PhD Proposal by Koustuv Saha
Title: Computational and Causal Approaches on Social Media and Multimodal Data: Examining Wellbeing in Situated Contexts
Ph.D. Student in Computer Science
School of Interactive Computing
Georgia Institute of Technology
Date: Wednesday, April 22, 2020
Time: 11:00 AM - 2:30 PM ET
Dr. Munmun De Choudhury (Advisor, School of Interactive Computing, Georgia Institute of Technology)
Dr. Gregory D. Abowd (School of Interactive Computing, Georgia Institute of Technology)
Dr. Thomas Plötz (School of Interactive Computing, Georgia Institute of Technology)
Dr. Emre Kıcıman (Information and Data Sciences, Microsoft Research AI)
Dr. Gloria Mark (Donald Bren School of Information and Computer Sciences, University of California at Irvine)
A core aspect of our lives involves interactions with the communities we are situated in. Understanding how we cope with psychological and cognitive demands is essential for both individual and collective wellbeing. Psychosocial dynamics of individuals are typically assessed using surveys, which though accurate in snapshots, suffer from recall bias, are reactive, and do not scale. These limitations are surmountable by social and ubiquitous technologies. My work leverages social media in concert with multimodal data, which facilitates analyzing dense and longitudinal behavior at scale. By adopting machine learning, natural language, and causal inference analysis, my work infers wellbeing and psychosocial dynamics of individuals and collectives, particularly those in situated communities, such as college campuses and workplaces. Broadly, my research bears design and technological implications for social computing systems and various stakeholders to support wellbeing and crisis intervention efforts in situated communities.
Before incorporating sensing modalities in practice, we need to account for confounds that may impact behavior change. One such confound is the phenomenon of “observer effect” or “Hawthorne effect”—that individuals may self-alter and deviate from their typical or otherwise normal behavior because of the awareness of being “monitored”. My proposed work studies this problem by leveraging the potential of longitudinal and historical behavioral data through social media. Focused on a multimodal sensing study of ~750 participants, I intend to conduct a causal study that broadly examines observer effect in social media behavior. I will adopt a theory-driven approach to model behavior change and measure how much and how long individuals are likely to modulate their social media behavior during study participation. Theoretically, this work provides insights into this phenomenon in terms of both social media behavior and longitudinal human behavior. Drawing on these insights, I expect to provide recommendations for correcting biases due to observer effect in social media sensing for human behavior and wellbeing.