​A half-day introduction to responsible AI and bias in machine translation models for African languages, in particular. Currently, there is significant interest in building Natural Language Processing (NLP) tools for African languages to enable various applications, especially those requiring local context. This is, however, hindered by the low availability of text datasets to support this effort, as opposed to more resourced languages like English, Swedish, etc. 

In this workshop, we introduce concepts of Responsible AI when building models for machine translation models for African languages. We will use a case study, discuss the approach we have taken to evaluate and quantify the gender bias within a Luganda-English machine translation system using Word Embeddings Fairness Evaluation Framework (WEFE). ​​