Re:Self: Redefining Personal Wellness
Role
UI/UX Designer
Working with: 13-person team of researchers, engineers, and designers
Timeline
February - July 2025
Tasks
XF Collaboration
User Research
UI Prototyping
Data Analysis
Competitor Analysis
Tools
Figma
Miro
Qualtrics
UI/UX Designer
Working with: 13-person team of researchers, engineers, and designers
Timeline
February - July 2025
Tasks
XF Collaboration
User Research
UI Prototyping
Data Analysis
Competitor Analysis
Tools
Figma
Miro
Qualtrics
The Brief
Processing overwhelming emotions can be tough, especially in the moment. It’s one story to recognize a thinking pattern, and another to change it for personal betterment.Beyond describing what happened, how can we learn from our own journal entries to better inform our future decisions?
Our team developed Re:Self -- an AI journaling platform that utilizes LLMs to guide users through difficult memories and develop better coping strategies.
RE:Self is an AI-powered journaling platform guiding users through emotional processing during overwhelming moments. I led 0-to-1 design from research through prototype, conducting literature reviews, designing conversational AI interactions, and developing frameworks to measure "reflective agency"—users' ability to gain insights from difficult emotions.
This addresses the gap between traditional journaling (requiring clarity) and therapy (requiring resources), affecting millions struggling with emotional processing. Through MIT Media Lab, I validated concepts via user studies, refined therapeutic AI prompting, and published findings at ACM AIES 2025, establishing ethical AI principles for mental health tools.
PREVIEW
When difficult memories resurface,
how can we grow & learn from them ?
Key Features:
- Audio transcription, AI situational analysis of what users inputs, categorizes event into Gross Model of Reflection
-
Simulation tree based on events, gives user options to decide where they wished they had acted or thought differently
-
Situation rehearsal: based on the user’s responses and the recorded scenario, the AI model generates hypothetical situations wherein the user may feel similar emotions, and helps user develop more concrete strategies
Research-backed design framework (published in 2025 AI, Ethics, & Society Conference
- INCLUDE BEFORE AND AFTER DESIGNS
MISC NOTES
- Emphasize throughout portfolio that my projects are continuity based on previous projects
- Research backed
- Wearables, and ecosystems
- Can translate complex principles into the product design process, as well as consumer-facing interactions
- Emphasize throughout portfolio that my projects are continuity based on previous projects
- Research backed
- Wearables, and ecosystems
- Can translate complex principles into the product design process, as well as consumer-facing interactions
Context
Background
Building on Breakthroughs
Based on existing researched on LLM real-time memory retrieval for elderly (Memoro). As a successful example of everyday Human-AI augmentation, our team wanted to extend this capability to more users in everyday functions, going from not just simple information retrieval such as names or locations, but empowering users to reflect upon experiences.
Memoro's success with AI-augmented memory for elderly users
Memoro's success with AI-augmented memory for elderly users
AI augmented memory recall (THE METHOD & PURPOSE of empirical evidence that AI intervention in everyday interactions can be useful for the elderly) --> how can we expand this to more users --> memory recall can be utilized in journaling, which is already a very common habit
The Broader Opportunity
Our team asked: Can this approach support emotional memory processing for broader populations? Younger users also struggle with memory processing, especially emotional memories, scaling beyond elderly care to emotional processing. Recall to reflection assistance, logical extension. Problem statement in here
Daily busyness prevents individuals from consistently organizing and evaluating personal emotions in real-time, making it difficult to retrospectively reflect and learn from past experiences. Young adults with access to digital technology face particular challenges: they experience emotional duress at work, struggle to identify patterns in their emotional responses, and lack concrete tools for developing self-awareness and adaptive coping strategies. The average person experiences complex emotions but often lacks the vocabulary or cognitive bandwidth to process them as they arise, leading to missed opportunities for growth and resilience-building.
