Conversational Surveys

We are joined by Ziang Xiao, a Postdoc Researcher in the Fairness, Accountability, Transparency, and Ethics (FATE) group at Microsoft Research Montréal. Ziang is also a Research Assistant Professor in Computer Science at John Hopkins University. His research focuses on understanding how humans behave at scale. Ziang joins us to discuss his research on conversational surveys.

He began by establishing the place of surveys as a long-time tool for understanding human behavior. He then talked about how the efficacy of surveys can be improved with tools such as a chatbot. Ziang discussed how chatbots are used for conversational surveys. To expound on conversational surveys, Ziang discussed the three kinds of interviews: structured, semi-structured, and unstructured. He discussed the strategies needed to make a survey conservational.

Ziang explained the challenges with conversational surveys and current research to overcome them. One such challenge is to analyze the survey results at scale. He also talked about people’s reactions to being interviewed by a chatbot. Ziang shared metrics that were used to evaluate the success of conversational surveys. He contrasted the results from both conversational surveys and traditional surveys.

Ziang shared his thoughts about the impact of generative models such as ChatGPT on the progress of conversational surveys. He also discussed the fringe benefits of conversational surveys and how businesses can exploit them for richer feedback. He discussed the opportunities and prospects of conversational agents in the future. He also emphasized the need for continuous development and privacy protection when designing conversational surveys.

Ziang mentioned Juji as a tool that can be used for building conversational chatbots. He mentioned other areas of future research for him. You can keep up with Ziang on Twitter @ZiangXiao or his personal website.