AI Co-Creation in Malawi: Access to Legal Information
Tackling the "beneficiary" question has been at the core of Civaam's work. How do communities truly own tech projects, without extracting local knowledge? Our approach mixes co-creation and co-production. This means we understand local communities as the true knowledge experts, and all processes must flow through them. Not as "consultants," but as builders.
Civaam’s Sara Wahedi supported the Oxford Institute of Technology and Justice to centre co-creation at the heart of their ‘Access to Justice’ pillar by developing co-creation frameworks that facilitated the development of AI systems and stronger localisation of datasets, led by the Women Lawyers Association of Malawi (WLA) and built by Microsoft’s AI for Good Lab.
The Approach
-
Before any tech decisions get made, everyone in the room needs to be speaking the same language. Do we all understand what an AI system actually is? What it does, and what it doesn't do? Epistemic justice starts here: without a shared vocabulary between technologists and community members, consultation becomes theatre rather than genuine co-creation.
-
This stage uses local processes and discussion to prototype tech systems, not build them outright. It draws on a range of technologies and processes: RAG explained as a library, a local corpus of text, or a mapping of community experiences. The goal is to figure out where a tech intervention would actually live, whether that's a backend AI tool supplementing local work or something else entirely. Essentially, this stage maps out what the gap really is, and whether technology is the right solution to close it.
-
If tech is the right solution, this stage builds the capacity for local organisations to stay at the helm of the conversation. That means everything from prompt documents to datasets built through low-resource fine-tuning, adapting the tool to local slang, dialect, and context. This work takes significant time. It means bringing community stakeholders into development and staying connected with them, not treating them as consultants brought in for a single round of feedback. It requires long-term, iterative testing and development. Technology can't be truly owned by local communities if it's built elsewhere and handed to them. Countless examples show where that path leads: straight to the tech graveyard.