CultureCompass

Inspiration

Mental health conversations are still heavily shaped by cultural expectations, family perceptions, and social stigma. Many young people want to open up, but fear being misunderstood or judged based on “what the culture expects.” This project was inspired by the gap between technology and culturally-sensitive emotional support. Existing chatbots are often generic, lack cultural nuance, and fail to adapt to different upbringing styles. I wanted to build something that understands culture, context, and emotions — a supportive space that speaks the language of the user’s lived experiences.

What it does

CultureCompass is an AI-powered assistant that provides:

Culturally-aware emotional support

Conversation-based stigma reduction

Personalized guidance based on background, traditions, and values

Resource suggestions for coping, communication, and self-awareness

Journal-style reflections to help users express what they cannot openly discuss

Anonymous, judgment-free conversations

It acts as a safe bridge between the user’s inner struggles and the cultural environment they live in.

How we built it

CultureCompass combines AI, NLP, and curated cultural datasets:

Tech Stack

AI Model: Open-source LLM (Llama / Mistral) fine-tuned for cultural dialogues

Backend: Python (FastAPI)

Interface: React + Tailwind

Database: MongoDB for journaling & anonymized responses

Deployment: Hugging Face Spaces + Vercel

Architecture

User starts a conversation

Input → NLP layer extracts emotional tone + cultural markers

AI generates culturally aware responses

Recommendations engine provides coping tips

User receives reflections + resources

flowchart TD A[User Input] --> B[NLP Emotion & Culture Detection] B --> C[AI Response Generator] C --> D[Recommendation Engine] D --> E[Frontend UI Response]

Challenges we ran into

Cultural bias in models: Many pretrained models default to Western emotional norms.

Maintaining empathy while being culturally neutral: Hard to balance non-judgmental responses vs cultural sensitivity.

Dataset limitations: Very few open datasets describe culture-based emotional behavior.

Real-time sentiment correction: Handling sensitive topics requires safe, accurate interpretation.

Time limitations: Building a prototype with meaningful depth in a short window.

Accomplishments that we're proud of

Built a functioning culturally-aware chatbot within the hackathon timeline

Successfully implemented emotion + cultural intent detection

Achieved smooth real-time conversation flow

Designed a clean, accessible UI

Created a foundation for a socially impactful tool addressing stigma in mental health conversations

What we learned

How deeply culture influences mental health conversations

Importance of removing bias from AI responses

Designing for empathy first, technology second

Value of creating safe, private digital spaces

How modern LLMs can be adapted for niche, culture-specific use cases

How to integrate NLP pipelines with lightweight frontend frameworks

What's next for CultureCompass

🧩 Add multilingual support (Tamil, Hindi, Bengali, etc.)

🧠 Improve cultural-awareness dataset with community-driven contributions

📱 Launch a mobile app with offline journaling

🔒 Advanced privacy framework for sensitive conversations

💬 Group-based cultural community rooms

🩺 Partnerships with mental health professionals

🎯 AI Therapist Mode with guided self-help sessions

🎨 UI personalization based on cultural themes

Built With

• javascript
• python
• react
• tailwindcss
• vite