Cultivate

A desktop plant that provides real-time feedback to language sentiment analysis

Tools Used
Raspberry Pi, Python, Figma, Fusion 360, Ultimaker 3D Printing
Date
2021
Type
Product Design
Role
Back-end development, ideation
Tools
Raspberry Pi, Python, Figma, Fusion 360, Ultimaker 3D Printing
Teammates
Benal Johnson, Celia Diaz, Emily Joens, Jennifer Wong
Duration
2 months
(Oct '21 - Dec '21)
Intro

Stimulating controversial discussion about gender inequity

Cultivate is a speculative design project aimed to stimulate discussion about gender inequity in the workplace.
I challenged myself to approach this speculative project as a legitimate product with a sophisticated design process and complex programming to learn how these tools can be used for unexpected purposes.
The final product is a whimsical 3D printed desktop plant that using a raspberry pi to analyze language sentiment and provide real-time feedback via wilting and flourishing of leaves.
Problem
(Paw Research Center)

42% of women have experienced gender discrimination at work

Workplaces strive for gender equity, yet women are still primary targets of micro-aggressions and biases in the workplace. Current solutions generally expect women to solve these problems on their own.
With the advancement of AI technology, there should be an objective way to confront gender inequity in the workplace—where no individual is responsible to monitor such behaviors.
I realize that AI tools can spark new issues such as those related to data privacy. I intentionally utilize a broad problem statement that does not address other issues to elicit genuine, unbiased direction from research and feedback.
AI technology seems like an obvious way to objectively confront gender inequity in the workplace.
I realize that the solution is not obvious and implementing AI tools can spark new issues such as those related to data privacy. I intentionally utilize a broad problem statement that does not address other issues to elicit genuine, unbiased reactions.

How might we support a more inclusive work environment?

User Interviews

Understanding women in the workplace

We met with four working women to understand the impact of inequity in the workplace and what might make women feel more supported.
Interviewee A
‍"I wish someone would stand up for me"
Need: Active support from others at work
Interviewee B
‍“We all see the problem but no one is doing anything"
Need: Methods to hold people accountable
Interviewee C
"I am afraid to say, ‘Hey, I don’t feel fully included.'"
Need: Another way to address inequality, not confrontation
Interviewee D
“It feels like no one even understands the problem”
Need: Education about the inequalities taking place
Ideation

Three categories of brainstorming

JMARS data is too complicated

During the ideation process, I performed several brainstorming activities with my teammates to ideate possible solutions to the needs brought up in interviews. Our ideas fell into three main categories:
  • Wearable Devices for consistently available support
  • Personal Storytelling Device to listen and share unbiased support/feedback
  • Ambient Intervention to unobtrusively communicate that intervention was needed + count number of interventions
Low Fidelity Prototypes

Building low-fidelity cardboard prototypes

We narrowed down these themes to three concrete solutions using several convergence-divergence exercises.
Wearable Watch to notify trusted colleagues when support is needed
Storytelling Plant to act as a trusted confidant, listening to stories and providing printed feedback/support
Accountability Plant + Tracker to record detailed info for instances of positive/negative speech and visually show results in real-time

Low Fidelity Prototype

JMARS data is too complicated

We narrowed down these themes to three concrete solutions using several convergence-divergence exercises.
  • Wearable Watch with tools to notify allies when support is needed.
  • Personal Storytelling Device to listen and share unbiased support/feedback
  • Reactive Plant to provide an ambient visual representation of positive/negative speech in real-time via flourishing/wilting of leaves
Usability Testing

Understanding how the possible solutions address user needs

JMARS data is too complicated

I worked with my team to test the prototypes with our target audience and better understand gaps between the product and user needs.
Key insights were:
  • Discomfort sharing personal stories at work
  • Interest in a product taking action (vs. pushing a button)
  • Value in objective nature of plant (based on data rather than opinion)
  • Appreciation towards ambient nature of plant to give feedback without interrupting conversations
  • Concern for data privacy with information recording—need more abstract data collection with no private information being saved
Final Direction

Storyboarding a subtly reactive desktop plant

JMARS data is too complicated

After reflecting on the insights, we voted and decided on our final concept—a desktop plant that subtly counts and reacts to language sentiment via wilting/flourishing of leaves.
I worked with my team to storyboard interactions based on gaps identified by user testing.
Iterations

Abstracting data collection and assessing necessary technology

JMARS data is too complicated

We iterated our design to based on feedback and storyboarding.
The main iterations were:
  • Abstract data collection—count of negative/positive instances rather than direct quotes and names
  • Data present on product's GUI screen
  • Implementation of colored LED lights to clearly indicate the difference between wilt and flourish
Fabrication

3D printing

JMARS data is too complicated

The final product was fabricated using a combination of 3D printed parts. The magic behind this product is a custom spool, which attaches each individual plant to a servo motor.
Well, actually the real magic is Benal Johnson, an inspiring Industrial Designer (and friend) who I feel lucky to have the opportunity to work with.

Fabrication

The final product was fabricated using a combination of 3D printed parts. The magic behind this product is a custom spool, which attaches each individual plant to a servo motor.
Well, actually the real magic is Benal Johnson, an inspiring Industrial Designer (and friend) who I feel lucky to have the opportunity to work with.

Programming
‍

Alright, this is where I brought the magic!
I programmed Cultivate using Python and Google Cloud API on a Raspberry Pi to capture, transcribe, and analyze speech and react to that speech in real-time. The reactions that I programmed are outlined below.
Programming

Writing script for plant to react to sentiment analysis

JMARS data is too complicated

Alright, this magic was all me!
I programmed Cultivate using Python and Google Cloud API on a Raspberry Pi to capture, transcribe, record, and analyze speech and react to that speech in real-time. The reactions that I programmed are outlined below.
Ideation
Programming

Writing script for plant to react to sentiment analysis

Alright, this magic was all me!
I programmed Cultivate using Python and Google Cloud API on a Raspberry Pi to capture, transcribe, record, and analyze speech and react to that speech in real-time. The reactions that I programmed are outlined below.

Speech to Text

minimal code snippet syntax highlighting example

Sentiment Analysis

minimal code snippet syntax highlighting example

Plant Reaction

minimal code snippet syntax highlighting example
Final Design

A controversial desktop plant that visualizes language sentiment

The final design is meant to look modern and simple, to blend into the most common aesthetic of current workspaces. We utilized red and green as our main colors, universal symbols for positive and negative to clearly indicate results of sentiment analysis. We used an elegant and familiar fonts to give a feeling of respect and inclusion.

Reflection

JMARS data is too complicated

Applying sophisticated technology and design methods to a complex and controversial topic was a huge challenge. In hindsight, there is room for improvement for tackling similar challenges in the future. Most significantly, testing should be done with a representative sample of everyone affected by the product—Cultivate should have been tested with all genders not just women.
Overall though, Cultivate was a success in starting the conversation about gender inequity in the workplace. It was a personal success too. This project pushed me way of my comfort zone and taught me the importance of speculative design to gain insights on user needs and predict future trends. Additionally, I learned the affordances of rapid prototyping to save time by finding mistakes early on the design process.
Next Project
TeachXR