This SBIR Phase 1 project proposes to demonstrate the feasibility of an advanced personalized learning system for pre-collegiate and remedial Algebra, with future extension to other scientific and mathematical disciplines. The project aims to model learning as a complex dynamics system, and to develop a user-model that incorporates student profile data with a detailed and large set of markers on usage and performance collected from an analytics engine. These markers can be derived from a student's learning and social activities, such as her interaction with various types of learning resources, participation on question and answer forums and so on. This model then guides a recommender system to select appropriate resources from a learning catalog based on the student's unique learning path. Whereas, the project uses open source recommender algorithms, the intellectual merit lies in developing a user model, identifying the specific markers that impact learning, automatically reconfiguring the learning material, and in demonstrating its viability for algebra learners. This model will be built leveraging some of the latest infrastructures in education technology including an authoring and delivery framework, a learning catalog of 10 million curated learning resources, and the latest advances in data-analytics and information filtering systems.
The broader/commercial impact lies in using the proposed technology to replace textbooks and other less-advanced learning systems, not only in mathematics, but also in other STEM disciplines, making it easier, faster and more affordable for students to learn. To an extent, it will level the playing field by providing equal access to a personalized 1-on-1 type of experience, even for those students who don't have the opportunity to get high quality instruction and mentorship today. It will significantly impact students who today struggle with mathematics and STEM subjects, by providing alternate and more relevant ways for them to visualize and learn. Additionally, Zyante's technology and expanded services will enable colleges/universities to offer high-quality online courses and successfully handle larger enrollments. All this will ultimately improve national outcomes and provide a better-trained workforce in disciplines needed to drive the economic success of the US. For these reasons, it is important that the proposed technology be readily commercialized.
1. Project Outcomes The SBIR Phase I project proposed development of learning materials and a user model to study the impact of "adaptive learning" targeted towards different learning styles. Through a series of studies, the project concluded that a blended style that includes includes a combination of activities yielded the most improvement, versus matching activities to a student's supposed learning style. The project went beyond the initial project goals and developed 6 full chapters of an Algebra zyBook, suitable to entirely serve the needs of an introductory Algebra course at UC Riverside. The algebra zyBook has been used by about 1,000 students so far, during summer 2014, fall 2014, and winter 2015. Moreover, the project developed another kind of adaptivity/personalization that auto-generates 5-7 problems of increasing difficulty. If a student gets an answer wrong, the correct answer is shown along with an explanation, and the student is given a new auto-generated question of the same difficulty. Feedback on these progressive activities has been extremely positive. Students self-reported improved learning and expressed a desire for additional exercises. The project also developed basic quizzing functionality. A student can generate a quiz that selects questions drawn from each section of a specified chapter. Student feedback indicates a strong appreciation for an ability to learn through self-quizzing. A pilot study that introduced a pre-quiz before a chapter yielded positive results. Interestingly, students spent more time on the chapter, and were more diligent in their attempts to answer questions ("guessing" less frequently, to the extent that the framework can detect random guessing, and showing themselves the answers less frequently too). Students who took the pre-quiz also performed better on a post-quiz. 2. Intellectual merit and broader impacts Many students entering college are not prepared for entry-level college courses. 20-30% of students in 4-year universities, and 60% of students in 2-year universities, enroll in developmental courses intended to help students prepare for requisite courses. However, many students never get past this barrier, with greater difficulties experienced in developmental math courses. Studies suggest only 25-30% of students enrolled in developmental math are able to successfully complete these courses. Targeting the re-design of developmental math materials has the potential for great impact. Algebra, statistics, and other materials based on the findings in Phase I may lead to more students succeeding in such core college courses, leading to higher college graduation rates, and thus more qualified graduates, especially in STEM. The findings can influence the development of not just math materials, but other analytical materials that involve both theory and practice, such as statistics, engineering, computer science, physics, chemistry, finance, accounting, and much more. 3. Additional information * Two papers are pending publication at the annual American Society of Engineering Educators (ASEE) 2015 conference. The project is continuing to analyze student improvement and anticipates additional conference and journal submissions. * The findings will steer content development of Zyante's zyBooks, currently used at over 200 universities by tens of thousands of students. * The Algebra materials (6 chapters) have been adopted by UC Riverside as the sole textbook/homework system, in use by 1,000 students annually. Subsequent development is expected after this project (possibly with Phase II support), followed by adoption by tens/hundreds of universities.
