The increasing integration of sensors and devices across domains as disparate as health, manufacturing, agriculture, and education is leading to unprecedented data availability and modeling for understanding and decision-making. This is the forefront of a new paradigm of satisfying information needs not only through access to documents but through automated scientific analysis over real-world evidence to directly answer questions, support decision-making and satisfy other information needs. In this talk, I will motivate this new paradigm of automated scientific analysis for satisfying information needs, discuss key technical challenges and provide point examples of causal methods (and their current limitations) that begin to address these challenges. Throughout the talk, I will draw on recent studies applying causal methods to extract knowledge and aid decision-making from both natural language (e.g., social media) and tabular (e.g., user behavior) data.
Emre Kiciman is a Senior Principal Researcher at Microsoft Research. His research interests span causal inference, machine learning, and AI’s implications for people and society. Emre’s main focus is on improving causal methods for decision-making across critical application domains; and, in the context of AI’s implications for society, his projects include work at the intersection of security and machine learning. Emre has a strong interest in computational social science questions and social media analyses, especially those that require causal understanding of phenomenon in health, mental health; issues of data bias; and understanding how new technologies affect our awareness of the world and enable new kinds of information discovery and retrieval. https://kiciman.