Inspiration
Mental health is a critical but often neglected aspect of healthcare in Africa. Many communities face significant barriers to seeking help, including social stigma, a lack of accessible and affordable mental health professionals, and a strong cultural preference for discussing personal issues within a trusted community. The traditional approach of clinical therapy can be daunting and inaccessible. Our inspiration comes from the Akan concept of "Sankofa", which means "to go back and get it." We want to create a tool that honors this idea by leveraging traditional wisdom and community values to make mental health support feel familiar, safe, and culturally relevant. We believe that by integrating AI with this approach, we can provide an immediate, private, and stigma-free first step toward wellness.
What it does
The Sankofa Listener is a text-based, AI-powered chatbot that offers compassionate and culturally-sensitive mental health support. It provides users with a safe, anonymous space to express their feelings and concerns via a simple messaging interface. The AI is trained to understand nuances in language and can:
- Offer empathetic responses and active listening to validate users' feelings.
- Provide simple, actionable coping mechanisms and mindfulness exercises grounded in traditional wisdom and local proverbs.
- Safely guide users toward local resources by connecting them to verified community support groups, elders, or mental health professionals.
- Identify signs of a crisis and provide immediate contact information for local emergency helplines.
The platform is designed to be accessible on basic smartphones and works with low bandwidth.
How we built it
We'll build the Sankofa Listener using a transformer-based language model fine-tuned on a carefully curated dataset. This dataset will include conversations with mental health professionals, transcripts of support group sessions, and a body of text containing African proverbs, folklore, and cultural narratives related to resilience and community.
- Technology Stack: The core will use a fine-tuned model a custom-trained open-source model). The user interface will be a simple web application or a Telegram/WhatsApp bot for maximum accessibility.
- Data Strategy: We'll partner with local mental health professionals and cultural experts to ensure the AI's responses are accurate, safe, and culturally appropriate. We'll use techniques like transfer learning to adapt a base model to our specific cultural and linguistic needs.
Challenges we ran into
- Cultural and Linguistic Nuance: One of the main challenges is ensuring the AI understands and responds to cultural contexts correctly. A phrase that is harmless in one culture could be deeply offensive or misconstrued in another. We need to be meticulous in our data collection and validation.
- Safety and Efficacy: Building a mental health tool requires extreme care. We must implement robust safety features to detect self-harm or crisis situations and ensure the AI never provides medical advice, but rather acts as a triage and resource navigator.
- Scalability: While a text-based bot is scalable, training the AI to recognize and respond to a wide variety of local languages and dialects is a significant hurdle. We have to prioritize based on the most common languages in a target region.
Accomplishments that we're proud of
We're most proud of developing a concept that directly tackles the stigma of mental illness. By framing the AI as a "listener" that draws on cultural heritage, we hope to make seeking help feel less like a clinical process and more like talking to a trusted, wise community member. This unique approach is what sets Sankofa Listener apart.
What we learned
We've learned that technology is only a small part of the solution. The most critical aspect of building a project like this is the human element—the deep understanding of cultural context, the collaboration with local experts, and a genuine commitment to creating a safe and trustworthy resource. We've learned that a successful AI in this space must be a bridge, not a replacement, for human connection.
What's next for Sankofa Listener
Our next steps would be to:
- Pilot the chatbot with a small, supervised group in a specific region to gather feedback and refine the AI's responses.
- Develop partnerships with more NGOs, community health organizations, and professional mental health associations to expand the network of verified resources the AI can recommend.
- Explore the addition of voice recognition to make the chatbot even more accessible for those with low literacy.
- Translate the core model and user interface to support additional local languages to expand the project's reach across Africa.
Log in or sign up for Devpost to join the conversation.