This is documentation of Minjung Kim’s thesis in Interaction Design (IxD) from School of Visual Arts, New York, USA (Sep 2019 — May 2020).
Date
Sep 16, 2019
Context
This time I’m going to look it out the initial ideation of my thesis. I’m considering a project that can be helped people to reduce or eliminate the bias with Machine Learning technologies.
Location
This week’s status quo: Initial Ideation
I’m focusing on the first ideation session with an open brainstorm freely.
Problem
Machine learning (ML) sometimes makes a judgment about people’s gender, including age, disability, religion, ethnicity, physical fitness, body size, politics, and technology devices.
Feedback
The type of feedback I’m looking for today is specifically the idea direction only. Actionable feedback would be perfect on this stage.
Content
1. Introduction
1.1 Motivation
1.2 Goals
— 1.2.1 Problem Areas to Solve
— 1.2.2 Desired Outcomes
1.3 Related Work
— 1.3.1 Existing Products and Technical Research
— 1.3.2 Experimental Projects
— 1.3.3 Design Inspirations
— 1.3.4 Scope and Limitation
1.4 Purpose and Research Question
1.5 Approach and Methodology
1.6 Target Group
1.7 Problem Statement Outline
2. Theoretical Background
TBD
3. Design and Research Method
4. Results
TBD
5. Discussion
TBD
6. Conclusion
TBD
Introduction
- 1 Motivation
Being a woman who is pursuing a STEM degree in Design and Technology, I realized that there are scores of lamented facts about the gender imbalance in terms of occupation rate, leadership culture, and even compensation. In July 2017, I participated in a sprint session as a technical volunteer in a developer conference which is called ‘Pycon’ for young coders. It was an annual gathering for the community using and developing the software open-source ‘Python.’ I looked around the room, and I was feeling nervous because there are not many girls. I met one girl who followed her mother, a developer. The girl told me, “I like computer games and want to be a software engineer as my mother.” Her eyes glowed as she talked, and thankfully she was getting a proper education and incredible motivation from her mother. However, I realized that, unfortunately, most girls don’t have the same opportunity as her. This story also made me rethink about when I was an undergraduate student majoring in Visual Design. The design industry used to be a boys club at the top, lacking diversity across both gender and race. With a lack of representation among their role models, underrepresented people can be deterred from pursuing creative positions. Thanks to the pioneering activists before us, this has been changing, and many of our favorite designers working today are other women so that we can see more female leadership in the design industry. Therefore whenever I realize and experience the gender gaps in the technology industry after pursuing further study in the field, here in the United States, I strongly have felt a passion for addressing this modern problem for young women as a designer. This project means continuously questioning ourselves and making equality a central topic in our lives for the next creative generation.
1.2 Goals
— 1.2.1 Problem Areas to Solve
—— Inclusion & Diversity in STEM educational products
— 1.2.2 Desired Outcomes
—— A digital product that is related to bias, and judgments in educational products.
—— Probably on Mobile
1.3 Related Work
— 1.3.1 Existing Products and Technical Research
- Google Cloud Vision API
Cloud Vision API allows developers to easily integrate vision detection features within applications, including image labeling, face, and landmark detection, optical character recognition (OCR), and tagging of explicit content. - Google Magenta
An open source research project exploring the role of machine learning as a tool in the creative process. You can make music and art using ML. - Easy Screen OCR
Create this smart application to help users to capture the screenshot and then extract the text from these pictures in a most efficient way. Quite simple to use and it deserves giving a shot. - Microsoft Custom Vision
- Amazon Rekognition
Facial recognition is a system built to identify a person from an image or video such as personal photo applications and secondary authentication for mobile devices. - Amazon Go
It is a chain of convenience stores with no lines, no checkout. 17 stores open and announced store locations in Seattle, Chicago, San Francisco, and New York City as of 2019
— 1.3.2 Experimental Projects Resources
—— Open CV
—— MIT — IBM Watson AI Lab
— 1.3.3 Design Inspirations
—— Images & Links Cluster
A self-attention-based neural network that can generate music with long-term coherence.
— 1.3.4 Scope and Limitation
—— One or two features?
1.4 Purpose and Research Question
- 0. (Challenging myself — Don’t want to create just an app prototyping)
- 1. How do people feel when ML makes a judgment about them?
- 2. How to help people express themselves in a more comfortable way?
1.5 Approach and Methodology
Machine Learning
The idea of machine learning can be organized into “Prospecting → Comparing → Minimizing.”
1.6 Target Group
Teenagers? Young girls? Disable?
1.7 Problem Statement Outline
1) How might we prevent discriminatory outcomes in ML?
2) How might we make teens imagine everything they can become beyond the stereotype of a career?
3) Can girls poise to close the gender gap tomorrow?
Ideation Map
Opportunity and Risk Areas
Opportunities
- Some people can be consciously committed to equality and work deliberately to behave without prejudice.
- Some people with biases will not necessarily always act in biased ways and it is possible to consciously override bias.
- Most of the positive companies care about diversity and inclusion.
- Biases can be overcome in structured settings to solve shared problems.
Risks
- Unconscious biases are powerful predictors of behavior.
- Still possessing negative prejudices or stereotypes but they tend to give up quickly before they try.
- As human beings, we tend to think we are not perfectly rational beings. — at least that’s what decades of cognitive research has shown.
- Some of the cognitive biases are common enough.
Fin.
