Executive Summary


A shift from the traditional education paradigm to a new model seems inevitable and possibly very urgent, as evidenced by the CoVid-19 pandemic.  While lectures may be easily delivered online, how to realize other essential components of education, such as human interactions, examinations, lab experiments, etc., effectively in a large scale presents a new challenge to educators and students.   


We hope the challenge may eventually turn into an opportunity, with a marriage of AI (Artificial Intelligence) and education, where we use the term “AI&ED” to designate the shift from traditional education to its next generation. The new model has to be Evolutional - It has to not only support lectures, but also support other components of traditional education including examinations, disciplines, human interactions, lab experiments, social skills, etc. Beyond that it can be personalized, transdisciplinary, and collaborative:


Personalized: Traditional education is modular and structured - Students need to learn (all) basics of a subject before continuing to the next subject.  A major problem with this approach includes:

  1. Students who are stuck in one subject cannot proceed to the next subject.

  2. Students often do not see real-world problems until in the latter stage of education.

  3. Students may waste time on learning parts of a subject that are less useful in solving problems.


A personalized, problem‐driven approach may be offered as an alternative to solve the above problems. Students may be more motivated in solving real-world problems, where the problems may come from a library or posed by students themselves. Students who learn how to solve a problem may generalize his knowledge to solve similar problems. By solving multiple problems students are expected to cover the same set of objectives following different paths.


Transdisciplinary: Traditional education is standardized, that students are trained in one discipline. This model unfortunately may create boundaries in supporting interdisciplinary problem solving. Trans‐disciplinary education differs from the traditional education model in that the boundaries between disciplines are eliminated.

Trans‐disciplinary education may be facilitated again by a problem‐driven approach, as described in (A), where resources (data, knowledge, skills, etc.) from different disciplines are synthesized to solve a problem.

Collaborative: Collaboration is of fundamental importance to achieving excellence, as everyone is limited in knowledge: Teachers can collaborate to improve teaching, students can collaborate to improve learning, and, of course, teachers and students can collaborate to maximize the benefits of education.  Collaborative education includes group-based projects, sharing of resources and knowledge, and incentive-based collaboration.

The above can be realized with AI technologies that may be classified into four categories: automatic problem solving (e.g., search, knowledge representation, reasoning, planning), humanized computing and communication (e.g., computer vision, natural language understanding, robotics), machine learning (e.g., data science), and semantic computing, where Semantic Computing (SC) addresses computing technologies that allow humans to search, create, manipulate and synthesize computational resources (including data, documents, tools, people, devices, etc.) based on semantics (“meaning”, “intention”).


We propose to provide an evolutional, AI-based platform that also supports personalized, trans‐disciplinary, and collaborative problem solving, where “resources” for problem solving (including data, knowledge, tools, people, devices, etc.) can be dynamically connected to solve problems. The connection between resources can be made via (1) semantic interfaces, which interpret and understand problems; (2) semantic analysis, which analyzes the resources available; and (3) semantic synthesis, which integrates multiple resources into a solution. The “domains” in the framework are organized and connected both horizontally and vertically, based on the structure of knowledge at all granularities.