Weaving algorithms

Abstract This paper details the process of indigenizing an Arizona water chatbot. Initial attempts to integrate tribal data, adopt an Indigenous worldview, and apply the Two-Eyed Seeing framework were unsuccessful, leading to inappropriate outputs and potential harm. The turning point came when we reframed bias as evidence and our embodied reactions as data, enabling an epistemological reorientation toward relational sense-making. We argue that, to serve Indigenous communities effectively, information design must address these fundamental epistemological conflicts. Our work offers a model for creating more attuned, decolonial AI chatbot interfaces that have wide-reaching potential.