Andre Ye

Project

Visual Discursive Datasets

Machine learning traditionally works with “verdictive datasets”, in which dataset samples are “verdicts” (determinate labels) given to contexts. However, philosophical, psychological, and sociological work demonstrates that many of the concepts we want to represent in machine learning — morality, aesthetics, norms — are *discursive*, meaning that they are formed and shaped through discourse (conversational landscapes from personal to societal levels). This project focuses on developing visual discursive datasets and exploring what computer vision models and human-AI interactions can be built from it.

What I learned

I have been able to further pursue independent research and use the wealth of resources around me to develop more innovative and unexplored ideas.

Advice for future applicants

Know and tell a compelling story about why your problem or idea matters!

More scholar profiles