Let's Get Organized! Automating the Design of Agent-Aware


Event Details
Thursday, April 23, 2015
Talk:
4:00 p.m. - 5:00 p.m., 115 Avery Hall

Reception:
3:30 p.m. - 4:00 p.m., 348 Avery Hall

Ed Durfee, Ph.D

Professor, University of Michigan

Abstract

Human organizations have served as inspiration for techniques to coordinate computational agents that are cooperatively pursuing collective goals.  Translating organizational concepts into computational form is not easy, though.  Most past work has taken a problem-driven view, where the organization is designed around the decomposition of a problem into roles and interaction protocols.  In contrast, we are pursuing an agent-driven approach, where an organization is designed to help an existing team of agents work together more effectively and efficiently.  Our approach emphasizes the importance of designing an organization based on an awareness of the capabilities, expertise, preferences, and limitations of the agents that will populate it, including the degree to which those agents themselves can be expected to understand their organizational responsibilities and adapt their behaviors accordingly.
In this talk, I will describe our work in translating our agent-driven organizational design approach into a principled computational form that formally defines the organizational design space, articulates concrete metrics for evaluating competing designs in that space, and specifies unambiguous organizational influences that guide agents’ behavior while leaving them latitude for exploiting their individual expertise and organizational awareness.   I will summarize our empirical studies for understanding the promise, and open challenges, of automating the organizational design process, and lessons learned about how intuitions about human organizations can be computationally realized.  This is joint work with Jason Sleight.

Speaker Bio

Ed Durfee is a Professor of Computer Science and Engineering, and of Information, at the University of Michigan in Ann Arbor, where he has served on the faculty for over 25 years.  He develops and studies computational mechanisms for planning and coordinating the activities of sophisticated, autonomous computational agents that adapt their behaviors to achieve individual and collective objectives in complex, time-critical environments.  He received his AB degree in Chemistry and Physics at Harvard University, and his MS (Electrical and Computer Engineering) and PhD (Computer and Information Science) degrees at the University of Massachusetts.  He is a former President of the International Foundation for Autonomous Agents and Multi-Agent Systems (IFAAMAS), and has served on many editorial boards and program committees in his field.  His awards include a Presidential Young Investigator Award (1991), the IFAAMAS Influential Paper Award (2008 and 2010), and being named a Fellow of AAAI (2001) and IEEE (2004).