The north star of AI development has long revolved around optimizing agents that are at ``human-levels’’ of accuracy on a given set of tasks — often learning by mimicking how humans perform. Such objectives contain undertones of AI being a means to the end of replacing humans across a wide breadth of tasks and are thus misaligned with the goal of creating trustworthy AI agents. This project focuses on creating agents that are optimized to it collaborate with humans and complement their problem solving abilities in order to ensure that future AI agents are better aligned towards benefiting society at large.
- Toward Automated Quest Generation in Text-Adventure GamesIn International Conference on Computational Creativity (ICCC), 2020
- Bringing stories alive: Generating interactive fiction worldsIn Proceedings of the Sixteenth AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE-20), 2020