Adaptive Learning: How It is Learned or What is Learned?
Joseph M. Scandura
There has been a major disconnect since their inceptions between theory, research and instructional application in structural learning (e.g., Scandura, 1971, 1973, 1977) and in what later became known as cognitive science (e.g., Brown & Burton, 1978, Anderson, 1988). With few exceptions (e.g., Scandura, Koedinger, Mitrovic & Ohlsson & Paquette, 2009), most of this work has been done in isolation.
Consequently, many ITS researchers have a basic misconception of SLT and its now proven benefits in developing and delivering dynamically adaptive tutoring systems (e.g., Scandura, 2013b). Because the general goals are similar, many investigators use the same lens to evaluate what are fundamentally different paradigms.
ITS development has traditionally been based on the assumption that it is essential for teachers to understand what is going on in student minds – to understand how students learn. Many ITS systems are based on often elaborate cognitive theories (e.g., Anderson et al, 1988, 1995, 1998). Development of dynamically adaptive tutoring systems based on SLT is fundamentally different. SLT is equally precise, but is more comprehensive in scope. It is explicitly concerned not only with learning but with teaching as well (e.g., Scandura, 2001, 1977). Differences between how experts and novices solve problems is critical in purely cognitive theories, where the goal is detailed explanations on what goes on in student minds.
This is important in SLT, but it is handled very differently. Teachers know that students with different degrees of expertise need different kinds of help at various times during the course of learning. Among other things experts generally deal with bigger “chunks” of knowledge. They typically get to desired results far more quickly and/or effortlessly. From an SLT perspective, all students learn in the same way. What differs is what students know at each point in time relative to what is to be learned. This is what determines what the teacher (or automated TutorIT tutoring system) should do next – for example, whether to test or teach, and what to test or teach. How something is taught is strictly secondary.
How is this possible? We have found a way (based on SLT) to systematically represent all knowledge in a uniform way that encompasses ALL levels of expertise simultaneously. All knowledge, whether expert, novice or anywhere in between, can now be represented systematically based on recently patented methods and technologies (Scandura, 2013a, US Patent No. 8,750,782, June 14, 2014). To be acquired knowledge is represented hierarchically, simultaneously at all levels of abstraction (represented as Abstract Syntax Tree-based SLT rules). Individual knowledge at any or all levels of expertise is measured relative such SLT rules. Using overlays is not new. An essential part of what is new is representing individual knowledge as overlays on hierarchical knowledge representations and automatically drawing inferences about mastery on other items (see our patent for details). Another essential is introducing higher order SLT rules that operate on and either generate or select SLT rules (cf. Scandura, 1971, 2007).
In short, the focus in SLT, and the AuthorIT and TutorIT technologies based thereon, is on what subject matter and instructional design experts (SMEs) believe should be learned for success. This to-be-acquired knowledge provides a uniform method for representing individual knowledge – whether expert, novice or anywhere in between. Methods reduced to practice in AuthorIT and TutorIT technologies make it possible to represent all knowledge hierarchically, simultaneously at ALL levels of abstraction.
Using these patented methods to develop TutorIT tutorials has made it possible to avoid complications cognitive scientists have long faced in building ITS. Focusing concern on how experts, naïve etc. students solve problems adds significant complications – precisely because there are any number of gradations and variations. The solution implemented in AuthorIT is to represent all levels of expertise simultaneously. TutorIT then automatically uses these hierarchical knowledge representations as a measuring device to identify what any given student in a given population does and does not know at each point in time delivering precisely what the student needs to progress. This new paradigm has proven to simplify the task of building and delivering dynamically adaptive tutoring systems by an order of magnitude.
Joseph M. Scandura
July 4, 2014
scandura@scandura.com