My research is focused on building and studying intelligent computing systems that demonstrate a human-like capability for in-depth understanding and production of natural language, and thus can achieve richer interactions with human users.
I am especially keen on building systems that decompose the meaning of language into complex conceptual structures that reflect humans' embodied cognition, memory, imagery, and knowledge about social situations. I use crowdsourcing and machine learning techniques to build these systems to scale, but I also often craft systems by hand because it can tell us how to use crowds and machine learning methods better.
Because these systems can comprehend narratives and participate in discourse, I use them to support safe and positive experiences for users of social media platforms, and to prevent the spread of harmful behaviors on them. I perform system evaluations and empirical studies with human participants using qualitative and quantitative methods.