CSIS Seminar

Limitations of Data-driven Frequency-based Approaches for Social Computing

Abstract

Frequency counts dominate the measurement of social media behavior analytics. Common computer science methods for the analysis of content such as visual analytics with word clouds, subjective text mining with sentiment analysis and topic classification all depend on frequency counts. However, such methods have two deficiencies. One of these is little to no theory regarding the mechanisms that generate the tallied observations. The second deficiency, related to the first, is little to no foundation for scaling the observed frequency counts. These limit the ability of social computing methods to address important practical problems, concerning for example the degree of distress in a compromised population or the extent of engagement in a political issue. This presentation situates frequency metrics of social media content in the context of a model of dyadic communication. The model focuses analysis on ratios of lexical choice rather than absolute frequency counts. This focus effectively supports practical applications such as comparative analytics of public response across critical world events, and reveals a role for so-called stop words.

Speaker Bio

Professor Valerie L. Shalin earned her Ph.D. in Learning, Developmental and Cognitive Psychology from the University of Pittsburgh. She currently serves in the department of Psychology at Wright State University, Dayton OH, where she is the Area Leader for the Human Factors focus in the Graduate HF/IO program. Dr. Shalin is an active participant in Wright State’s Knowledge-enabled Computing Center of Excellence (Kno.e.sis). In addition to her role as a co-editor of the seminal volume on Cognitive Task Analysis, she is co-recipient with her student of the 2016 Human Factors prize on the topic of Big Data, recipient of the NASA Ames team award for contributions to the first Mars Exploration Rover Mission and has over 70 referred journal articles and conference/symposium papers. Her current and past work has been funded by grants from the Air Force, ARI, INRIA, ONR, NASA, and NSF, encompassing task domains such as space exploration, disaster response and medicine. She has served on the editorial board for Human Factors and Journal of the Learning Sciences, refereed for twelve professional journals spanning psychology and engineering, and reviewed grant proposals for The US Army Medical Research and Material Command, Joint Congressional Projects, NSF and NASA. She has graduated thirteen Ph.D. students, all of whom focus on the analysis of human behavior to support the design of workplace technology and collaboration.