Problem

Why existing apps fail here

Calm, Headspace, and Insight Timer all use a catalog-browse model. Users see dozens of sessions, apply filters, and make multiple decisions before ever closing their eyes. This works for a calm, well-rested user. It fails completely for Think Round's visitors — who arrive in the Healing Rooms after processing difficult emotional experiences and may be anxious, raw, or overwhelmed.

Existing approach — what was proposed

Catalog + browse model

User browses sessions, filters by duration and type, reads descriptions, selects. Requires 5–8 active decisions before starting. Appropriate for a calm user. Harmful for a user in acute emotional distress.

This approach — what I designed

AI-first · 4 questions only

User answers 4 questions. AI reads their state and generates a complete session — environment, sound, breathing, technique, duration, and personalized opening message. Zero browsing. Zero catalogue.

My Contribution

What I specifically did

Physiological Research

Vagus nerve, parasympathetic system, CO₂ mechanism, breath hold safety per state

Survey Design

4 questions justified, exclusions documented, full decision mapping

AI Rule Architecture

Hard rules, soft rules, 7-step decision flow, prompt structure

UX Journey

Both paths designed, interactive prototype built, UI critique delivered

Research Phase

What I researched and why it mattered

Before designing a single screen I researched the physiological science behind breathing-based meditation — not just which techniques exist, but why each one works, for which emotional states, and where they become actively harmful.

Vagus nerve and the parasympathetic system

The vagus nerve is the body's primary brake cable — running from the brainstem down through the chest to the gut. It fires during every exhale, slowing heart rate and switching the nervous system from fight-or-flight to rest-and-restore. This is why the exhale phase is the active calming mechanism, not the inhale.

CO₂ mechanism and neural excitability

During breath holds, CO₂ builds in the blood — dilating cerebral blood vessels and reducing how easily neurons fire. This produces heaviness and drowsiness. It is the mechanism behind Dr. Andrew Weil's 4·7·8 sleep technique. But it only works if the nervous system is already calm. For anxious users, the same CO₂ triggers the amygdala's threat response — worsening rather than calming the state.

Breath hold safety — the hardest design rule

When someone is anxious, their chest is already tight and they already feel short of breath. A breath hold creates air hunger that the brain reads as suffocation. This became a hard rule in the AI system: if Q2 = Anxious, no breath holds — ever, regardless of any other survey answer or user goal.

Survey question design — minimum viable information

I mapped every session decision (environment, sound, breathing, technique, duration, voice tone) back to the minimum survey questions needed. 4 questions unlocks all decisions. Gender, age, and ethnicity were excluded deliberately — they do not change breathing technique or environment choice. Their inclusion would add friction at the worst moment without adding value.

AI as the recommendation system

The recommendation system is Claude AI receiving a structured prompt containing the design rules and the user's 4 survey answers. There is no separate decision tree or ML model. The prompt IS the algorithm. A developer writes it once. Claude applies it to every user automatically, returning a complete session configuration in 2–4 seconds.

Design Framework

Three locked layers — the core system

The entire system is built on three layers kept strictly separate: emotional states (what the user feels), focus areas (what the AI infers), and breathing profiles (what the AI selects). Confusing these layers was a persistent problem I corrected throughout the process.

Q2 Emotional state	Focus area	Breathing profile	Pattern
Anxious	Anxiety core	Anxiety calm	4 · 6
Tense	Stress relief core	Stress relief	4 · 4 · 8
Tired	Emotional reset secondary	Anxiety calm	4 · 6
Scattered	Mind clarity core	Deep focus / Balance	4·4·4 / 4·4·4·4
Sad	Emotional reset secondary	Anxiety calm	4 · 6
Neutral	Follows Q4 intention entirely — goal is the primary driver when mood is neutral
Q4 override: Ready for sleep	Sleep aid core	Sleep ease	4 · 7 · 8
Q4 override: Be kinder to myself	Self-compassion secondary	Anxiety calm	4 · 6

Safety exception: Anxious + Q4 = Ready for sleep → Focus area becomes Sleep aid but breathing stays Anxiety calm 4·6. Never Sleep ease 4·7·8 for anxious users — the 7-second hold would trigger the amygdala's threat response.

AI Rule System

How the AI makes decisions

The prompt contains two types of rules. Hard rules cannot be overridden under any circumstance — they exist to protect the user's physical and emotional safety. Soft rules are strong preferences the AI balances when survey answers pull in different directions.

Hard ruleQ2 = Anxious → NEVER include breath holds, regardless of any other answer

Hard ruleQ1 = First time → NEVER use Sleep ease 4·7·8

Hard ruleQ2 = Anxious → NEVER choose desert or cosmos environment

Hard ruleDuration must NEVER exceed what the user stated in Q3

Soft ruleSad or tired → prefer enclosed sheltered environments (forest, bamboo)

Soft ruleMind clarity → prefer silence or single Tibetan bowl over ambient sound

Soft ruleFirst time / Occasional → more guidance and explanation before each breath phase

Soft ruleSleep goal → birdsong never selected — birds signal morning, not rest

Think Round Alignment

How this serves Think Round's mission

Think Round's vision — The Center for the Human Family and the Paradise Project — aligns with this work across four dimensions. The VR meditation experience is not a standalone product; it is the technology that powers their Healing Rooms and experiential spaces.

Think RoundHealing Rooms — trauma support, ACEs and PACEs programs, intergenerational trauma

→

This workEmotional reset and self-compassion focus areas — open monitoring and loving-kindness practice are evidence-based trauma support tools

Think RoundMindfulness skills development in curriculum — pre-natal through grade 3

→

This workPhysiological breathing research — vagus nerve, parasympathetic system — is the scientific backbone of any credible mindfulness curriculum

Think RoundParadise Project dome — holographic earth, constellations, 3D renderings, immersive audio

→

This workNight Sky, Forest, Ocean, Cosmos VR environments — digital extension of the dome's immersive vision using the same spatial technology

Think RoundInterfaith and cultural literacy — serving all of humanity regardless of background

→

This workZero demographic questions — no gender, age, or ethnicity collected. The same session quality is delivered to any human regardless of who they are

What's Next

Remaining work

The research foundation and design framework are complete. The following deliverables and design phases remain before the project reaches development handoff.

Immediate priority

User personas (1–2 profiles)
Formal problem statement
Annotated wireframes (key screens)
Survey rationale document

High priority

Trauma-informed design research (ACEs/PACEs)
Physical space integration design
Competitive analysis document
Voice design and tool selection
Children / family version scoping

Production phase

VR environment specifications (6 spaces)
Sound library selection and licensing
Returning user flow design
User testing (5–8 participants)