| Zachary Pardos
加州伯克利大学信息学院和教育研究院副教授
|
|
【主讲】Zachary Pardos，加州伯克利大学信息学院和教育研究院副教授
【时间】2019年5月31日（周五）下午14：30 – 16:00
Dr. Pardos is an Assistant Professor at UC Berkeley's School of Information and Graduate School of Education. He is an expert in knowledge representation and personalized learning technology leveraging big data in education. His current projects focus on increasing upward economic mobility in the California public postsecondary system and using behavioral and semantic data to map out paths to cognitive and career achievement. He earned his PhD in Computer Science at Worcester Polytechnic Institute. Funded by a U.S. National Science Foundation Fellowship (GK-12), he spent extensive time with K-12 educators and students working to integrate educational technology into the curriculum as a formative assessment tool. He has published over 50 peer-reviewed articles related to learning analytics and holds several academic leadership positions in the community, including posts as an editorial board member for the IJAIED and JEDM journals, executive committee member for the Artificial Intelligence in Education Society, and program committee member of the 2019 ACM conferences; Learning @ Scale, Learning Analytics and Knowledge, and RecSys. Dr. Pardos comes to UC Berkeley after a post-doc at MIT Computer Science Artificial Intelligence Lab (CSAIL). At UC Berkeley he directs the Computational Approaches to Human Learning (CAHL) research lab and teaches courses on data mining and analytics, digital learning environments, and machine learning in education.
【Speaker】Zachary Pardos, Assistant Professor, UC Berkeley's School of Information and Graduate School of Education
【Topic】Reasoning About Knowledge from Learner Pathways
【Time】Friday, May 31, 2019, 14：30 – 16:00
【Venue】Room 404, Weilun Building, Tsinghua SEM
【Organizer】Department of Management Science and Engineering
【Abstract】 The aggregate behaviors of students in a learning environment can collectively encode information about the pedagogical objects with which they interact. In this presentation, I will demonstrate ways in which the synthesis of data from higher-ed can illuminate the terrain of higher ed and support students in their decision making and way-finding. An application of recurrent neural networks and skip-grams, techniques popularized by their application to modeling language, are brought to bear on millions of historic student course enrollments to create vector representations of these objects. Analysis of the produced vector space reveals predictive information about students' degree stop-out and course preparation. These course selections also convey a substantial level of semantic relational information about courses, rivaling the textual information conveyed from course descriptions. This information can be visualized, reasoned about, and ultimately surfaced to students. Our course information platform (askoski.berkeley.edu), adopted by the UC Berkeley Office of the Registrar, uses this automatically inferred semantic information to help students navigate the University's offerings and provides personalized course suggestions based on topic preference, course history, and program requirements. Considerations for balancing the tension between exploration and on-time graduation at a liberal arts university will be discussed, as well as the applicability of these techniques to other educational technologies.
|
