Department of Computer Science
IT 910 Special Topics in Artificial Intelligence
Research Issues in Agent Development
Meeting time: Monday
Dr. Gheorghe Tecuci, Professor of Computer Science
Office hours: Monday
Office: ST-II, Rm. 421
Prerequisite: CS580, or other graduate level course in artificial intelligence, or permission of instructor.
This course addresses open research issues in agent development, such as mixed-initiative problem solving (that combines the humanís experience, flexibility and creativity with the agentís speed, recall, accuracy and consistency); knowledge representation at multiple levels of abstraction and formalization (to facilitate human-agent communication, domain modeling, problem-solving, knowledge acquisition and learning); agent personalization (through learning of userís preferences, biases, and assumptions); synergistic integration of modeling, learning and problem solving; synergistic integration of teaching and multistrategy learning (such that the user helps the agent to learn and the agent helps the user to teach it); mixed-initiative learning of userís language; non-disruptive learning (during agentís operational use); agent-supported human-information interaction (to cope with large volumes of information during mixed-initiative problem solving); ontology import, learning, and merging; agent problem solving and learning with an evolving representation language; mixed-initiative knowledge discovery; evaluation of agent-based systems. The actual topics that will be covered will be decided based on the research interests of the students.
Each student is expected to actively participate in discussing all the research issues. In addition, each student will select one research issue as the topic of the studentís project. The project will consist of a study of the current state of the art and of new advances proposed or made by the student. The study will be presented in class toward the middle of the semester,
and the research contributions will be presented toward the end of the semester. Some of the projects may include experimentation with or enhancement of the agent software developed in the Learning Agents Laboratory. This agent software will also be used to provide context for the above research issues. However, no prior knowledge of this software is required. It is desired that the projects will
result in research reports and publications.
The final grade will be computed as follows:
Class participation: 30%
Midterm research overview: 35%
Final research paper: 35%
Lecture notes provided by the instructor.
Various papers on agent development.