CAREER: Socially-Aware Language Technologies To Support People in Supporting Others for Better Online Communities

Stanford University, Stanford CA

Investigators

Abstract

This award is funded in whole or in part under the American Rescue Plan Act of 2021 (Public Law 117-2). This research will invent a set of new socially aware language technologies designed for online peer support groups, including machine learning classifiers that automatically predict helping skills from text in noisy and low-resourced settings, and language generation techniques that effectively generate tailored and contextualized assistance for supporters. The people who provide support are the key to the success of these online groups, which are used by millions of people with health concerns. However, online supporters often do not receive rigorous training and tailored feedback, which might lead to unsupportive or even negative helping behaviors. Existing mechanisms of training or scaffolding largely rely on human supervision, making it hard to scale up to help the large number of supporters who support millions of people in need of care. This work has the potential to advance the state-of-the-art and scale up to many different other domains with minimal human effort. By developing, deploying, and evaluating new interventions that empower supporters in socially important domains, this work broadens the scientific understanding of technology use for mental health peer support. By combining computer science with the study of online peer support groups, this work will appeal to students who might not otherwise be attracted to science and engineering careers, including women and members of underrepresented groups. This work will accomplish the vision of supporting people in better supporting others in several representative text-based online peer support groups by: (1) developing innovative natural language processing techniques to predict supporters' helping skills and examining how helping skills related to positive outcomes; (2) designing contextualized language generation approaches that provide tailored assistance for supporters by highlighting which helping skills are needed in a given situation and suggesting example responses with actionable feedback; and (3) creating an open-source and human-in-the-loop tool to empower supporters and evaluating how the tool can be used for both training and real-time scaffolding via lab studies, field experiments, and real-world deployment. The result will be a novel synthesis of social science theories and beneficial advances in natural language processing (NLP) communities, to pioneer this emerging research field that uses NLP to support mental health and well-being. Concretely, it will develop scientific knowledge of how supporters use different helping skills to help seekers, and a deep understanding of how such support exchange relates to positive outcomes. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.