With the rising use of Artificial Intelligence within different functions in society, the issues of bias within these systems are becoming more widespread and notably impacting a larger number of people with a higher degree of seriousness. To mitigate and address these issues, this paper analyzes current practices in dataset compilation, use and AI system design before highlighting some state-of-the-art work being done in this domain. This is followed by recommendations to improve and build upon that work and propose and inclusivity matrix along with an evaluation metrics, vernacular sharing and a call for small-data-based AI approaches as concrete steps in addressing the issue of bias in these systems.
