Plugin + Website
Pufferfish: An AI privacy protection plugin allowing users to control + learn about their data
roles
Product Designer
Researcher
tools
Figma
Figma MCP
Cursor AI
Claude
team
5 UX Researchers
1 Product Designer
overview
A research-first 0-1 plugin, deployed with Cursor
I collaborated in an interdisciplinary team of 6 to research user behavior in digital and AI spaces. Our findings indicated that users were concerned about their data, but knew very little about their rights with respect to it. To bridge this gap, our team proposed a privacy-focused plugin focused on transparency, user privacy protection, and ethical data use practices.
After we concluded research with the proposed design solution, I independently designed and deployed the final product.
Design Process
* Research Problem Space
* Define Insights
* Ideate
* Prototype
* User Testing + Insights
* Iterate
* Deploy
Research Methods
We conducted extensive early-stage research to scope out core issues and underlying assumptions with users and their relationships with AI and privacy. I was responsible for sourcing people for the legal and corporate sector of the interviews, conducting 4 of the 18 interviews with 2 legal attorneys, 1 software engineer, and 1 designer.
12
Think-aloud sessions using directed storytelling
18
User interviews across young adults and adults from varying backgrounds
01
Survey
2000
Responses analyzed
Insights + Research
Overall: Privacy distrust in the context of AI is both a cognitive perception issue (trust and knowledge) and an interaction issue (navigation and controls)
Insight 1: Privacy Controls were difficult to understand.
Users struggled to identify a clear path towards and understand what various privacy controls meant.
Traditional gifts
+
+
Insight 2: Most users lacked privacy knowledge, and as a result, assumed either the worst or best about the use of their data.
We initially assumed that the people we spoke to knew about AI and privacy, which informed their pessimism. We found many assumed the worst, like that privacy controls would not exist, but did not know about the topic. Making an opportunity to educate is equally as important to trust-building as the platform’s UX.
Research
-
Insight 3: Distrust reduced product adoption and negatively impacted users’ overall perception of the parent company
Participants reported limiting and avoiding AI product use due to their fears. Users need immediate, clear assurance about what's protected.
Research
-
Ideation
Concept Definition and Information Architecture
We translated research insights into product directions.
User Testing + Iteration
Plugins + Websites + Editing on Other Interfaces
After prototyping a few potential means of interaction between the plugin and text input, I conducted A/B testing to understand how users would understand the distribution of content.
Testing
-
Initial user feedback on the flows of <scheduling envelopes of gifts + messages> revealed confusion around whether the system sent actual mail + if so, how our app was different from typical mailing.
Iteration
-
We reassessed our PRD and realized there was a need to pivot in our design language + core features.
* Our design language relied too heavily on envelopes, leading to user confusion of function
* The original mail-sending only addressed (1) facet of relationships: stabilizing future communications, but not cementing the present or recalling shared pasts.
Design System
Emphasizing clarity with subtle playfulness
Designed to reduce anxiety; studied the Google Design System to maximize legibility.
the future
Packets Create Moments of Shared Joy
Users schedule-send digital packets containing movies, audiobooks, photos, messages, or music to others or themselves in the future.
the present
Build Events Together— they're not shared with the Public
Store events with candid pictures, music that reminds you of what happened, an auto-generated map of your journey, and more!
All events are automatically saved to you + your friend's Shared Journey. No need to go through the trouble of sending images, updating photo albums, sharing permissions, and more after a hangout.
the past
Your Home Screen: Journeys synthesize Packets + Events into Shared Narratives
Combines the previous two features into one scroll per friend, turning fragmented interactions into cohesive, multidimensional collections you can both review and recall :).
Reflection
Plugin Interactions, Vibe coding, and Research
I didn't initially realize I would be introducing an unfamiliar form of interaction to users. How could I induce them to click and manipulate in-chat text while potential interactions existed in the widget separately as well? Most interactions in apps and websites are physically grouped.
Moving forward, I’m eager to rigorously probe at layouts to investigate if they might require novel flows for interaction before designing, always consider information distribution, and conduct A/B tests for potentially novel layouts earlier in the design process.
I learned to blend design thinking with crafting a functional plugin through Cursor AI, Claude, and Figma MCP. Being able to test my prototypes in code led to an unexpected bonus: I caught edge cases I hadn’t previously thought of, and was able to adapt my designs accordingly.
