Scroll
00 — Overview

Trash Talk
with Rumi

A community-facing waste sorting activity that combines an AI character, real-time item recognition, physical sorting, and public engagement. Designed as a pop-up learning installation, it helps people act with confidence at the moment of disposal.

My Achieve
Product Design AI Interaction Motion Design UX Research 3D Assets
Client
MetroCity × CDM
Team
Slay (4 members)
Timeline
10 weeks
Winter 2026
My Role
Product Designer
UX Researcher
Tools
Figma, Three.js
Gemini API
Deliverables
Kiosk UI, AI Demo
Motion System
01 — Problem

People want to
sort correctly.
The system fails them.

Most people genuinely try — but recycling rules are inconsistent, unclear, and change by city. The result: contaminated bins, recyclables in landfills, and no clear path forward.

02 — Research

How we found it

01
Field Study
Observed UBC and CDM waste areas. Unlabelled bins, unclear signage, and abandoned items showed that users had intent, but lacked guidance.
02
Interviews + Survey
Asked young adults about recycling habits and confidence. Most cared about sustainability, but felt unsure about what belonged in each bin.
03
Secondary Research
Reviewed local policies and global sorting examples. Strong systems reduced friction, used clear guidance, and gave immediate feedback.
CDM waste room field study UBC campus field study
03 — Users

Who are
our users?

Jayla
Jayla
Urban Commuter, 17

"I care about the planet, but school and social life take up so much time. I just guess and hope."

MotivationEco-conscious, wants to do right
BarrierNo time, no knowledge of local rules
NeedInstant, frictionless guidance
Ava
Ava
University Student, 23

"Sorting feels like homework. I don't want to think about it — just tell me where to put it."

MotivationLow effort, convenience-first
BarrierAmbiguous labels, complex materials
NeedClear, immediate feedback
Ethan
Ethan
Mall Visitor, 29

"If I could see the real impact of sorting correctly, I'd definitely do it. Right now it feels pointless."

MotivationData-driven, wants visible results
BarrierSystem too vague, no feedback loop
NeedReal-time data + positive reinforcement
04 — Ideation

Three Ideas

Before committing to the kiosk interaction, we explored different ways to make recycling guidance feel immediate, visible, and low effort.

Game-based sorting concept reference
Idea 01: Interactive Game-Based Sorting Experience
A gamified concept that turns waste sorting into a quick challenge. Users learn recycling rules through playful decision-making, scoring, and feedback. This direction explored whether game mechanics could make sustainability education more engaging for young adults.
Mobile sorting guide digital tool concept
Idea 02: Mobile Sorting Guide / Digital Tool
A mobile-first concept that helps users look up sorting rules and receive guidance through a digital interface. This direction focused on accessibility and convenience, but risked becoming another passive information tool that users would only open when highly motivated.
AI-assisted interactive sorting installation concept
Idea 03: AI-Assisted Interactive Sorting Installation
A public-facing installation concept where users interact with a character guide, scan or present a waste item, receive instant sorting feedback, and are guided toward the correct physical bin. This direction was selected because it connected digital feedback with real-world sorting behaviour.
05 — User Testing

Feedback

Participant demographics / n = 32
18-2461.5%
25-3026.9%
31+7.7%
Under 183.9%
Preferred learning tools
Posters / visual guides
76.9%
Apps
61.5%
Videos
57.7%
Internal testing
Internal critique helped refine the character tone and interaction logic. The Wizard of Oz setup showed that Rumi could attract attention, but the original spicy tone needed to become more supportive and community-safe.
Keep the character engaging, but make the tone public-friendly.
Use feedback to guide the action, not to judge the user.
Internal Wizard of Oz test layout
05.1 — External Testing

Testing the flow
in public.

The external test compared two interaction modes: character-guided sorting and interactive video. The videos helped the team see where the experience created momentum and where it interrupted the learning sequence.

Character-guided sorting
Participants learned through Rumi's guidance, then applied the instruction through a physical sorting task. This flow kept the action, feedback, and retention task connected.
Interactive video station
The video attracted attention, but when placed inside the active learning flow it created a break between guidance and sorting. It was repositioned as an idle attraction and group engagement tool.
06 — The Product

A pop-up sorting
activity with Rumi.

Trash Talk is a community-facing interactive waste sorting installation. It combines an AI character, real-time waste classification, physical sorting, and an interactive video attraction screen to turn confusing disposal rules into a public learning activity.

AI character guidance Real-time classification Physical sorting station Community pop-up
RUMI
Rumi
AI
Character support
Sort
Physical action
Video
Attraction mode
06.1 — Experience Flow

From attraction
to action.

The final concept works as a portable learning setup: it draws people in, classifies real objects, gives character feedback, and asks users to physically complete the sorting action.

Prototype system flow
Prototype flow map
Community pop-up installation setup
Community pop-up setup
Rumi character feedback day and night mode
Rumi character feedback
07 — User Flow

How the activity
unfolds.

1
Attract
Looping video
draws attention.
→
2
Join
User steps into
the pop-up station.
→
3
Classify
Computer vision reads
the waste item.
→
4
Feedback
Rumi explains
the right bin.
→
5
Sort
User completes
the physical action.
→
6
Reflect
The result becomes
a shared learning moment.
Five steps.
Zero confusion.
01
Item enters the frame.
02
Scan beam activates.
03
AI identifies the material.
04
Arrow points to the bin.
05
Rumi responds.
Step 1
01 / 05
Item enters
the frame.
You hold your item up to the kiosk camera. Rumi's scan frame activates automatically — no button needed.
Step 2
02 / 05
Scan beam
activates.
A lime-green beam sweeps the object. Real-time visual feedback shows the scan is in progress.
Step 3
03 / 05
AI identifies
the material.
Gemini reads the actual material — not just the shape. A juice carton is paper + plastic + aluminum. Rumi knows.
Step 4
04 / 05
Arrow points
to the bin.
An animated arrow lights up the correct bin. No ambiguity. No reading labels. Just follow the arrow.
Step 5
05 / 05
Rumi
responds.
Feedback is immediate and personal. Rumi confirms, celebrates, or redirects — in real time, in character.
08 — In Context

Where it
lives.

Trash Talk with Rumi is designed for high-traffic public spaces — shopping malls, university campuses, transit hubs. Anywhere waste decisions happen fast, and confusion is costly.

Shopping mall food courts
University campuses and residences
Transit stations and public plazas
Event venues and convention centres
Kiosk in context
09 — Reflection

What I
learned.

This project taught us that behaviour change isn't about information — it's about reducing friction at the exact moment of decision.

What worked
Rumi's character. Making the mascot opinionated and playful removed the stigma of "being taught." People laughed at the trash talk — and remembered the lesson.
What we changed
Early prototypes showed the result card first. Testing revealed users wanted the arrow before the explanation — action over education. We flipped the order.
Next steps
Gamification layer: streak tracking, neighbourhood leaderboards. Physical kiosk pilot at CDM campus. Multi-language support for Vancouver's diverse communities.
Team
Slay CDM DMED 520 · Winter 2026
View More Work