The development of AI models has often been siloed and associated with lack of representative data samples across different social classes and cultures. Conventional deep learning models often do not perform well when there is such undersampling of data. This can impact marginalized populations as the data curation is not representative or inclusive of all groups. In this panel, we discuss how curating new datasets and corpora representing the Hispanic population, can contribute to inclusive AI systems that mitigates biases towards underserved populations and move towards democratizing AI. The panel group will discuss this across different case studies in: long-term health tracking, disease classification in dementia and its subtypes, and population diversity in biological models, and cognitive behavioral changes from wearable sensors across diverse demographics.
Chair: Dhireesha Kudithipudi (UT San Antonio)