My research combines Interactive Learning Environments (ILE), Educational Data Mining (EDM) and Information Visualization (Info-Vis) to inform analysts, educators and researchers about user behavior in software, specifically in CBEs, which include intelligent tutoring systems, computer aided instruction tools, and educational games.InVis is a novel visualization technique and tool I created for exploring, navigating, and understanding user interaction data. InVis reads in user-interaction data logged from students using educational systems and constructs an Interaction Network from those logs. Using this data InVis provides an interactive environment to allow instructors and education researchers to navigate and explore to build new insights and discoveries about student learning.I conducted a three-point user study, which included a quantitative task analysis, qualitative feedback, and a validated usability survey. Through this study, I show that creating an Interaction Network and visualizing it with InVis is an effective means of providing information to users about student behavior. In addition to this, I also provide four use-cases describing how InVis has been used to confirm hypotheses and debug software tutors.A major challenge in visualizing and exploring the Interaction Network is network's complexity, there are too many nodes and edges presented to understand the data efficiently. In a typical Interaction Network for twenty students, it is common to have hundreds of nodes, which to make sense of, has proven to be too many. I present a network reduction method, based on edge frequencies, which lowers the number of edges and nodes by roughly 90\% while maintaining the most important elements of the Interaction Network. Next, I compare the results of this method with three alternative approaches and show our reduction method produces the preferred results. I also present an ordering detection method for identifying solution path redundancy because of student action orders. This method reduces the number of nodes and edges further and advances the resulting network towards the structure of a simple graph.Understanding the successful student solutions is only a portion of the behaviors we are interested in as researchers and educators using computer based educational systems, student difficulties are also important. To address areas of student difficulty, I present three different methods and two visual representations to draw the attention of the user to nodes where students had difficulty. Those methods include presenting the nodes with the highest number of successful students, the nodes with the highest number of failing students, and the expected difficulty of each state. Combined with a visual representation, these methods can draw the focus of users to potentially important nodes, which contain areas of difficulty for students. Lastly, I present the latest version of the InVis tool, which is a platform for investigating student behavior in computer based educational systems. Through the continued use of this tool, new researchers can investigate many new hypotheses, research questions and student behaviors, with the potential to facilitate a wide range of new discoveries.