Designing an Adaptive AI Suggestion Interface for Distraction-free Task Completion
Role
UX Designer
Timeline
Apr 2025 3 weeks
Team
5 Students (including me!)
Skills
User Research Interaction Design UX Design AI Prototyping
TL;DR
Introducing the Suggestion Center: an ergonomic dashboard that prioritizes tasks over doomscrolling.
When using our phones, we multi-task between multiple apps, leading to distraction, stress, and cognitive overload. This over-active usage especially poses concerns while we are preoccupied—socializing, eating, traveling, and much more. In order to combat this, I conceptualized the Suggestion Center with my project groupmmates. With the promise of AI, we investigated how adaptive interfaces could be used to make task completion simpler and safer: less fumbling around for things and more focused on getting things done.
Periodically shifting AI suggestions anticipate your next task. Just tap one to engage in a task!
You can move tasks into the focus list above. With temporary and focused tasks, you can engage in tasks one at a time or simultaneously.
Highlights
Used AI to advocate for my ideas with AI-assisted prototypes.
LLM tools like v0 and Cursor allowed me to exercise my learnings in truly interactive prototypes that would be a nightmare to make in Figma.
Gathered key academic insights about adaptive interfaces.
Through secondary research with help from AI, I found that adaptive interfaces were actually quite a mixed bag–but if done in a way that users can predict, it can lead to efficient results.
Conceptualized a framework that allows for controlled adaptivity.
Leveraging stronger mental models about adaptivity, I was able to consider and construct a suggestion-based approach that would allow users to feel ownership and control over their task completion.
Motivation
I explored how adaptive UI could make our phones safer and emphasize their utility.
Interface Exploartions
This project revitalized a thought I've had for a while: why can't iPhones have built-in CarPlay?
We redesigned the interface to be more predictive and helpful, serving AI suggestions based on user context and goals.
Research Insights
By using AI to gather research papers, I realized that Adaptive UI is a mixed bag.
Key Insights
Adaptable interfaces: promise vs. reality
User control is essential - adaptable (customizable) interfaces were favorable in multiple metrics (1) - but too much requires a more system complexity and work for the user.
Building scalable, recognizable systems.
Changing UI is nice, but it shouldn't be from scratch every time. Defined component-based systems (2) serve to improve trust in the product.
On-demand suggestions are most helpful.
Interface adaptations shouldn't be direct or forceful—they need to be subtle and predictable (4). Adaptive UI has a lot to learn from contemporary studies on recommendations systems.
Secondary research is underrated, and AI helped me use it to make and share revelations with my teammates.
To discover valuable professional domain knowledge, I used Asta, a tool made by the Allen Institute for AI, which also operates Semantic Scholar. Asta was incredibly transformative in boosting my capacity to utilize relevant research to develop and advocate for important design considerations to the other members in my project.
Crafting the Suggestion Center
With prototypes on Figma, v0, and Cursor, I got my teammates to buy in to my vision.
Check out version 1 and version 2 of my vibe-coded prototypes. For the second one, you have to double-click suggestions to pin them.
The Job's Not Done
This was a class project—but I believe in this idea, and I'm still iterating in my spare time.
Reflection
UX concepts are often unrealistic, but I think that daring to wonder is beautiful.
Professor Haijun Xia (UCSD) showed me many videos from the past decades that envisioned the future of interfaces, and it was wild to see how off base they were. But I think we're getting closer.
While our approach is bold, the idea of using suggestions to guide tasks can be used more subtly.
While the Suggestion Center departs from current paradigms in order to eliminate distractions like doomscrolling, the concept of using suggestions to guide tasks can be leveraged in current interfaces.
I discovered MercuryOS after making this project (6 years too late) - it's made me even more invested in how I can take this idea further.
References
Selected work
Like what you see? Let's connect!