Boa Yoon

Turning early confusion into early action with AI-powered abuse detection and guidance

Overview

PROJECT TYPE

Individual Project

DURATION

8 weeks

MY ROLE

User Research

User Interview

UXUI Design

Prototyping

Usability Testing

TOOLS

Figma

Figjam

Slack

Challenges

41% of women and 26% of men experience verbal abuse from their partners.

However, victims often lose time trying to recognize the abuse and figure out how to respond on their own, which can make the situation worse.

Main Features

Analysis of verbal abuse patterns and users’ emotional states

Set goals together and practice coping strategies to help escape abusive situations.

Quick reporting without the abuser knowing

When verbal abuse is detected, a message is sent using a preset name and code. If the user replies with “1,” it automatically reports to 911.

An AI friend that helps solve problems together

After easy sign-up, it provides guidance on how to deal with verbal abuse whenever needed.

Design Process

Discover

Competitive analysis

App review analysis

User Survey

Affinity mapping

Key issues extraction

Define

Problem statement

Persona creation

User journey map

Storyboard

HMW statements

Prioritization

Selecting final direction

Develop

Ideation

Crazy 8’s

User flow

Low fidelity

Prototypes

High fidelity

Style guide

Deliver

Usability testing

SUS

Test analysis

Key issues

Iterations

Final UI refinements

Final prototype

Discover

Research

Partners, family and friends. Verbal abusers are often the people we see every day.

After reviewing more 5+ academic studies, I found that confusion and lack of guidance make it hard to leave abusive situations. Delayed response often leads to ongoing emotional pain and worse outcomes.

Literature Review & Academic Paper

Competitor research & App Review

However, victims are still left to find a way to escape their abusers on their own.

I reviewed 10+ verbal abuse apps and found that victims had to find information and report abuse on their own. I then read 100+ app reviews and community posts, grouped issues into four types to understand what problems people were facing.

Competitor research

"I'm constantly being monitored by my abuser"

"Is there an app I can use without getting caught?"

"Is this abuse?"

"How should I respond?"

Raw Data Collection (App Reviews & Community Posts)

Affinity mapping

User Survey

Survey results showed that 60% of users needed guidance on how to get out.

I surveyed four people who had experienced verbal abuse with simple questions. I focused on which was harder, noticing the abuse or knowing how to respond. This helped me understand what support people need most in that moment.

Survey results

Define

Persona & User’s Journey Map

When no one steps in first, delayed action leads to anxiety, depression, and dependence.

I focused on teenagers and low-income women — the groups most at risk based on my research. This showed that a lack of clear guidance and places to get help slows early response and helped me identify the most important need for early guidance.

Problem Statement

After recognizing verbal abuse, victims are left to respond alone without knowing what steps to take next.

HMW

How might we spot verbal abuse and guid what to do next?

Ideation

2x2 Matrix & MoSCoW Prioritization

Choosing solutions that protect legally, not just technically.

Many users requested recording features, but consent laws vary globally. Using a 2×2 matrix to weigh legal safety against user needs, I prioritized automatic voice detection that helps everyone without legal risks.

Develop

Sketches & Crazy 8’s

How do you help someone being watched by their abuser?

I sketched designs that look like normal chat apps to stay hidden. Through Crazy 8s, I explored ways to enable quick emergency reporting while keeping the interface unnoticeable to abusers nearby.

Character Design

I named it Lumaid, combining Light and Aid.

Like a friend who brings light in the darkness, the bright sky background offers hope during serious moments, while dark mode's starry night provides calm comfort.

Storyboard

Putting Myself in the User’s Shoes

I quickly drew a storyboard on my iPad while imagining real user situations. Seeing everything visually helped me understand the flow and pinpoint the problematic parts.

User Flow

Different danger levels need different responses.

I designed two flow versions to match urgent versus less urgent situations and focused on practical steps users can take with their AI companion.

Low fidelity design

Wireframes helped us test the structure before adding details.

Each screen was annotated to explain design intent, interactions, and accessibility needs.

Style guide

Looking normal keeps users safe.

I used soft colors and easy-to-read text to help users feel calm.

Deliver

High-fidelity design

One-tap
SSO login

Custom AI name
Hide from abusers

Global sign-up with
any phone number

Quick report button
on home screen

Your AI friend,
always here

Pick options to
make a safety plan

Dark mode
for chats

Find nearby
help centers fast

Chat with friends
No history saved

Auto-connect
your contacts

Check your situation & practice how to respond

Video call to check
if user is safe

Usability test

I ran usability testing with 7 participants from different age groups.

Task

Complete onboarding

Send a discreet help message

Respond to emergency detection alert

Review detected abuse and solutions

Task Completion Assessment

I checked how easily users completed each task.

During each session, I watched how participants used the product and how easily they finished tasks. When they struggled or failed, I asked short follow-up questions to understand what made the tasks difficult.

Task Result

Users felt relieved that Lumaid helped them first, and some became deeply engaged.

With an average SUS score of 72.5, it's easy to use. Some UI elements were a bit hard to see, but overall people really liked it. Users loved checking verbal abuse and status, reporting emergencies quickly without telling anyone, and the easy option selections.

Conclusion

Users could use Lumaid easily, but they needed simpler options and clearer visual cues for faster reporting.

Iteration

Based on data and feedback, I made the chat options shorter and easier to see.

I focused on helping users get help quickly and accurately. I improved the design to be visually clear at a glance and easy to use in stressful situations.

Simplifying options for faster contact

Feedback showed panicked users struggled with too many steps. I simplified options and enabled quick communication with trusted contacts.

Improving guide visibility

Users couldn't see the guide direction. I enhanced contrast for better visibility.

Adding text to report button

Users didn't recognize the report button. I added a text label to make it instantly clear.

Adding a back button

Users wanted to exit the chat screen. I moved the menu button to the right and added a back button.

Reflection

Takeaways

01 Understanding Users in Crisis

Users need a way to respond quickly and simply in emergencies. Through this project, I learned that too many options can actually increase confusion. I gained a deeper understanding of the importance of designing fast, intuitive flows that consider the user’s situation.

02 Prioritizing Core Features

Since not every problem can be solved at once, I focused on the most essential functions: detecting verbal abuse, sending secret signals, and enabling safe messaging. Within limited time and resources, I learned how to streamline key user flows and guide important actions naturally.

03 Improving Through User Feedback

Early usability tests revealed unexpected issues, which I was able to quickly iterate on. Reflecting real user voices taught me the value of designs that are immediately clear at a glance and reinforced the need for continuous improvement.

Next Steps

The goal is to expand practical ways for users to get support, like “responding with an AI friend.” I also plan to partner with domestic violence hotlines to enhance quick connection and response capabilities.