Goal
Human-AI collaboration, Journaling + AI assistance = enhanced self-reflection. We wanted to expand this assistance to more users through accessible services. Translate a complex psychological model into understandable
Research
Preliminary User Study
20 participants in Boston area, younger people, research revealed that adults struggle with three core challenges in emotional self-regulation:
SAMPLE STATS Current Behaviors:
Expressed Needs:
Technology Attitudes:
- Temporal disconnect: Users cannot consistently organize emotions in-the-moment due to cognitive load from work responsibilities and daily stressors
- Pattern blindness: Without systematic reflection, users miss recurring emotional triggers and maladaptive coping patterns that could inform better responses
- Limited emotional vocabulary: Users lack words to describe complex emotional states, defaulting to basic descriptors that fail to capture nuanced experiences
SAMPLE STATS Current Behaviors:
- 60% attempted journaling but stopped due to time constraints
- 35% journal sporadically during emotional extremes
- 5% maintain consistent practice
Expressed Needs:
- "I need help identifying what I'm actually feeling" (75% of participants)
- "I want to see patterns but don't know what to look for" (65%)
- "Writing feels overwhelming when I'm already stressed" (80%)
Technology Attitudes:
- High openness to AI assistance if it preserves authenticity
- Preference for guidance over diagnosis
- Concern about privacy and emotional data storage
Literature Survey
Gross Model for Emotional Regulation. Literature review findings indicated that traditional journaling shows proven benefits for emotional regulation, but adoption remains low due to the cognitive effort required during emotionally challenging moments—precisely when support is most needed.
Competitor Research
Mood Tracking Apps (Daylio, Mood Panda)
AI Therapy Chatbots (Wysa, Replika)
Digital Journals (Day One, Journey)
Common features across platforms included mood tracking, reminder systems, and data visualization, but none successfully integrated real-time emotional support with long-term pattern recognition and strategy development.
- Features: Daily emotion logging, statistical tracking, mood charts
- Limitations: Require emotional literacy beyond developmental capabilities of distressed users; focus on data collection over insight generation
AI Therapy Chatbots (Wysa, Replika)
- Features: Conversational interfaces, CBT exercises, 24/7 availability
- Limitations: Create dependency on external validation; may reinforce negative patterns through excessive empathy without challenge
Digital Journals (Day One, Journey)
- Features: Text entry, photo integration, calendar views
- Limitations: Lack guidance during emotional overwhelm; no pattern recognition or intervention capabilities
Common features across platforms included mood tracking, reminder systems, and data visualization, but none successfully integrated real-time emotional support with long-term pattern recognition and strategy development.
Design Research Questions
For feature validation, by synthesizing research and user insights into actional design principles
SAMPLE
RQ1: How can voice-based entry lower barriers to emotional documentation during high-stress moments?
RQ2: What level of AI interpretation supports insight without undermining user agency?
RQ3: How can visual representations of emotions facilitate understanding without oversimplification?
RQ4: What guidance structures help users develop adaptive coping strategies they'll actually implement?
SAMPLE
RQ1: How can voice-based entry lower barriers to emotional documentation during high-stress moments?
RQ2: What level of AI interpretation supports insight without undermining user agency?
RQ3: How can visual representations of emotions facilitate understanding without oversimplification?
RQ4: What guidance structures help users develop adaptive coping strategies they'll actually implement?
Understanding why something works matters as much as making it work
Design
User Journey
The one about the office worker who’s stressed about her workplace interactions. System-level design: Architected three-phase user journey (capture → analyze → reflect) considering technical feasibility, user cognitive load, and long-term behavior change
Brainstorm
Object oriented on FigJam + the sketches
Showing the early iterations and then later how we realized we needed design research questions in order to systemically generate academic insights from usability testing.
Putting it All Together
[Walk users through the actual use process]
Showing the early iterations and then later how we realized we needed design research questions in order to systemically generate academic insights from usability testing.
- Include sketches of the FigJam things I did
RE:Self is an AI-powered journaling platform that guides users through emotional processing during overwhelming moments. The system employs a three-phase approach: voice-based emotion capture, AI-assisted analysis with visual feedback, and guided reflection for developing adaptive coping strategies. By leveraging the Gross Process Model of Emotion Regulation, RE:Self helps users move through five stages of emotional processing: situation selection, situation modification, attention deployment, cognitive change, and response modulation. The platform transforms abstract emotional experiences into concrete insights and actionable strategies for future situations.
