Yunbin Bae's Portfolio

startup spresto

0-1 AI UX Research & Strategy

to Build Trust.

How to Build User Trust in AI for High-Stakes Situations despite AI's Inherent Technical Limitations?

role

UX Research Lead

timeline

24 Weeks

Oct 2021 ~ Mar 2022

Type

Generative Research

Method

Contextual Inquiry

Feedback Analysis

Survey (Qualitative)

Literature Reviews

Concept Evaluation

Stakeholder Interviews

team

2 Product Designers

1 Product Manager (Me)

1 Lead Engineer

3 Developers

4 Marketers

overview

Create security solutions for hotel spycam epidemic by Building trust in AI to address user’s privacy concerns

business objective

Successful 0-1 launch of hotel booking platform with physical/cyber security solutions to protect guests’ privacy during hotel stays and meet initial 3 month business goals.

research focus

Validate Concept & product-Market Fit by uncovering the needs for a safe hotel experience
Build trust and confidence among women in their 20-30s to use our platform despite its imperfections

Discovery

Q4 2021

Exploratory

Q1 2022

Solutions

Q1 2022

research Question

Why are women still loyal to exploitative platforms, despite their negative sentiment about privacy violations?

insight

Set realistic expectations for users’ trust, as perceived “imperfections” in their technology deter adoption.

methods

Feedback Analysis: App Review (Sentiment & Thematic)
Contextual Inquiries
Survey (Qualitative)
Stakeholder Interviews

methods

Stakeholder Interviews
Literature Reviews
MVP Concept Evaluation

insight 1

Mediate user expectations on AI (that AI is not perfect!) to build initial trust.

insight 2

Embrace and highlight security solutions as a unique selling point, industry-first platform to address privacy concerns.

design solution

Integrate security solution in nav menu

Explainable AI: Prioritize AI method (patented) over others & grant easy access to user guides

Sense of Control: Handle model error feedback internally

Ethical AI: Eliminate AI personification

Advertise long-term automated solutions

How mIght we

Build trust & confidence among women in their 20-30s to use our hotel booking platform with security solutions, despite its imperfections?

product strategy

Addressing privacy concerns enhances, rather than diminishes, the joyful travel experience.

context

For 20+ years, With 95% of victims being women,

South Korea's spycam porn epidemic

has been covertly enabled by hotel booking industry prioritizing profits over privacy.

“I guarantee that I’ve been filmed naked at least once in a hotel room.”

— P5, contextual inquiry

behavior-needs gap

However, women in 20s~30s remained primary users of such exploitative monopoly platforms.

research question

Why are women still loyal to exploitative platforms despite their negative sentiment about them?

2018 national protest

01. Discovery

03. Solutions

02. Exploratory

insight #1 from literature reviews

Mediate user expectations on

AI/technology to build initial trust.

When designing for marginalized populations, i.e., women in their 20s and 30s, it's essential to foster a balanced sense of trust in technology, ensuring users neither blindly trust nor completely distrust it. This is particularly critical with AI, as its potential for exploitation poses risks to all users (humans), necessitating heightened caution.

by Akshay Kore

Design for Marginalized Populations

from “Designing Civic Technology with Trust (2021)”

by Christopher A. Le Dantec

Overtrust

Distrust

AI Capability

Trust in AI

Calibrated Trust

Historicize

Engagement

Focus on

Current Experience

Mediate expectations

with marginalized populations

Preserve Relationships

mvp concept evaluation

The MVP failed to effectively convey our unique positioning as the industry's first to address spycam concerns.

In MVP version 1, our security solutions, such as the spycam detection feature, were downplayed—only accessible on first-level screens after a booking was made.

stakeholder interviews

sPresto feared backlash from 'anti-feminists' or being seen as too political, potentially undermining the carefree, joyful experience.

Technical PM

@sPresto

We can’t risk losing potential customers who may not care about this issue; because anyone can book hotels!

insight #2

Embrace and highlight our security solutions as a unique selling point, industry’s first platform to address privacy concerns.

research goal

Uncover the needs for a safe hotel experience

to validate concept & product-market fit

research strategy

Utilize attitudinal insights from an existing app (spycam detection)

to guide direction for our hotel booking platform launch

Restricted by company policies from engaging with wider networks,

I conducted research on a spycam detection app, leveraging its existing user base and feedback on safety features. This was valuable as both platforms target individuals with heightened privacy concerns, particularly in high-risk environments like hotels.

Spycam Detection App (Launched)

Hotel Booking Platform

with security solutions

(To-be-launched)

Feedback Analysis

Reviewed 400+ spycam detection app’s Feedback for Sentiment & Themes Analysis

After reviewing 30+ sPresto’s PR & news articles, I analyzed feedback on spycam detection app (2.6 ratings) to identify main sentiments as Fear, Doubt, Helplessness and main themes as distrust (and even hostility) against society, business, and technology.

contextual inquiries

Inquiries about 20s-30s women’s experience in public spaces and spycam detection app revealed the core reasons of the distrust.

