Navigation and simulation technology solves challenges of online learning systems
Sunnyvale, CA and Cambridge, MA, June 17, 2013 — At the Sixth Conference of Massachusetts Institute of Technology’s Learning International Networks Consortium (LINC), Fujitsu Laboratories of America, Inc. and MIT have jointly announced a first-of-its-kind, revolutionary asynchronous, personalized learning platform – Guided Learning Pathways.
Guided Learning Pathways is a result of the companies’ joint research program focused on overcoming the challenges in traditional online learning systems, including finding appropriate learning materials and personalizing learning pathway of learners. Guided Learning Pathways is designed to address these critical problems for learners. As a result, two technologies have been developed and applied in the research. One is navigation technology, which can organize massive online learning materials into multi-layer topics. The other technology developed is the students’ learning behavior simulation based on an advanced probabilistic learner model. Fujitsu Laboratories of America and MIT will continuously research and develop Guided Learning Pathways, and apply it to the massive online learning systems for colleges and enterprises.
“We are pleased to collaborate with MIT to address a critical need in today’s online education environments” said Yasunori Kimura, President and CEO of Fujitsu Laboratories of America. “Our joint solution will enable students to rapidly acquire knowledge that is customized to their learning requirements”.
Full details of the research will be presented at the conference being held from June 16-19, 2013.
From high schools, through college, and continuing to early and mid-career, increasingly learners throughout the world are using the Internet to gain new knowledge and skills. The goals may be to study for a pre-college Advanced Placement (AP) test, to learning subject matter toward a bachelor’s degree, to obtaining knowledge and skills that are directly needed in one’s job, and much more. The trend towards online learning has grown recently due to the prevalence of Open Educational Resources (OER), where educational content is available free of charge. Additional growth is due to very recent creation of, Massive Open Online Courses (MOOCs), where over 100,000 students may take a course at one time.
These 21st Century Internet-enabled developments have not, however, been matched by comparable developments in methods of presenting educational materials to students in ways compatible with their varying learning styles and educational needs. Almost always the educational content is accepted to be a ‘course’ as in college course, where all the students learn a “course’s worth” of identical material and at the same speed. This old-fashioned industrial model acts on students as if they were identical mechanical widgets passing through a production process. But in fact no two students learn the same way nor do they naturally at the same speed.
GLP Concept Design
Online technology is allowing the ultimate personalization of the Internet experience—why not for education as well? E-commerce sites recommend things that you can buy that other shoppers—similar to you—also bought. Search engines tailor results that fit your personal interests. The same concepts and technologies can be brought to bear on education, to recommend to learners which “learning nuggets” to study.
Nuggets are learning materials that teach a single, atomistic concept with a domain like calculus (for example, derivatives). Examples of nuggets are a homework problem to be done online; a video snippet (e.g., lecture, real world application, Khan Academy video), shorter than ten minutes; a pop quiz for self-assessment of content knowledge; an animation, possibly interactive; a simulation, also possibly interactive; a web-based lab experiment; a short educational game; or a small section of textual material, typically less than one page in length.
By breaking domains into atomistic concepts and populating each concept with a wide variety of learning nuggets, GLP will be able to eliminate the Industrial Age “course” and tailor each individual’s education to suit their interests. As shown in Figure 1, each learner can follow his or her own pathway through the material. We can imagine that learners engage with nuggets that fit their non-academic interests, and intelligent tutors in GLP understand their strengths and weaknesses. Badging and rewards can also bring a fun element to learning and show learners how academic topics relate to the “real world”. For example, a learner studying calculus who loves music might go to an interactive simulation, which examines the singing behaviors of musicians. Among other things, it shows spectral patterns of the singing, the first and second derivatives of various elements of each song (e.g., decibels, real time change in frequency), and it allows the learner to alter recorded voices according to some calculus parameters to hear what the singer would have sounded like under those revised circumstances.
Figure 1: Learning Path and Nugget Recommendations
Many technological challenges exist for creating GLP; our research project has addressed some critical issues. First, while there are huge amount of OER available which are promising for GLP, it is difficult to organize them as learning nuggets into appropriate topics in learning pathways. Second, successful GLP applications require effective recommendations. However, field evaluations of educational technology are very resource intensive—evaluating recommendation algorithms would require a large number of students from diverse backgrounds in order to see significant results. Before taking a platform to students, though, we wanted to be able to narrow down the types of recommendation algorithms to field test.
Newly Developed Technology
For populating each topic in learning pathways with well-organized and reused learning nuggets from OERs, we provide an advanced approach to explore and organize learning nuggets with multi-layer topics having different granularity based on a probabilistic topic model (Latent Dirichlet Allocation) framework, which is more efficient and effective compared with conventional keyword search. Because for a new learner with limited knowledge to a new domain, it is very hard to obtain appropriate search keywords and find wanted contents. In order to evaluate different recommendation algorithms outside of the classroom, we created a simulation model based on a stochastic, Bayesian Knowledge Tracing algorithm. We also include a topic-graph generator that allows us to generalize across domains and test large-scale systems. Finally, we utilize an implicit rating system for learning materials, in which learning nuggets are not rated by learners directly, and instead their ratings are calculated based on learning outcome of learners. These components’ parameters were set using published field evaluations and give us a general sense of which recommendation algorithms we should test in a deployed system.
