Date:
Location:
Speaker: Chien-Ju Ho (Washington University in St. Louis)
Title: Towards Behavior-Informed Machine Learning
Abstract: Machine learning (ML) has seamlessly integrated into various facets of humans' everyday lives, largely drawing from human data for its training. Consequently, these ML systems frequently exhibit and reflect human behavioral biases, leading to concerns across a variety of applications. In this presentation, I will discuss my recent efforts to develop behavior-informed machine learning which considers and incorporates human behavior's impacts into ML system design. Specifically, my focus will be on two crucial aspects of human behavior in the ML lifecycle: the generation of data used for training machine learning models, and human decision-making processes that occur in conjunction with machine assistance. The goal of my work is to develop ML systems that are robust to behavioral training data and capable of augmenting and enhancing human decision-making capabilities.