With great excitement I invite you all to participate in what I hope will turn into a continuing discussion on curriculum matters. This column, as the name suggests, will attempt to address issues relating to the teaching of AI, whether it is in an undergraduate or a graduate level course, or an advanced tutorial on a specific AI topic, or an AI training seminar at your organization. At times I will focus on the practice of AI in the industry and what is expected of a person that fills an entry-level AI job. Other times, I will examine the role of AI in the context of a college/university curriculum. I will try to bring forward emerging ideas, issues, techniques, texts, and tools that are being used in the teaching of AI.

Since the issuance of the call for papers for the 1994 AAAI Fall Symposium on Improving the Instruction of Introductory Artificial Intelligence and the subsequent symposium itself, it has been widely recognized that an introductory AI course is a difficult course to teach well. Even prior to that at least two groups of researchers were involved in developing general purpose tools for use in the learning of AI concepts.

PAIL (the Portable AI Lab), is an integrated collection of well established AI tools and techniques intended for use as a resource for teaching and learning AI. The second one, called FLAIR (Flexible Learning with an AI Repository), contains some visualizations, in addition to educational materials.

The FLAIR group has been hosting yearly, NSF sponsored, summer faculty enhancement workshops on the topic of teaching undergraduate AI (a four-year program that concluded this summer). An idea that emerged out of the summer workshops was a WWW accessible repository of pedagogical materials at Temple University. One of the significant outcomes of the Fall 1994 symposium was to create a more centralized and complete repository under the umbrella of AAAI. The repository exists at http://www.aaai.org/Resources (follow the link to the Education Repository). You will be able to access PAIL, FLAIR, as well as numerous other resources through this growing repository. There are also provisions for submitting your own materials (or links to it).

The 1994 Fall Symposium was followed up by a special track in the 1996 Florida AI Research Symposium (FLAIRS, no connection to the one above!). Despite the success of the track, this year's track at FLAIRS (the symposium) was abandoned due to low submissions. After having scanned the proceedings of 1997 SIGCSE conference I was further disappointed by the lack of any tracks on AI at that conference (there was one workshop on Genetic Algorithms). Does that imply that we have discovered the best way there is to teach AI? Or is it the implication that the number of AI and AI-related course offerings are dwindling at most educational institutions?

In my mind, what we teach in (or about) AI, like the definition of AI itself, is an evolving issue. As it has in the past, at least in my mind, AI continues to be at the frontiers of computer science and technology. Some may argue, rightfully, that at the moment in computer science the frontiers are led by advances in the hardware and networking arena. However, as you can see from the articles in this volume, AI is once again playing the lead role in advanced applications based on new hardware and network technologies.

While attending the first conference on Autonomous Agents I was taking note of the concepts and technologies being employed in the implementation of autonomous agents. I became woefully aware of the lack of instruction in the current computer science curricula that would lay the foundation for furthering the development of agent technologies. In my mind, the issue of what forms the syllabus for an AI course is becoming increasingly important.

As is evident, AI courses will continue to play a major role in the over all computer science curriculum. This volume is devoted to the topic of Autonomous Agents. I would like to urge you to look at the articles in this issue from the eyes of an AI educator, or, more generally, a computer science educator/trainer. Try to identify the underlying concepts being employed in the design and implementation of various agents and then try to answer the questions: How many of these concepts are currently taught in your curriculum? What would it take to incorporate the missing topics in your curriculum? What pedagogical materials would it take for you:

1. to understand those topics?
2. to help you teach them better?
3. to help your students understand them?

What course(s) will serve to introduce these in your curriculum? Perhaps the AI course is the best place for introducing these ideas.

In the next issue, I will attempt to summarize the embedded-agent oriented approach to teaching AI. I would welcome any experiences, and/or suggestions on this topic. You are also invited to present your ideas, concerns, or suggestions for other possible topics. You may do so in the form of a query, a short submission, or a longer manuscript.