An LLM-Powered Self Reflection Tool
Timeline
3 Months
Tools
Figma
Python
Flask
SQLite
HTML/JS/CSS
GSAP.js
P5.js
Collaborators
Yash Mittal
Keywords
Product Design
Web App Dev
AI Product
UX Research
**Please note that this page is still being updated 🛠️**
Every day, humans produce an astonishing amount of valuable information—our thoughts. According to the National Science Foundation, each person generates between 12,000 and 60,000 thoughts daily. Imagine harnessing this untapped reservoir of data to enhance self-awareness.
Introducing Penceive AI.
Landing Page / cycling thoughts and hover interaction animated with GSAP.JS
A serene companion for capturing fleeting thoughts and transforming them into a beautifully organized tapestry of self-reflection. A pin-like wearable captures your spoken words on the go, effortlessly transcribing and enriching them through advanced language processing. Dive into a calming interface where you can sift through, categorize, and explore your ideas, accessing a calendar of your mental journey.
Systems map / How it works
Recollect and reflect– navigate your thought timeline
Thought Timeline / accessing your thoughts through a calendar – built using p5.js.
Users can explore their thoughts chronologically, accessing entries by the date to reflect on their daily mindset. The interface features a dynamic particle system, symbolizing scattered thoughts coalescing into a coherent whole. A subtle breathing animation gives the design a living, organic feel, enhancing the reflective experience.
Ask your past self a question – search through your thoughts.
The Process
THE VISION
From the outset, we understood the ethical responsibility inherent in the role Penceive would play in users' lives. As AI becomes increasingly integrated into our daily routines, it is essential to design interactions that enhance, rather than diminish, our humanity. Penceive is not intended to replace mental health professionals or serve as advisors to users. Instead, our goal is to create a tool that makes self-reflection more accessible, empowering users to become more self-aware. We aim to build a non-intrusive companion that listens like a trusted friend, not to offer opinions, but to encourage meaningful self-reflection.
INITIAL IDEATION
The idea for Penceive was inspired by the concept of the 'Mind Palace' from the Netflix show Sherlock, which I then connected to the 'Pensieve' in Harry Potter. Both concepts stem from the same underlying goal: accessing and organizing thoughts in a way that benefits the user. Memory is a vital tool for adding context to our decision-making processes, and self-reflection allows us to process these memories, helping us make sense of the overwhelming thoughts we generate.
EXPERT INTERVIEWS
We began by engaging with a diverse range of experts, including professors in Human-Computer Interaction, Human-Centered Design, Psychology, and Cognitive Science. These conversations helped us understand the implications of our proposed concept. Through these discussions, we gained valuable insights and identified potential risks our technology could pose to users. This process allowed us to define the scope of our product and proceed with caution.
From our conversations, these were some of the important points we started to thinking about:
Data Privacy
Data collection and storage would require robust measures as user’s would be inputing extremely personal data.
Technical Feasibility
How do we create a system that would be able to host a personalized LLM that would be access the user is constantly updating?
User Interface
How do we create an interface that encourages consistent self-reflection? How do we display information to help user's gain more from the data?
Testing Mechanism
How do we test out our ideas and hypothesis for a novel concept like this?
Unforeseen implications
User's would use our technology to recall events and emotions – how do we create a friendly AI tool that would not change the narrative that people remember
Ensuring User input consistency
How do we make sure that User's make consistent entries into Penceive without being intrusive? How do prompt them effectively?
USER RESEARCH THROUGH LO-FI PROTOTYPING
To gain deeper insights into user interaction with our proposed concepts, we provided participants with dictaphones and asked them to carry them at all times for a week. We encouraged them to record any thoughts or observations they deemed noteworthy.
Using the Whisper.ai API we transcribed the entries. We took anonymized entries from the dictaphones to understand the kinds of user
Manually categorized thoughts