Towards an Actionable Understanding of Conversations
Conversations are central to our social systems. Understanding how conversationalists navigate through them could unlock great improvements in domains like public health, where the provision of social support is crucial. To this end, I develop computational frameworks that can capture and systematically examine aspects of conversations that are difficult, interesting and meaningful for conversationalists and the jobs they do. Importantly, these frameworks aim to yield actionable understandings—ones that reflect the choices that conversationalists make and their consequences, beyond the inert linguistic patterns that are produced in the interaction.
In this talk, I will describe two complementary efforts in this direction, addressing the internal dynamics of conversations and their broader impacts. First, I will present an unsupervised methodology to model conversationalists' interactional choices; I will then formally describe the problem of drawing causal links between conversational behaviours and outcomes. I will talk about these projects in the context of a crisis counseling service, where the conversational stakes and challenges faced by counselors are especially salient.
Host: Ben Zhao
Justine Zhang is a PhD Candidate in the Information Science department at Cornell University. She focuses on developing computational frameworks to study conversations. Her research engages with a wide range of fields, spanning natural language processing, computational social science, political science, psychological counseling, and economics. Previously, she completed a bachelor's degree in computer science at Stanford University. She is a recipient of the Microsoft PhD Fellowship.