Modeling Human Tutors

• In Event: 34.075 - Symposium on Intelligent Tutoring Systems, BIG DATA–Learning Analytics, and Automated Humanlike Tutoring: Similarities and Differences

Abstract

TutorIT is based on the Structural Learning Theory (SLT) designed to work more like a human tutor. It provides specific feedback based on what each student needs and when. Good human tutors know precisely what the student must learn. To be successful, the student must know what to do at each step of the process – what decisions to take, what actions and when. At each stage of the process, TutorIT lets the student know whether an answer is right or wrong, and directly or indirectly what should be done and when.
TutorIT uses patented processes to quickly determine what part(s) of what needs to be learned any given student knows at each point in time. Student knowledge changes incrementally as result of learning. TutorIT does this with optimal efficiency. TutorIT tests and/or provides remediation on every step and every decision students must make to be successful. When a problem comes up, TutorIT pinpoints precisely what the student is missing. Like a good human tutor it remediates on the spot.
Given 4802-3489, for example, TutorIT is not satisfied with just the answer. TutorIT evaluates each step in the process and like a human tutor takes appropriate action. It determines, for example, whether or not the student knows a basic fact, can regroup from the next column and what to do if the top digit in the next column is 0. The student must demonstrate not only what to do but when to do it. Unlike other so-called adaptive learning, the process stops only after a student demonstrates the level of mastery predetermined by the author. Where desired, mastery levels can even easily be adjusted by the teacher.
AuthorIT and TutorIT rest on a theoretical foundation that is very different than Intelligent Tutoring systems (ITS) or Big Data (e.g., Author, 2007, 2014a). Given any content domain, the AuthorIT authoring system guides authors as they identify WHAT students need to learn for success (Author, 2014a). The result is an arbitrarily detailed knowledge representation that accommodates the needs of all students in any given target population. In short, AuthorIT is used to represent the knowledge to be acquired. AuthorIT is based on Structural Analysis (SA) in the SLT used to represent to be learned knowledge. Unlike other forms of task analysis, all knowledge in the SLT has both declarative and procedural aspect. This is accomplished by representing ALL knowledge hierarchically (e.g., Author, 2007). SA accommodates multiple solution methods.
TutorIT takes any such knowledge representations as input and makes all pedagogical decisions automatically – without any programming. The result is a highly efficient authoring system AND a highly efficient tutoring system that applies universally to ALL content.
AuthorIT and TutorIT are distinguished by two unique features. First is AuthorIT’s unique ability to systematically represent what needs to be learned for success -- simultaneously at arbitrarily many levels of expertise. Second is TutorIT’s unique and general capability to make use of such representations to make all pedagogical decisions (what to test, what to teach and when) automatically.

Author

• Joseph M. Scandura, Merge Research Institute