Squirrel AI Learning Attends the 2019 Stanford University Mediax Conference: Connecting the Learner and the Learning with Algorithms and Analytics
The 2019 Stanford University MediaX Conference was held at the School of Education, Stanford University on October 23, North American time. The theme of the conference is: Algorithms and Analytics: Connecting the Learner and the Learning.
The speakers that were invited to the conference include: Richard Tong, Chief Architect of Squirrel AI Learning, Daniel Schwartz, Dean at the School of Education, Stanford University, Mark Musen, Professor of Biomedical Informatics at Stanford University, Ajay Madhok, Founder of Reboot Digital, and other experts and scholars in the fields of artificial intelligence and education research.
Richard Tong, Chief Architect of Squirrel AI Learning, was invited to deliver a keynote speech. He shared with hundreds of researchers, technology practitioners and other audiences from Stanford University about the experience of Squirrel AI Learning in using AI to bring great changes to students’ learning and future education, and introduced in detail to the participants how the research collaboration projects between Squirrel AI Learning and the world’s top universities and research institutions overturned the traditional teacher-centered education model, so as to provide students with personalized and high-quality education.
The current overseas collaboration projects of Squirrel AI Learning include: joint technology development with SRI international, virtual personal assistant to help students find the root causes of mistakes, multimodal comprehensive behavior analysis, etc., as well as SimStudents simulation students collaboration with Carnegie Mellon University (CMU), etc.
MediaX is a member project of the Human-Sciences and Technologies Advanced Research Institute, which is affiliated to the School of Education, Stanford University. It brings together more than 20 interdisciplinary laboratories at Stanford University, high-quality and rich academic resources, professors, innovative companies and researchers to explore the application of information technology in the future industry. MediaX members include Facebook, Fujitsu, Omron, Hong Kong University of Science and Technology, Squirrel AI Learning by Yixue Group and other internationally renowned technology companies and universities.
Human Learning Code
Martha Russell, Executive Director of MediaX at Stanford University and a Senior Researcher at the Human-Sciences and Technologies Advanced Research Institute, Stanford University, first delivered an opening speech.
Russell currently leads the business alliance and interdisciplinary research of MediaX at Stanford University, covering a variety of business development and innovation in the fields of information science, agriculture, communications and microelectronics. Focusing on the power of a common vision, Russell developed planning/assessment systems, and provides professional consulting services on technological innovation for regional development.
In her speech, Russell expressed three views around the “Human Learning Code”:
First of all, human experience is based on the sharing of information among people and groups, and rational learning is the basis for creating a successful user experience.
Secondly, learning analytics is not only about “Analyzing Learning Data”, but also about having an in-depth understanding of which learning activities are effective, for whom and when.
Finally, the field of learning analytics has the potential to improve students’ success through a deeper understanding of the academic, social emotion, motivation, identity and metacognitive background behind each student. These insights can also be applied to business, entertainment and health.
Using Intelligent Adaptive Education Algorithms
Richard Tong is the Chief Architect of Squirrel AI Learning. He has served as the Head of Implementation, Greater China Region for Knewton and the Director of Solution Architecture at Amplify Education. In addition, he is also a Member of the IEEE AIS (Adaptive Teaching System) Standard Working Group and Chairman of the Interoperability Group (IEEE 2247.2).
Richard said that at present, the development speed of AI is beyond people’s expectations and imaginations, and any applications and scenarios related to AI are experiencing blowout growth.
Soon, the role of AI will be increasingly reflected in education and other fields that related to human interest. Especially in the future of the education field, every student will have an adaptive AI tutor, even those children with special needs are no exception.
Richard believes that AI and human teachers can learn from each other’s strengths and close their gaps, and combine each other’s advantages to provide students with all-round quality teaching. For example, AI has more advantages in diagnosing students’ learning state and providing intelligent adaptive education, while human teachers are better at psychological support and encouragement.
In order to achieve these goals, AI needs different algorithm modules, such as multi-dimensional probability knowledge state prediction algorithms can help AI diagnose learning state, path optimization algorithms based on knowledge map can help AI better recommend tasks for students; active learning and human-in-the-loop methods can enhance cooperation between AI and human beings.
Since 2014, Squirrel AI Learning has been independently developing an intelligent adaptive learning system for Chinese K12 students. Its main goal is to accurately diagnose the mastery of students’ knowledge points, and then recommend personalized learning content and learning path planning.
As a leading intelligent adaptive education company in China, Squirrel AI Learning is also making efforts in research and development. At present, Squirrel AI Learning is using deep learning to enhance the Bayesian-based tracking algorithms of students’ knowledge points and KST algorithms, etc.; using SimStudent and Apprentice Learner to establish recommendation strategies through reinforcement learning; and introducing human-in-the-loop methods in machine learning.