- Include sketches of the FigJam things I did
RE:Self is an AI-powered journaling platform that guides users through emotional processing during overwhelming moments. The system employs a three-phase approach: voice-based emotion capture, AI-assisted analysis with visual feedback, and guided reflection for developing adaptive coping strategies. By leveraging the Gross Process Model of Emotion Regulation, RE:Self helps users move through five stages of emotional processing: situation selection, situation modification, attention deployment, cognitive change, and response modulation. The platform transforms abstract emotional experiences into concrete insights and actionable strategies for future situations.
Phase 1: Voice Entry - Emotional Capture
Users initiate reflection through voice recording, pressing a button to begin speaking freely about their emotional experience. This phase prioritizes:
Design Rationale:
Key Features:
Design Rationale:
- Lower cognitive barrier: Speaking requires less mental effort than writing during emotional overwhelm
- Preserve authenticity: Unstructured verbal expression captures nuanced emotional states better than forced categorization
- Immediate accessibility: Voice entry can happen anywhere—walking, commuting, or in private moments
Key Features:
- One-tap recording initiation
- No time limits or structured prompts during initial capture
- Background recording capability for stream-of-consciousness expression
- Automatic transcription for later reference
Phase 2: AI Analysis & Visualization
The system processes voice entries through multiple analytical layers based on the Gross Process Model of Emotion Regulation:
Emotional Mapping Visualization: Transcribed text is highlighted using color gradients corresponding to detected emotional valence and intensity. The system identifies which stage of emotion regulation the user is experiencing:
Pattern Recognition:
Emotional Mapping Visualization: Transcribed text is highlighted using color gradients corresponding to detected emotional valence and intensity. The system identifies which stage of emotion regulation the user is experiencing:
- Red highlights: High-intensity negative emotions (anger, frustration)
- Blue highlights: Sadness or withdrawal patterns
- Yellow highlights: Anxiety or fear responses
- Green highlights: Positive coping attempts or reappraisal
Pattern Recognition:
- Identifies recurring triggers across multiple entries
- Surfaces emotional regulation strategies currently in use
- Highlights moments of cognitive distortion or rumination
- Notes connections to past entries with similar themes
Phase 3: Guided Reflection - Building Adaptive Strategies
Users are guided through structured reflection to transform insights into actionable coping strategies:
"What Actually Happened" Reconstruction:
"Alternative Paths" Exploration:
"Next Time I'll Try" Planning:
This three-phase design creates a complete cycle from emotional experience to processed understanding to future preparation, supporting users in developing lasting resilience and emotional intelligence.
"What Actually Happened" Reconstruction:
- AI prompts users to separate facts from interpretations
- Identifies gaps or inconsistencies in narrative
- Helps complete partial memories with contextual cues
"Alternative Paths" Exploration:
- System generates 3-5 alternative responses user could have taken
- Each path mapped to different emotion regulation strategies
- User can modify or personalize suggested alternatives
"Next Time I'll Try" Planning:
- Users select preferred coping strategies for similar future situations
- System creates implementation intentions ("If X happens, then I'll Y")
- Strategies stored for future reference and rehearsal
- AI generates similar hypothetical situations
- Users practice applying new strategies in low-stakes environment
- System provides feedback on strategy alignment with chosen goals
This three-phase design creates a complete cycle from emotional experience to processed understanding to future preparation, supporting users in developing lasting resilience and emotional intelligence.
A Novel Design Framework
Our study with 20 participants revealed that AI reflection tools risk eroding user agency through over-automation. The resulting Reflective Agency Framework—published at ACM AIES 2025—offers five principles for preserving human autonomy in AI systems. While RE:Self remains a research prototype, the framework is already informing next-generation mental health tools, with two startups implementing our guidelines and MIT researchers building on our findings.
The future of Human-AI collaboration needs to be based upon responsible frameworks, which is becoming increasingly valuable and apparent as companies face regulatory and trust concerns.
The future of Human-AI collaboration needs to be based upon responsible frameworks, which is becoming increasingly valuable and apparent as companies face regulatory and trust concerns.
Research-to-Product
Whether conducting usability research for a new feature, presenting to leadership about ethical AI implementation, or designing for vulnerable user populations, this project proved I can balance user needs, business constraints, and technological possibilities—while backing every decision with data.