Human-AI Tools for Expressing Human Situations and Contexts to Machines
PI: Haoqi Zhang
Associate Professor of Computer ScienceNorthwestern University
Co-PI: Darren Gergle
Professor of Communication Studies and (by courtesy) Computer Science
Northwestern University
Framework components: Algorithm and Interaction
Despite significant advances in machine sensing and machine learning technologies—and the availability of rich application programming interfaces (API) for creating context-aware applications—it remains difficult for designers to express their concept of a human situation 
(e.g., places to hold a private conversation; good places to take a bike ride with young children) to machines so that applications can be aware and responsive to the situation across a variety of distributed contexts. Failing to do so could result in costly errors in safety and privacy or lead to inequitable access to AI supported experiences.
To bridge this gap, this project advances new programming environments and tools that support designers using their understanding of human situations to construct machine representations using available context features. Focusing on addressing potential failures in AI safety, this project seeks to (1) help designers form rich and accurate conceptions of human situations to encode into machines systems; (2) provide support for refining concept expressions and evaluating machine learning models in actual use cases; and (3) identify and address issues of differential access across settings.