← Selected work
HUMAN–AI INTERACTION · PUBLIC ARTCASE STUDY / 03

Gaze to the Stars

An AI-mediated public art experience that transforms MIT’s Great Dome into a platform for collective storytelling.

MY ROLE

Experience design & AI development lead

CONTEXT

MIT Media Lab · Critical Matter Group

TOOLS & PRACTICE

LangChain · Computer vision · Projection

Visit the project ↗
MIT’s Great Dome at night, its projection reflected across the Charles River
MIT MEDIA LAB · CRITICAL MATTER GROUP

Within every eye
lies a hidden story.

Woven like constellations in the night
01

What if architecture could listen?

Inside a private pod, people share personal stories with an AI embodied by MIT’s Great Dome. Their stories are encoded within eyes and projected onto the Dome—turning a moment of reflection into a public experience of being seen.

  1. Enter the pod

  2. Talk with the Dome

  3. See your story

  4. Explore shared stories

02

A protected space for reflection

The experience begins
with a place to pause.

An intimate, individual encounter.
The illuminated pod, ready to listen.
THE CONVERSATIONAL POD

Privacy sets
the tone.

Inspired by the intimacy of a confessional and the calm of meditation spaces, the pod combines biometric sensing with storytelling.

The enclosure creates a protected environment for participants to share personal dreams, struggles, and aspirations.

INSIDE THE ENCLOSURE

An infinity room,
inspired by stargazing.

Light, sound, and spatial framing guide attention away from external distractions before the conversation begins. The mirrored interior gives the participant an immersive setting for self-reflection.

The view inside the spherical interface.
The enclosure connects the facial interface, mirror space, display, sensing hardware, and sound.
03

A conversation with the Dome

The building takes
on a listening voice.

The AI’s designed persona is a 109-year-old goddess embodied by the MIT Dome. It introduces itself, guides a meditation, and invites participants to reflect.

LangChain workflows, voice synthesis, and real-time conversation support the exchange. The system responds and asks follow-up questions while keeping the participant’s own story at the center.

FOUR OPENINGS INTO A PERSONAL STORY
  • “What do you dream of becoming?”
  • “What are you unwilling to feel?”
  • “What have you been feeling lately?”
  • “What are you longing for?”
Introduction and meditation lead into a personal prompt, reflection, a follow-up question, and a closing response.
04

Where stories meet the gaze

An eye becomes a vessel
for a personal story.

The eye was chosen to retain emotional presence while aiming to preserve anonymity. Two kinds of input come together: the participant’s gaze supplies the visual canvas, and their spoken narrative supplies the message.

SAM2 and custom gaze tracking extract iris structures and eye movements.
COMPUTER VISION

A moving canvas

Tracking keeps the iris structure and movement at the center of the visual translation.

VISUAL INPUT

The gaze

Eye capture, iris segmentation, and particle mapping form the visual layer.

SEMANTIC INPUT

The story

A transcript summary becomes Braille, arranged as a code ring within the iris.

The visual and semantic streams converge in the encoded eye render.
THE ENCODED RESULT

A story carried
within the iris.

Personal narratives become hidden messages that invite viewers to look closer.

Narratives encoded within participants’ eyes
The encoded gaze at architectural scale.
05

The moment of being seen

Private reflections,
shared with the city.

200+personal stories projected
onto MIT’s Great Dome

The installation brings individual reflections into a collective public experience. Each eye carries a personal narrative onto a landmark that the city can see.

Seeing my story on the Dome made me feel that my life had been seen by the city. I wasn’t alone.

A participant in Gaze to the Stars
06

A constellation of shared experiences

Individual stories.
A larger human story.

The experience continues beyond the physical installation through an interactive emotional atlas linking stories, eyes, and participants.

Embeddings, UMAP, and HDBSCAN clustering surface shared themes across the narratives. The visualization offers a way to explore connections between personal experiences.

At left, narrative clusters. At right, the interactive atlas connects each eye to a story and a wider constellation.
FINDING SHARED THEMES

Machine learning groups related narratives, revealing patterns across individual experiences.

EXPLORING THE CONNECTIONS

Visitors can rotate the digital Dome and follow links between eyes, stories, and clusters.

PROJECT FILM

See the complete experience.

Explore the project
Gaze to the Stars · Project film
Project credits & recognition
My contribution

As a research assistant in the Critical Matter Group at the MIT Media Lab, I led development alongside Professor Behnaz Farahi. I designed the end-to-end participant experience, from the conversational pod to the Dome projection, and developed the human–AI interaction framework while contributing to the software system.

Team

Behnaz Farahi, Haolei Zhang, Sergio Mutis, xdd, Suwan Kim, Yaluo Wang

Recognition & publication

NeurIPS 2025
ISEA 2026

NEXT PROJECT / 01

Amimi

Reliable, accessible relationship support for young couples, woven into everyday life.