Export
Keith Decker
· 4 previous affiliations
- University of DelawareSince 1996
- Associate Professor · Since 2002
- Assistant Professor · 1996–2002
- Carnegie Mellon University1995–1996
- Postdoc · 1995–1996
- University of Massachusetts Amherst1987–1995
- PhD, Computer Science · 1987–1995
- Rensselaer Polytechnic Institute1984–1987
- Master's Degree, Computer Science · 1984–1987
Research Overview
Keith Decker focuses on distributed artificial intelligence, multi-agent systems, and the architecture of information agents. His observed work from the mid-to-late 1990s explores how autonomous software entities can cooperate asynchronously to retrieve information and support complex decision-making. The most recent papers in the evidence address the role of intermediary agents in decentralized networks and electronic commerce.
Research Trajectory
The available evidence for Keith Decker's research trajectory is concentrated in the mid-to-late 1990s, focusing heavily on the design, modeling, and deployment of multi-agent systems. During this period, his work addressed the foundational challenges of how distributed software agents can interact, share information, and coordinate tasks in decentralized environments such as the early Internet. Decker's foundational interests in distributed computing are reflected in early themes such as the research direction of Critical Path Prioritization for Parallel Execution and the area of Dynamic Redundancy Cancellation in Parallel Systems. These concepts of parallel efficiency and fault tolerance naturally extended into his subsequent work on autonomous software agents.
In 1996, Decker served as the senior author on Distributed Intelligent Agents, which detailed the Retsina framework. This work demonstrated how a distributed collection of software agents could cooperate asynchronously to perform goal-directed information retrieval and integration. The system was designed to support complex decision-making tasks, with applications ranging from everyday organizational management to financial portfolio oversight. This research aligns with broader challenges in distributed computing, touching upon themes such as the research direction of Scheduling Complex Workflows with Directed Acyclic Graph Dependencies by addressing how multiple autonomous entities manage dependent tasks.
In the same year, Decker was the first author of Modeling Information Agents: Advertisement, Organizational Roles, and Dynamic Behavior. This paper expanded on the theoretical underpinnings of agent interaction, formalizing how information agents advertise their capabilities, assume specific organizational roles within a multi-agent system, and adapt their behavior dynamically in response to changing environmental conditions. This focus on dynamic adaptation in uncertain environments shares conceptual ground with the broader area of Robust Control for Partially Observable Robotic Systems and work in Modeling Time Quality Tradeoffs in Anytime Computation, where systems must make optimal decisions under strict resource and information constraints.
Decker's focus on agent coordination culminated in the 1997 paper Middle-Agents for the Internet, where he again served as first author. This work introduced a taxonomy of "middle-agents"—intermediaries such as matchmakers, yellow page agents, blackboard agents, and brokers—that facilitate the flow of information in electronic commerce and open multi-agent systems. By analyzing the performance characteristics of these different middle-agents, including their privacy, robustness, and adaptiveness, Decker provided a framework for connecting ultimate information providers with requesters in large-scale, distributed networks.
Current Directions
Based on the provided corpus, the most recent paper in the evidence is from 1997. This late-1990s work primarily investigates the architecture of intermediary agents and decentralized information flow. However, the broader taxonomy of Decker's associated research themes includes more recent developments, such as the research direction of Policy Optimization in Learned Latent Imagination. While the direct paper evidence is sparse, these thematic associations suggest an overarching interest in how autonomous agents learn, plan, and optimize policies in complex, data-intensive, or partially observable environments.
Profile updated
| Title | Venue | Year |
|---|---|---|
| Middle-Agents for the Internet | IJCAI | 1997 |
| Distributed Intelligent Agents | IEEE Expert | 1996 |
| Modeling Information Agents: Advertisement, Organizational Roles, and Dynamic Behavior | Agent Modeling | 1996 |
Loading new papers…
Loading research neighbors…
See something wrong? Report a mistake