WhisperingGarden.

Turning passive waiting into an active response. An IV-pole interface guides people through the hospital garden to a listening point, where they hear voices from others and can leave one of their own.

My role
Product & design lead: research, interaction, hardware
Period
Mar–Jun 2026
Place
Erasmus MC, Rotterdam
Team
With Mingxuan Xu, Tiancheng Xi and Yuxin Li
Whispering Garden listening objects in the planted atrium of Erasmus MC
Comfort

Summary

Problem

During an infusion, people are tied to a chair, a line and a schedule. The wait feels passive and lonely.

Insight

Five patient interviews showed three layers of loneliness. A small sense of choice changed how the wait felt.

Concept

An IV-pole interface that leads people to a garden listening point to hear and leave voice messages.

Result

A working prototype. 5 / 5 participants chose a theme, found the point and listened to a full message.

What I did

3 contributions

  • Patient research & framework

    Led five interviews with long-term infusion patients and translated the findings into the Warmth, Voice and Choice framework.

    Research

  • Product direction & interaction

    Defined the product direction, the interaction sequence from IV pole to garden, and the emotional feedback language of light, pixels and voice.

    Concept

  • Hardware & prototyping

    Led hardware selection, integration and physical prototyping, up to a working loop of proximity sensing, pixel feedback and voice playback.

    Prototype

ESP32 × 2ArduinoHC-SR04WS2812B 8×8 LED matrixDFPlayer MiniPhysical prototypingIn-depth interviews

Product decisions & metrics

Tested with 5 participants
  1. Decision 01The interface lives on the IV pole, not a phone.

    Patients already keep a hand on the pole, and a screen would pull their eyes down instead of out into the garden. Trade-off: cues must stay glanceable, so little information fits.

  2. Decision 02Voice and Choice first, Warmth later.

    Temperature was far harder to build reliably, and a working end-to-end loop mattered more for testing. Trade-off: the physical layer of loneliness is still open.

  3. Decision 03Four themes, not an open feed.

    A small, real choice gave back a sense of agency without adding decisions. Trade-off: less variety per visit.

North starFull listens per infusion

Share of patients who hear a complete message during treatment. 5 / 5 in testing.

Supporting
  • Messages left per session (3 / 5 hesitated to record)
  • Self-reported agency and loneliness, before vs after
Guardrail
  • Tubing adjustments while interacting (2 / 5 in testing)
  • Extra work for nurses

From a five-person test in a planted setting: directional, not proof of effect.

01 / Context & opportunity

Waiting is not an empty moment.

During an infusion, a person is tethered to a chair, a line, and a schedule they cannot control. The physical constraint amplifies uncertainty and makes time feel strangely passive.

First-hand observation and patient conversations revealed familiar coping behaviors: looking at a phone, resting, or talking to someone nearby. Yet the needle, tubing, and IV pole continually pull attention back to treatment.

The hospital's planted atrium already offered light, greenery, and space to move. The opportunity was to make that environment respond gently to the person within it, without asking them to stare at another screen.

Concept visualization of the Whispering Garden in the hospital atrium
The setting. The planted atrium of Erasmus MC, where patients on an IV line can still move, see green and hear others.
Target users

People living with recurring or long-duration IV treatment, especially those who attend alone.

Experience goal

Make waiting time something a person can actively explore.

02 / User research

The need beneath the waiting.

Five in-depth interviews with long-term infusion patients shaped a four-stage emotional journey: before treatment, settling in, the long wait, and leaving.

The strongest signal was not simply boredom. Patients described three overlapping forms of loneliness: physical isolation, missing social feedback, and a deeper loss of agency. A small sense of choice could change how the wait was experienced.

Emotional journey

Interview synthesis
  1. 01

    Pre-infusion

    Anticipation and uncertainty

    “I was nervous before the needle went in.”
    Loneliness
  2. 02

    Settling in

    Body and equipment take focus

    “I felt a bit more in control when I adjusted my sitting position.”
    Loneliness
  3. 03

    The long wait

    Loneliness and low agency peak

    “Nobody explained what was next and the remaining time, so I felt a bit lost and anxious.”
    Loneliness
  4. 04

    The end

    Relief mixed with uncertainty

    “I feel like a burden asking the doctor if I’m done.”
    Loneliness

Filled cells: how strongly loneliness and low agency came up at each stage, from five interviews

  • 01 / MentalAnxious, lonely, unsettled
  • 02 / BodilyCold and physically restricted
  • 03 / TemporalUnclear process and remaining time
  • 04 / BehavioralMovement and one-handed activities constrained

Three layers of loneliness

Research synthesis → design dimensions
  • Physical

    Cold, constrained, unable to move freely

    WarmthNot built
  • Social

    Little companionship or immediate response

    Voice
  • Existential

    Few meaningful choices during the wait

    Choice

Why no Warmth? We explored heat in early concepts, but safe, controllable warmth right next to a patient on an IV line was too hard to build within the project. We put our effort into Voice and Choice, the two needs we could build and test with real people.

03 / Interaction concept

A garden that answers back.

Choose a voice theme, follow gentle cues into the garden, discover a listening point, and decide whether to listen or leave a message of your own.

Light and pixel expressions communicate a state from a distance. Voice makes the presence of another person tangible. Together, they give feedback without demanding continuous screen use or complex movement.

The concept evolved from a handheld device to a wearable cue and, finally, an IV-pole interface paired with a garden listening point. The final arrangement brought discovery and conversation into one place.

