 We know from the research that individual tutoring can boost performance to letter grades. Every learner is different in their prior knowledge and experience, current interests and future goals. This means the learning experiences they need are also different. Unfortunately, our educational system treats learners the same. One size fits all. We know it's not a good fit, but we accept it because the education systems need to meet the demands of teaching at scale. So we have a tension between the demand to educate large numbers of learners and the desire to tailor instruction to individual needs. Can we do both? Yes, thanks to the fact that the learning process is replete with information. Information on what students are doing and thinking, how they're interacting. By capturing and leveraging this information, we can simulate the ideal of personalized learning instruction and do it at scale. But right now, even with the growth in technology, the amount of data that is captured is just a thin slice of what's possible. So we have a challenge, three challenges. First, find ways to capture the rich data on the learning process. Two, analyze the data to understand and detect relevant patterns. And three, translate those results to effective action. Imagine a learner studying a given topic on a computer-based system. We can collect millions of data points from every click of her mouse, tap of the keyboard, and even target of her gaze. Multiply that by all the learners studying that topic around the world, and we have orders of magnitude more data. Think of these learners as in a network. We can then use the power of the network, for example, identifying clusters of learners who share patterns, and then learn more about the group, as well as make better inferences about each individual. Analyzing the data. To analyze the data, we have methods that have been developed that can detect patterns in these large data sets and actually make surprisingly accurate predictions. Will Joey fail his course? Will Sue remember her training when she gets back on the job? While these predictions are useful, they're not sufficient because we also need to understand why. We also need explanations or diagnoses for why learners succeed and why they fail. The data can tell this as well if we combine detecting patterns in the data with models for learning. And thanks to decades of research on learning, we have accurate quantitative models of how people learn. We've developed a system called the Learning Dashboard based on this. As a learner is working and their knowledge of the material grows and develops, the Learning Dashboard collects the available data and analyzes it to develop an up-to-the-moment, accurate estimate of their learning state. What they know, how well they know it, what misconceptions they hold. And this is where the power of our model-based approach comes into play. In addition to just predicting the learner's performance, it explains it. It identifies skills that need more practice and misconceptions that need to be addressed. Thus enabling the system to adapt the material to the individual's needs, provide customized recommendations and targeted feedback. This attention to the individual is critical because learners take different paths through the material and every learner needs help finding their best path. In fact, what we know about novices is they often don't know what they don't know. So this system can help people be more effective independent learners. It can also help teachers. Think about the system telling a teacher what concepts the class is understanding well and which concepts they're struggling with. Which particular students need extra support and which students need more challenge. With this information, the teacher can target his or her time a precious resource more efficiently and effectively. At Carnegie Mellon, we've studied the system in practice and demonstrated its effectiveness. Students and introductory statistics using this system showed significantly greater learning gains in half the time compared to students in a traditional course. We replicated these effects three times and colleagues have shown similar results at seven different institutions. So I think of the learning dashboard as our first step in a vision for how technology can enable personalized learning at scale. Improving students learning and instructors teaching are just two of the feedback loops we can leverage. With learning data at the center, we can also help course builders improve the learning resources based on evidence and learning scientists deep in our understanding of the human mind, thereby transforming education to more effective and efficient system where data and evidence drive practice. And so my question, what I'd like to learn from you, is how do you see personalized learning helping you and your organization improve and adapt. Thank you.