Minodu, a project of DFKI, Kara University, and local communities in Togo, investigates such under-represented knowledge to co-develop adaptive agricultural practices in a rapidly changing climate. While we cannot solve structural homogenization through AI our work offers practical strategies for reinforcing linguistic and epistemic plurality.

Our session combines a brief project introduction, a discussion with our Togolese team members, and an open, participative fishbowl discussion to collectively explore how epistemic diversity can be safeguarded in the age of AI.

We discuss:

1. The creation of a Kabiyè glossary for climate and agricultural terminology, codeveloped with a Togolese linguist;

2. An audio-first communication strategy aligned with oral knowledge traditions and local media ecologies;

3. Low-tech co-creation methods such as community media access points and participatory prototyping.

These practices illustrate how community-rooted, language-sensitive design processes can counteract epistemic loss and challenge the cultural biases embedded in AI.