From object to place

Three iterations, one clearer interaction
  1. Early sketch of a removable handheld device attached to an IV pole
    01 / Handheld

    Warmth close to the body

    Useful as comfort, but not strongly connected to the garden experience.

  2. Prototype exploration of a wrist-worn directional indicator held near an IV pole
    02 / Wearable

    A directional cue

    Checking the wrist added effort; separating audio from the garden confused participants.

  3. Final working IV pole with pixel interface and separate garden listening point
    03 / Final prototype

    Interface on the pole

    The pole guides discovery; a single garden point holds the voice exchange.

The interaction, step by step

IV pole → garden
  1. IV pole01

    Choose

    Select Comfort, Humor, Advice, or Daily.

  2. IV pole02

    Seek

    Emoji size and haptic intensity shift with distance.

  3. IV pole03

    Arrive

    Color and a short pulse confirm the listening point.

  4. Garden04

    Listen

    Lift the handset and hear a thread of voices.

  5. Garden05

    Respond

    Record a new message or reply to one you heard.

A participant using the IV-pole interface
On the pole. The pixel display and a haptic pulse lead the way while the patient keeps a hand on the pole.
Garden listening device with a white handset, wooden housing, and an illuminated orange pixel expression
In the garden. The listening point: lift the handset to hear others, or leave a message of your own.
04 / Hardware & interaction

Built to respond in the real world.

The core proof is a working interaction loop: proximity sensing, pixel feedback, and voice playback responding within the same physical experience.

30-second working prototype demo. Proximity, pixel feedback and voice interaction running together.

System architecture

Sense → coordinate → express
01 · SenseProximity and input 02 · CoordinateState control 03 · ExpressLight, pixels and voice HC-SR04Ultrasonic proximity sensing Physical buttonsPhysical input ESP32 × 2State control andbutton-signal synchronization WS2812B 8×8 matrixPixel-matrix expression DFPlayer MiniAudio playback
Two ESP32 boards split the load and stay in step through a button signal, so light, pixels and voice respond as one experience.

Four voice themes

8 × 8 expressions from the original LED artwork
  • Comfort
  • Humor
  • Advice
  • Daily

Participants found Humor the most attractive theme. Comfort and Daily looked too similar, which the next version needs to fix.

Interaction states

  1. 01Theme selected

    Comfort · Humor · Advice · Daily

  2. 02Searching

    Expression changes with distance

  3. 03Point found

    Color confirms arrival

  4. 04Listening / recording

    Garden-side expression reflects audio state

What it took to make it run

Two ESP32 boards split the load and synchronize through a button signal. LED power, DFPlayer communication, power-bank auto-shutdown, and wiring inside the wooden housing were all real integration constraints.

The pixel display made discovery feel playful, but its cues had to stay glanceable. The device should direct attention toward the garden, not hold it on a screen.

Visitors trying the voice device at a hands-on prototype demonstration
Hands-on demonstration. Visitors trying the voice interaction at the garden listening point.

Inside the pixel display

The Arduino sketch maps an 8 × 8 LED matrix and two physical buttons to expression and listening states.

#include <Adafruit_NeoPixel.h>

#define LED_PIN      15
#define BUTTON_1     14
#define BUTTON_2     27
#define LED_COUNT    64
#define MATRIX_W      8
#define MATRIX_H      8
#define BRIGHTNESS   35

Adafruit_NeoPixel strip(LED_COUNT, LED_PIN, NEO_GRB + NEO_KHZ800);
05 / Validation & iteration

What worked. What resisted.

Five participants with infusion experience completed the full interaction journey in a planted environment. The sessions focused on clarity, emotional resonance, and willingness to engage.

  • 5 / 5
    Selected a theme and found the listening point
  • 5 / 5
    Listened to the full voice message
  • 3 / 5
    Were hesitant to record their own message
  • 2 / 5
    Repeatedly adjusted the IV tubing while interacting
The experience sparked curiosity. Humor was the most attractive theme.
Participant
Haptic and visual feedback complemented each other: one was explicit, the other intuitive.
Participant
Comfort and daily icons looked too similar; the white icon was too bright.
Participant

Next iteration

Keep / change
Keep

Choice generated more agency than passive content. The humor theme sparked curiosity, and visual and tactile cues complemented each other during discovery.

Change

Reduce the need to walk, hold the handset, or record. Test a seated, hands-free version; calibrate proximity feedback; make the theme symbols easier to distinguish.

06 / Reflection

A small choice changed how the wait felt.

We set out to make waiting time something a person could actively explore. With a working device in a real atrium, every participant chose a theme, found the garden point and listened to a full message.

Leading the hardware taught me where experience and engineering meet. Every cue had to stay glanceable, every wire had to fit inside a small wooden box, and light, pixels and voice had to respond as one.

What worked

From the five sessions
  • ChoiceChoosing created agency

    Picking a voice theme gave people something to decide during a wait that normally decides for them. It generated more agency than passive content, and 5 / 5 found the listening point.

  • SensesTwo cues, one message

    Haptic and visual feedback complemented each other during discovery: participants described one as explicit, the other as intuitive.

  • VoiceMessages held attention

    5 / 5 listened to a full message, and the Humor theme sparked the most curiosity.

Where I'd take it next

Next iteration
  • Make speaking up easier

    3 / 5 hesitated to record. A seated, hands-free listening point with lighter ways to reply, like a quick reaction, could lower that barrier.

  • Test where it matters

    Evaluate it in a real infusion setting over a longer period, and question the two-controller architecture before scaling it up.

Waiting doesn't have to be empty. A small, real choice can make it feel like your own time again.