A distributed task environment for teaching artificial intelligence with agents

Abstract

It is not uncommon to teach Artificial Intelligence (AI) by asking students to implement agents that embody intelligent behavior. This helps students gain a fuller understanding of the many concepts taught in the course. There are two issues with this approach that deserve attention. First, students come into an AI course knowing how to program in different languages and having different levels of programming ability. Second, it's useful for the students to have a single task environment for all of the agents they program. A solution to both issues lies in a distributed system where the agents are clients communicating with a server that handles a configurable task environment. This allows the students to program their agents in any language and on any platform they desire, so long as they can communicate with the task environment server. If the task environment can be configured to provide additional levels of complexity and difficulty, this allows students to program at a level they are comfortable with. They can then challenge themselves by incorporating more advanced capabilities into their agents. This paper presents just such a distributed and configurable task environment that was developed for an undergraduate AI course.

Developing Critical Insights into Artificial Intelligence

“…The notion of trying to teach AI by asking students to implement software that embodies intelligent behaviour and helps students gain a fuller understanding of the many concepts taught in the course underpins the work of Hill and Alford (Hill and Alford 2004). This approach uses Russell & Norvig's agentbased interpretation of the Wumpus World as a vehicle to stimulate interactive learning among students of mixed experience and ability.…”

Give students a clue

“…[12][p.18] In this paper we describe our project and associated support software. Others have developed frameworks for teaching AI, such as Hill and Alford's "task environment" [6] and Kumar and Meeden's "robot laboratory." [8] The project we present here requires no investment in hardware or software and does not require students to learn a new language or AI framework so that it can be completed within a semester.…”