Recruitment

Korean Women in 20-30s recruited from survey.
Survey participants are from:
- 20-30s South Korean national feminist online community comprised of female-verified users
- Female college students & recent grads

Process

Critical-Incident Technique

Behavior Observation - spycam detection app

In-Depth Interviews about user experience, frustrations, unmet user needs, etc.

stakeholder Interviews

Contrary to sPresto’s expectations,

users disapprove of their technology & business model.

Making us pay for detecting spy cams? This is ridiculous. CEO is male and it exploits women’s fear to make money out of it.

Contextual Inquiry

Participants

CEO @sPresto

In high-stake scenarios i.e., privacy in public spaces...

finding #2

Users expect perfect, “cure-all” solutions in order to trust and use AI/technology. Otherwise, users tend to avoid using it entirely.

“Anyways, this detector app is never a cure-all for this societal issue.”

“I won’t bother myself detecting public bathroom. Perhaps because I’m lazy… or maybe I’m less sensitive to this issue than I thought I was...”

“It is draining to detect spy cams every time I go to public places.”

finding #3

Users often question the integrity of the business model especially
if AI/technology appears imperfect.

“I am very skeptical about this app’s technology. Does it merely use smartphone built in camera feature, deceiving users?”

“OMG, this app is a scam. Spycams are bad but the people who exploit our fear for molka are way more evil!”

“This must be a scam. They don’t allow you delete your account.”

“Just by looking at the Home feed, I’m sure man built this this app!”

insight

Set realistic expectations for users’ trust,

as perceived “imperfections” in technology deter user adoption.

how might we

Build trust and confidence among women in 20-30s to use our hotel booking platform with security solutions, despite its imperfections?

After the app launch,

it rapidly gained

1 Million +

Downloads in 3 months.

4.9/5.0

Appstore ratings in 3 months.

$5.6M

Revenue generation from hotel bookings.

132%

increase in user satisfaction after spycam detection integration in navigation tab.

Reflecting on my

AI UX Project...

Calibrate user trust on AI

I learned the importance of mediating user expectations on AI. Ever since ChatGPT opened a new AI era, it is becoming more crucial to design AI for humans.

Plan ahead for longitudinal research

AI evolves as users engage with it over time. As a result, users’ mental models about how it works as they engage with it over time may change (it’s a continuous loop).

I would ensure to offer AI solutions of value by understanding how mental models change over time and evaluate AI performance as users and use cases increase.

Future Research Directions

Thank you

to Dong-Hyun Son for having me in your company and believing in my ability to make impact when it was my first time working in tech; to Chisu and Sung-hwan for having me in your team and giving me honest feedbacks and guidance when needed; to our beloved designer Juwon and engineer Nijat!

It was such a joy to work and learn something new every day at sPresto :)

She is an exceptional professional—passionate, diligent, and motivational. I hired her for her initiative as a UX Researcher and promoted her to Product Manager within a month. Through her data-driven strategies and user-centric approach, she transformed customer skepticism about our mission into glowing reviews, with users praising our empathy for their privacy concerns. She excels at aligning user needs with business goals while fostering a positive, collaborative team environment. Anyone who works with her will benefit from her remarkable research skills and leadership.

— Dong-Hyun Son, Chief Executive Officer @sPresto

“

”

Pre-ChatGPT Era: My First AI UX Research Project (2021–2022)

when AI UX research was still uncommon, I focused on designing for marginalized groups. This approach felt natural, as humans, especially vulnerable populations, are particularly susceptible to AI’s influence.

Persona / User Segmentation

product strategy

Addressing privacy concerns enhances,

rather than diminishes, the joyful travel experience for women in 20-30s.

Based on research, I developed MVP ver. 2 to prominently showcase our security solutions.

design solution #1

Discoverability: Integrate spycam detection in navigation menu’s 3rd tab

Highlighted unique selling point & increase discoverability.

insight from MVP concept eval.

Embrace & Highlight our security solutions as a unique selling point, industry-disruptor that address critical user needs, privacy concerns in hotels.

design solution #2

Explainable AI: Grant easy access to user guides while prioritizing AI method which has better accuracy

Managed user expectations from 100% (=perfect) accuracy to a realistic level, achieving 4.9 ratings with positive reviews.

insight from literature reviews

Mediate user expectations on AI/technology to build initial trust.

design solution #3

Sense of Control: Handle model error feedback internally, not through public sharing

Reduced negative sentiments and misinformation, improved the credibility of the spycam detection app, resulting in a 150% increase in app ratings (2.8 -> 4.2) & higher user satisfaction reflected in reviews.

insight from Contextual inquiry

“I am very skeptical about this app’s technology. Is it deceiving users by merely using built-in camera feature in smartphones?”