NEWS In August 2009, I will be an Assistant Professor at Worcester Polytechnic Institute, in Social Science and Policy Studies, Learning Science and Technology, and Computer Science.
I am:
My research is at the intersection of Educational Data Mining and Human-Computer Interaction. I develop and use methods for mining the data that comes out of the interactions between students and educational software, in order to better understand how students respond to educational software, and how these responses impact their learning. I study these issues within intelligent tutors and educational games.
In recent years, my colleagues and I have developed automated detectors that make inferences in real-time about students' motivational and meta-cognitive behavior, using data from students' actions within educational software (no sensor, video, or audio data).
In particular, we have developed software that can automatically detect "gaming the system", attempting to succeed in an interactive learning environment by exploiting properties of the system rather than by learning the material.
We have also developed software that can detect, when a student is idle, if the student is off-task, or if the student discussing the subject matter with a teacher or another student. In addition, we have developed models that can determine whether an incorrect answer genuinely reflects a lack of knowledge, or is simply a slip.
We develop these detectors using methods drawn from data mining, machine learning, and psychometrics [Read More]. These detectors are generally developed using labels generated either through quantitative field observations and text-based replays of log data [Read More]. We have demonstrated that our detector of gaming the system can accurately assess this behavior when students are using different tutor lessons than the detectors were originally trained on.
These models of student behavior have proven themselves useful tools for making discoveries about human learning and learners. For instance, my colleagues and I have discovered that gaming the system is generally associated with poorer learning, but not always - specifically, gaming does not lead to poorer learning if students game material they already know. We have also found that a student's choice to game is more influenced by differences in their learning software, and responses to momentary affect (specifically boredom and confusion), than to stable or semi-stable student attributes, such as attitudes towards mathematics or performance orientation.
This work in turn leads to the development of more effective learning software, software that can adapt effectively and sensitively to differences between students. In my doctoral work, I developed an interactive software agent, Scooter the Tutor, who responds when students game, significantly improves gaming students' learning. [Read More]. In another project, my colleagues and I improved the accuracy of knowledge estimates used in mastery learning, via incorporating contextual models of guessing and slipping; in 2008-2009, we will study whether this leads to better and faster learning of complex skills and concepts.
Another ongoing project involves students' learning and engagement in educational games. In one recent study, my colleagues and I found that students can experience equally positive moment-to-moment engagement in an intelligent tutor and an award-winning educational game, suggesting that games' impact on engagement may not be because they improve moment-to-moment affect relative to tutors. We have also developed a model of affective dynamics in games. In 2008-2009, we will be studying the effects of incorporating game features into intelligent tutors.
Please check out my publications web page for recent papers.
Other projects I was involved in, in the past, included
- Developing cognitive tutors to teach data representation and interpretation
- Visualizing students' implementations of data structures, at a conceptual level [Read More].