Currently, Squirrel AI Learning has set up a laboratory to carry out joint technology development with the Stanford Research Center. Squirrel AI Learning and the Institute of Automation of Chinese Academy of Sciences have set up a joint laboratory for AI intelligent adaptive education. This year, Squirrel AI Learning has set up a joint laboratory with Carnegie Mellon University (CMU), aiming to apply the most advanced artificial intelligence research results to teaching practice.
What is important to measure?
Daniel Schwartz is the Dean at the School of Education, Stanford University and an Expert in human learning and educational technology. At present, the laboratory that managed by Schwartz is conducting research on the basic problems of learning. Schwartz’s latest book, ABCs How We Learn: 26 Scientific Proven Methods, How They Work, and How Long They Work, refined learning theory into practical solutions and was listed by NPR as one of the “Best Books” in 2016.
Schwartz said that in a vibrant future, it’s important to measure whether and how people choose to learn beyond rigorous classroom instructions: Firstly, what we need to be measured now is the learning process, not just the results. This is important, because we need people to know how to learn.
Secondly, people actually know a lot of good learning strategies, and the key question is whether they choose to use these strategies. Finally, the core of design thinking is to avoid ending prematurely, for example, do not commit to your first idea. It turns out that it is actually good for learning, and can teach students the learning strategies they are willing to adopt.
Intelligent Agents, Knowledge Maps and Open Learning Data
Mark Musen is a Professor of Biomedical Informatics at Stanford University and the Director of the Biomedical Informatics Research Center at Stanford University. Musen is engaged in research related to intelligent systems, reusable ontologies, metadata for the release of scientific data sets, and biomedical decision support. His team developed Protégé, the world’s most popular technology for building and managing terminologies and ontologies.
Musen is also a Lead Researcher at the Center of Extended Data Annotation and Retrieval (CEDAR). CEDAR is a center of excellence supported by NIH big data to knowledge program. Its goal is to develop new technologies to simplify the compilation and management of biomedical experimental metadata.
Musen said that knowledge maps provide a formal representation of human knowledge and have the ability to link the concepts in these maps to related data sets and structures, people and intelligent agents can search these data sets and structures to discover new relationships between what we know and the data that may support these conclusions.
Technologies such as CEDAR can make the metadata describing experimental data sets easier to discover and reuse, thus providing opportunities to “Post” data linked to knowledge maps, thus enhancing the way we learn from scientific results
Musen believes that the results of the scientists’ work are not their published papers, but the data generated by their experiments. Current technologies make scientific data easier to find, access, interoperate, and reuse (FAIR). Future technologies will enable intelligent agents to help scientists understand information from data, plan new research, and make new discoveries.
Connecting the Learner in Learning Management Systems
Ajay madhok is a Founding Partner of Reboot Digital, a digital creative and marketing agency, and an Advisor of Playground Global, a venture fund. He has been involved in establishing four joint ventures, two of which were acquired and one was merged. At present, he is studying innovation models that large enterprises can practice maintaining competitiveness and correlation.
Madhok mainly introduced a digital learning management system designed by his team, which can provide the learning experience environment that students need. The goal of the system is to engage students to participate in guiding their own learning, help them set goals, track progress through dashboards, and select materials and challenges that meet their current capabilities.
Madhok said, “Effective learning depends on understanding students’ previous knowledge, experience, motivation, interests, language, and cognitive skills. Attracting students in the learning environment with emotional support can promote their sense of belonging, adaptability, initiative and learning achievements. Meanwhile, through the arrangement of interrelated concepts, students and learning are linked, so as to achieve personalized learning experience layer.
Meeting of Three Algorithmists and “Robin Hood“
Bruce Cahan is a Consulting Professor at the School of Engineering, Stanford University, where he designs and applies new theories to create financial and insurance markets, so as to improve the quality of regional living systems. Cahan is also the CEO and Co-founder of Urban Logic, a non-profit organization that uses finance and technology to change the way of systems thinking, behavior and feeling. Cahan has served as an International Financial Lawyer at Weil Gotshal＆Manges in New York for more than 10 years and as a Banker at Asian Oceanic in Hong Kong.
Cahan said that by 2050, AI workers could be divided into three categories: receivers collect and sell generic “Big Data”; amplifiers broadcast and seek to conform to any behavior or view that the government or business community want to fund; tuners ask whether, under the influence of all receivers and magnifiers, the results are beneficial to the people or small businesses that affected and manipulated by them.
According to Cahan, Stanford University is organizing research, teaching and practice to ensure that we apply artificial intelligence to business and social undertakings. Cahan believes that receivers, magnifiers and tuners cannot coexist naturally. Various ethical problems that arise in the development of AI, as well as challenges and threats to human society are all need to implement solutions as soon as possible.
Cahan proposed a concept of “Robin Hood AI Clinic” in his speech. Just as medical schools use teaching hospitals, or law schools use law clinics to check whether the diagnosis and treatment options learned and practiced by students and faculty really help real humans, Cahan believes that humans also need an institution like “Robin Hood AI Clinic” to perform similar teaching functions, so as to iterate and improve the possibility of AI comprehensively solving actual human problems and opportunities.