Machine Learning as Grounded Theory: Human-Centered Interfaces for Social Network Research through Artificial Intelligence

Lorenzo Barberis Canonico,Nathan J. McNeese,Chris Duncan
DOI: https://doi.org/10.1177/1541931218621287
2018-09-01
Proceedings of the Human Factors and Ergonomics Society Annual Meeting
Abstract:Internet technologies have created unprecedented opportunities for people to come together and through their collective effort generate large amounts of data about human behavior. With the increased popularity of grounded theory, many researchers have sought to use ever-increasingly large datasets to analyze and draw patterns about social dynamics. However, the data is simply too big to enable a single human to derive effective models for many complex social phenomena. Computational methods offer a unique opportunity to analyze a wide spectrum of sociological events by leveraging the power of artificial intelligence. Within the human factors community, machine learning has emerged as the dominant AI-approach to deal with big data. However, along with its many benefits, machine learning has introduced a unique challenge: interpretability. The models of macro-social behavior generated by AI are so complex that rarely can they translated into human understanding. We propose a new method to conduct grounded theory research by leveraging the power of machine learning to analyze complex social phenomena through social network analysis while retaining interpretability as a core feature.
What problem does this paper attempt to address?