By using intelligent agents to simulate instruction, agent-based learning environments can serve as a powerful research tool to investigate teaching and learning. The agent metaphor provides a way operationalize and simulate the “ human” aspect of instruction in a more ecologically valid way than other controlled computer-based methods. Additionally, from an architectural perspective, since agents are independent objects in the learning environment, it allows for more flexibility in research design. In particular, agent-based learning environments with multiple agents, in systems such as MIMIC (Multiple Intelligent Mentors Instructing Collaboratively), allow for investigating the effect of multiple mentors or multiple perspectives on a learning topic. Preliminary results from MIMIC research indicate that multiple agents can serve to effectively operationalize instructional theory. In terms of overall impact, creating agent-based learning environments to investigate instructional issues is at the leading edge of revitalized research integrating artificial intelligence with education, and in exploring new paradigms for researching teaching and learning.