What Do Knowledge Graphs Really Know?

Mark Musen (Stanford)

Abstract

Since the earliest days of AI, there has been a struggle to define what we really mean by knowledge and how to determine when a computational agent displays intelligence. Allen Newell had considerable influence in asking the AI community to think about knowledge in behavioral terms. He argued that knowledge representations do not manifest intelligence until some process comes along and operates on them. In that perspective, knowledge graphs are not intrinsically knowledgeable at all. We need agents to do something with our graphs to cause intelligent actions. With all the excitement about knowledge graphs, we’ve been focussing intently on the graphs. In the next round of research, we need to be thinking more about the knowledge. Knowledge graphs give us the ability to represent unimaginable numbers of entities in the world and the relationships among them. It’s time to apply the same kind of principled thinking that has led us to the development of knowledge graphs to the construction of intelligent agents that can demonstrate, through their behaviors, what our graphs really know.

The slides are available here.

Bio

Mark Musen is Professor of Biomedical Informatics and of Biomedical Data Science at Stanford University, where he is Director of the Stanford Center for Biomedical Informatics Research. He conducts research related to intelligent systems, computational ontologies, open science, and biomedical decision support. His group developed Protégé, the world’s most widely used technology for building and managing terminologies and ontologies. He served as principal investigator of the National Center for Biomedical Ontology, one of the original National Centers for Biomedical Computing created by the U.S. National Institutes of Heath. He directs the Center for Expanded Data Annotation and Retrieval (CEDAR), founded under the NIH Big Data to Knowledge Initiative. CEDAR develops semantic technology to ease the authoring and management of biomedical experimental metadata. He was the recipient of the Donald A. B. Lindberg Award for Innovation in Informatics from the American Medical Informatics Association in 2006. He has been elected to the American College of Medical Informatics, the International Academy of Health Sciences Informatics, and the National Academy of Medicine.