Building a Tool that Draws from the Collective Wisdom of the Internet to Help Users Respond Effectively to Anxiety-Related Questions

Online anxiety support communities offer a valuable and accessible source of informational and emotional support for people around the world. However, effectively responding to posters' anxiety-related questions can be challenging for many users. We present ourwork in developing aweb-based tool that draws from previous question-response interactions and trusted online informational resources to help users rapidly produce high-quality responses to anxiety-related questions. We describe our efforts in four parts: 1) Creating a machine learning classifier to predict response quality, 2) developing and evaluating a computational question-answering system that learns from previous questions and responses on support forums, 3) developing and evaluating a system to suggest online resources for anxiety-related questions, and 4) interviewing support community moderators to inform further system design. We discuss how this tool might be integrated into online anxiety support communities and consider challenges with the tool's functionality and implementation. We also provide the dataset we used to train the system to provide opportunities for other researchers to build on this work.