Figure 2: Learning material navigation with multi-layer topics
Atomistic topic maps are currently emerging in other MIT projects, such as some led by MIT’s OEIT (Office of Educational Innovation and Technology) and ODL (Office of Digital Learning), and the GLP team is speaking with these groups to explore future collaborations, and also introducing GLP to edX project. We expect that our ideas will blend with theirs to form stronger and more innovative learning projects that will be deployed within the MIT community. In general, the future seems bright for the many GLP concepts that have been demonstrated by independent organizations, such as personalized digital learning, massive enrollments in online classes, and collaborative, online learning. We look forward to unifying these concepts into a single platform and creating additional GLP components to serve learners around the world. As shown in Figure 2, we needed to define a software architecture to hold all the parts together, yet one that would also allow for GLP’s long-term growth and success. Based on GLP platform, Fujitsu is also planning to apply corresponding technology to its future business, which is not only involving traditional school education but also informal and lifelong corporate learning.
Figure 3: Software Architecture of GLP Platform
Nugget App & Service manage how learners view and interact with nuggets, or individual learning materials. Each nugget teaches a single concept topic and is “bite-sized”. This app requests the nugget (i.e. a video, set of lecture notes, animation, simulation, etc.) from the appropriate location on a server, and displays it with appropriate user controls. It collects analytics on how the user interacts with the nugget.
Nugget Recommendation App & Service calculate which nuggets would most likely be useful for each individual learner, and ranks them accordingly. The apps can use factors such as learning styles, non-academic interests, and results from other learners. The returned result is a list of nuggets, much like a search engine’s results page.
Intelligent Tutor App & Service present assessment-type problems to each learner. As learners solve the problems, the tutor builds up a model of each learner’s understanding. When appropriate, it provides personalized hints and explanations to improve each learner’s understanding.
Content Topic Map App & Service display the content topics relevant to each learner’s learning goals. Each content topic map consists of individual topics; for example, within the calculus map are topics like derivatives, integrals, and limits. The topics are thus a subset of all topics within GLP—for example, someone studying calculus would not see the topics related to basic algebra. Each topic has a set of learning nuggets associated with it that learners can study from.
Content Topic Recommendation App & Service determine which topics a learner is prepared to study, from the ones needed to reach her learning goals. A learner is prepared to study topics where she has mastered all of the pre-requisites at a sufficient level.
Education Dashboard App & Service allow educators to manage groups of learners. They can set a shared learning goal, but also adjust learning goals and activities for individual learners. Educators can also see group and individual progress.
User Registration App & Service allow users to register for GLP. Each type of user will follow a different process. For example, learners will experience a learning styles assessment and a knowledge level assessment, whereas content creators might experience a verification of their expertise.
Glossary and Notes
A massive open online course (MOOC) is an online course aiming at large-scale interactive participation and open access via the web. In addition to traditional course materials such as videos, readings, and problem sets, MOOCs provide interactive user forums that help build a community for the students, professors, and TAs. MOOCs are a recent development in distance education.
Open Educational Resources (OER) are freely accessible, usually openly licensed documents and media that are useful for teaching, learning, educational, assessment and research purposes.
edX is a massive open online course platform founded by Massachusetts Institute of Technology and Harvard University to offer online university-level courses in a wide range of disciplines to a worldwide audience at no charge, please see: https://www.edx.org/ ;
Fujitsu is the leading Japanese information and communication technology (ICT) company offering a full range of technology products, solutions and services. Over 170,000 Fujitsu people support customers in more than 100 countries. We use our experience and the power of ICT to shape the future of society with our customers. Fujitsu Limited (TSE:6702) reported consolidated revenues of 4.5 trillion yen (US$54 billion) for the fiscal year ended March 31, 2012.
For more information, please see: : www.fujitsu.com
About Fujitsu Laboratories of America, Inc.
Fujitsu Laboratories of America, Inc. is a wholly owned subsidiary of Fujitsu Laboratories Ltd. (Japan), focusing on research on Internet, interconnect technologies, software development and solutions for several industry verticals. Conducting research in an open environment, it contributes to the global research community and the IT industry. It is headquartered in Sunnyvale, CA.
For more information, please see: : www.fla.fujitsu.com
About the Professor
Professor Richard Larson is Mitsui Professor of Engineering Systems and Director of Center for Engineering Systems Fundamentals. The majority of his career has focused on operations research as applied to services industries including technology-enabled education. Since 2010, Professor Larson has started collaboration with Fujitsu Laboratories of America, Inc, on education research, which is a key component of Human-centric Intelligent Society project. Please see: http://linc.mit.edu/
About Fujitsu Laboratories Limited
Founded in 1968 as a wholly owned subsidiary of Fujitsu Limited, Fujitsu Laboratories Limited is one of the premier research centers in the world. With a global network of laboratories in Japan, China, the United States and Europe, the organization conducts a wide range of basic and applied research in the areas of Next-generation Services, Computer Servers, Networks, Electronic Devices and Advanced Materials.
For more information, please see: : jpfujitsu.com/labs/en