Essay by Subrahmanian in Science Magazine Examines Machine Learning for Predicting Human Behavior

An essay by V.S. Subrahmanian published in Science magazine on Feb. 3 examines the use of machine learning to better predict human behavior.

Subrahmanian, a professor of computer science with an appointment in the University of Maryland Institute for Advanced Computer Studies (UMIACS), co-authored the piece with Srijan Kumar, a fourth-year doctoral student in computer science.

“Predicting human behavior: The next frontier,” states that recent advances in machine learning are revolutionizing how we understand human behavior, both online and offline. In the piece, the authors identify several challenges in using the technology in the near future. This includes “noise” in big data that can sometimes overwhelm predictive models, and rare-event prediction, such as companies monitoring their internal networks to identify the “rare” users who may steal important information.

“In the coming years, machine learning is poised to help humanity make transformative leaps in many disciplines, ranging from national security to human health,” says Subrahmanian. “However, many technical and ethical challenges remain in order to enable ordinary humans to better understand the strengths and weaknesses of such predictive models.”

Subrahmanian has used machine learning for predictive modeling on topics like identifying malicious users and activities on social media, forecasting leadership changes in terrorist organizations, or determining the location of IED weapons caches in Iraq and Afghanistan.

Much of the work takes place in the Laboratory for Computational Cultural Dynamics (LCCD), one of 16 centers and labs in UMIACS.