Delft Data Science Seminar about Online Education
You are invited to participate March 31st, 2017. During this day you are welcome to join not only our own Delft Data Science Seminar about Online Education.
Delft Data Science Seminar
Friday 31 March 2017 | 10:00 – 15:00 | Location: Sport & Cultuur
TU Delft’s Delft Data Science organises an event regarding Online Education. Massive Open Online Courses are a very innovative way to make education available for everyone. But how do we keep the quality high with such a large demand for Online Education? Delft Data Science’s researchers and international speakers will share their new findings in this seminar with other academics, partners and with you.
The Online Education Seminar is part of a two part program, including a Open Science Seminar (15:00 - 17:00h) for which you can register as well. Learn more about the Open Science Seminar
Justin Reich is the Executive Director of the MIT Teaching Lab since 2015. In addition, Justin is the Co-director of EdTechTeacher since 2006. Justin will speak about early Interventions to Support Online Learners on Friday during the Seminar.
The frame of mind that a learner brings to a new learning experience has a powerful impact on their performance and persistence in those courses. In two pilot studies, we experimentally tested a series of early interventions based on social psychology and behavioral economics in several Massive Open Online Courses. We used a value affirmation intervention to attempt to close gaps in persistence and performance between students from the UN's list of least and most developed countries. We used a planning intervention to help all students be more successful in meeting their goals. Based on the success of these two pilots, we are currently conducting a replication study in all new courses from MITx, HarvardX, and Stanford's OpenEdX. By conducting the experiments over dozens of courses with tens of thousands of students, we can leverage the log data from the edX platform and text data from participant responses to interventions to understand heterogenous treatment effects in ways that go beyond what can be accomplished in traditional lab experiments. The structure of MOOC data also lends itself to advances in open science, and using pre-registration and other transparency methods to increase the robustness of causal inferences. While new forms of data have a powerful role to play in education, the future of data science involves new data enhancing experimental paradigms, not replacing them.
Marcus is a professor for Advanced Learning Technologies at the Open University of the Netherlands and is currently involved in several national and international research projects on competence based lifelong learning, personalized information support and contextualized learning. Marcus' will present on the topic Data Enhanced Learning during the Seminar on Friday.
Reflection and personalized feedback are some of the most efficient means for enhancing human learning. Data tracking in digital systems and sensor-based wearable computing enable data-driven learning support for reflection and personalization. In the talk I will demonstrate three approaches for supporting reflection in a) an online learning environment b) a real-time feedback environments and c) a sensor-based Augmented Reality training environments.
Using log data to support reflection in action and about action has been recently used in several master courses of the Open Universiteit and the results show that this can enhance students responsiveness and collaboration patterns in online learning environments. Real-time sensor analysis and feedback based on this data has been used in the Presentation Trainer prototype. In several evaluations we have shown that real-time feedback can enhance human performance especially on complex skills. In a third approach we use wearable sensor systems to record expert performances and use the captured performances to train apprentices in AR. For this approach I will present a framework for scaffolding experts performances for specific training purposes in AR learning environments.
You can find his presentation here.
Jacob Whitehill is an assistant professor at the Computer Science Department of Worcester Polytechnic Institute. His research is in machine learning and computer vision, including deep learning, and its applications to affective computing, automatic facial expression recognition, human behavior analysis, and educational data mining. Jacob will give a presentation with the title What Does It Take to Optimize Human Learning? at the Seminar on Friday.
In order to improve large-scale educational platforms to be more effective at optimizing students’ learning, it is useful to find more accurate methods of estimating learners’ states (e.g., what do they know; how motivated/frustrated/engaged are they?) and predicting learners’ future behavior (e.g., will they persist or drop out?) based on their past behavior. Moreover, in order to maximize the benefit of personalized learning over “single-path” instruction, it is important to collect a large and diverse set of learning resources — which could consist of explanations, motivations, hints, practice problems, etc. — from which to draw. In this talk I will present three projects that tackle these perception, prediction, and resource collection challenges: (1) Automatic perception of student engagement from face videos; (2) Automatic prediction of MOOC student dropout from clickstream data; and (3) Crowdsourcing novel tutorial videos on math topics. The talk will conclude with a discussion, from a data science perspective, of some of the practical obstacles to creating optimized learning systems — e.g., selection bias, expense of randomized-controlled trials, and privacy.