Artificial intelligence (AI) and machine learning (ML) is becoming pervasive in our life. There is an increasing need for developing such a system. Yet the development processes of ML models or complex AI systems are often not well-supported. Researchers and developers have to go through tedious processes of examining details to understand the current performance and look for actionable insights. In this project, I study how visual analytics and interactive visualization can help exploratory analysis in ML/AI development process for improving the systems.

Social media websites provide rich environments for social scientists to study human behaviors and phenomena. However, given the enormity of many social media datasets, making data manageable for analysis by social scientists is still an unsolved problem. In this project, we focus on understanding the needs of social scientists, working to understand the contexts and research questions of their respective fields, and building visual analytics tools that facilitate their understanding of large text datasets.

Supercomputing has become a critical tool for many scientific fields such as climate science and physics. Nevertheless, there are few models that provide insight into how scientists think about their use of supercomputers and how it influences their workflow. In this project, we leverage ethnographic approaches to study the current practice of the use of high performance computing and to gain a deeper understanding of how to design usable data abstractions for next-generation scientific workflows.

The use of informal text communication (e.g. chat, texting, microblogs) is on the rise worldwide, including in collaborative work settings that span cultural boundaries. Socio-emotional communication is a critical determining factor in the cohesiveness of international work teams. However, text communication of any kind has often been viewed as impoverished and lacking in socio-emotional cues. To challenge this stereotype and provide evidence that technology is being repurposed, we study the patterns of emoticon use in informal text communication datasets and develop measures to detect social behavior patterns of emoticon use in the datasets.