A Community Centered Approach to Bias Evaluation
Using funds from FAccT 2024 Scholars program, I implemented a 1-year research project with 4 MSc and 3 BSc women students from Addis Ababa Institute of Technology and Addis Ababa Science and Technology University. The project was focused on evaluating gender bias in Machine Translation Datasets for three Ethiopian languages. Importantly, the project focused on bias in translation among the three languages (as opposed to the status quo of measuring bias when translating from English) and curated evaluation datasets and frameworks from the ground up relying on the languages’ features and the community cultures. The PI, Dr. Hellina Hailu Nigatu, received the Wangari Maathai Impact Award at the 2026 Deep Learning Indaba conference for this work.