Paper
Demo
01. Could taking a more diverse perspectivelead to a better solutionfor short-form video addiction?
02. Shorts are fun and give us quick information.But is there a way to enjoy them, and then sill get back to real life easily? Is there a way to return to real life with ease afterward?
If I think about it in terms of stimulation, customers start light with the appetizer, get more stimulation with the main dish, and finish smoothly with dessert.
So, I decided to apply the same idea to the way people watch shorts.
I made something called Video Omakase — a course-style playlist.
Applied the same idea to the way people watch shorts.
It starts and ends with lower-stimulation videos, and puts more exciting ones in the middle.
The goal is to help users enjoy shorts and then smoothly return to what they need to do in real life.
How can we find the low stimulus content?
Habituation
So I turned to a psychology concept called habituation.
It explains that when people are repeatedly exposed to the same type of content,
their brain releases less dopamine, which makes the content feel less stimulating over time.
But during our pilot test, I found an issue.In the middle of the course, I had included highly personalized new shorts — the kind platforms usually recommend. For some participants, these videos were so engaging that they said it felt like falling back into the usual shorts-watching loop.That’s when I realized — the difference in stimulation between familiar and new videos shouldn’t be too big.
System Flow
So I used Gen AI to help us discover new shorts that feel less stimulating.
- First, we collected the user’s watch history on a specific topic. (E,g,. a user is a fan of Beyoncé and the user watched lots of Beyoncé-related shorts)
- Then I asked ChatGPT-4o to extract broader or less specific keywords from the metadata, like “dance tutorial” or “woman singer,” which are related to Beyoncé but less directly.
- Then, I used those keywords to search for new videos through the YouTube Search API.
- Finally, I combined those new videos with the ones the user had already seen to create the full Omakase course.
The user selects a topic and a viewing duration, and the system generates an Omakase-style video course.
It starts and ends with familiar Shorts while the middle section introduces loosely related new content.
After the set duration, the curated Omakase session automatically ends.
Ran a user study with 30 participants who experienced three viewing modes:
the standard YouTube algorithm (A),
a low-stimulus playlist of previously watched videos (B),
and our Video Omakase system (C).
After each session, participants rated satisfaction (boredom scale) and how easily they could stop watching (stop-intention).This allowed us to evaluate both enjoyment and the system’s ability to support healthier disengagement.
Participants
- 15 self-identified shorts addicts
- 15 regular viewers (non-addicts)
Design
Participants were exposed to 3 types of short-form video playlists:
A. YouTube Recommendation Course
B. Low-Stimulus Course (previously watched videos only – based on habituation theory)
C. Omakase Course (curated balance of novelty & familiarity)
Measures Collected
- Boredom Level (1: Not bored at all, 5: Extremely bored)
- Stop Intention (1: Couldn’t stop watching, 5: Could easily stop)
- Qualitative Insights (open-ended feedback)
(A) Stop Intention
1: Couldn’t stop watching
5: Could easily stop
B > C > A(Significant: A vs C (p=0.028),
B vs C (p=0.029))
- B was easiest to stop.
- C balanced between A and B.
- A made it hardest to stop watching.
(B) Boredom Level
1: Not bored at all
5: Extremely bored
B > C > A (Significant: A vs C (p=0.032),
B vs C (p=0.015))
- B caused the most boredom.
- C was moderate.
- A was least boring.
- Course A was described as highly stimulating, making it hard to stop watching.
- Course B felt repetitive and less engaging due to familiar content.
- Course C was seen as a balanced mix—less intense than A, but more interesting than B.
Why Beta Test?
The Chat+ PC version is a new feature being integrated into Samsung's One UI 6. Thus, it had to be includedas part of Samsung's One UI 6 beta test program.
Why Focused Group Interview?
I conducted FGI to gather user perspectiveson future features, which was a limitation in the beta test.
Duration:Sep 6, 2023 - Sep 26, 2023
Total Applicant: 332
Registration Method: Available through Samsung Members application after downloading Samsung's new OS version (One UI 6.0)Key Survey
Questions: Inquiries about improvement suggestions for existing features and potential new features
1. Due to the characteristic that testing was only possible on PC, unlike other One UI features, there was relatively super low participation to find meaningful insights.
2. Compared to other chat services, users faced multiple service restrictions (including limited PC compatibility, 500-message daily limit, and messaging limitations with iOS users), resulting in a high volume of usage-related inquiries.
3. Split management between Samsung (mobile) and SK Telecom (PC) led to delays in issue identification and resolution.
4. Limitations of voluntary participation and feedback collection.
...
Despite these constraints, additional research was conducted
by adopting a research methodology to address the major issues identified during the beta test and by implementing stricter researcher criteria.
01. Participant Recruitment and SelectionTo address low beta test participation rate, user research was conducted targeting loyal mobile service users, with interviews organized by four major occupational groups identified from top 10% users, anticipating active participation due to their familiarity with the service and potential to build a cross-platform loyal user base. A total of 29 participants took part in interviews across four occupational groups.
02. Pre-task Instruction ProvidedTo overcome the limitations of PC-only testing and voluntary participation experienced during the beta test, I introduced preliminary tasks aimed at improving service understanding and discovering use cases. These pre-testing activities were complemented by detailed usage guides, with appropriate compensation provided to encourage participant engagement.
1. Installation Guide
2. Usage Limitations: Daily message limit: 500 messages, iOS users cannot receive messages
3. Feature-specific Usage Guides: Message sending (group, 1:1, broadcast), Trying out the settings feature, Synchronization management, etc
4. Occupation-specific Use Case Guides: Please use this app during your work hours throughout the test period
Affinity Diagram analysis of FGI sessions revealed significant user concerns about malicious messages on Chat+ PC, with more than half of participants reporting difficulty distinguishing harmful messages from legitimate ones and experiencing higher rates of malicious content compared to other services, leading some users to consider switching platforms.
- Plan to develop a rating method to quantify stimulation levels and test its effectiveness in future experiments.
- We plan to explore scenarios where these features could be more impactful.