Artificial Intelligence (AI)

The MindWalk IDT Protocol uses an AI algorithm to drive its personalized learning experiences. Here is more detail.

Intelligent Inference Engine

The Intelligent Inference Engine serves as the “brain” of the learning process.  It analyzes learning behavior and outcomes in real-time and, based on its analysis, chooses what will happen next across a range of learning parameters.  These include which content to present, the desired level of skill proficiency and the desired learning modality. The intelligent inference engine uses a combination of commercial off-the-shelf authoring tools, database tools, and custom code to achieve the planning and presentation of course material.  At present, the student primarily interacts with skill assessment (tests, exercises, simulators) and topic presentation material). The skill assessment material is responsible for determining correctness of the student’s answers and actions, and assigns a “score” or judgment of a student’s mastery level of a particular concept. Presently, the assessment is a real number in the range  [0.0 , 1.0], where 0.0 represents no mastery, and 1.0 represents total mastery of the concept. The results of the student assessments are relayed to the curriculum planning and assessment routines.  Based on student performance, the assessment and planning routines determine which material should be next presented to the student.   The selected topic identifier is passed back to the skill assessment component, which then presents the designated material to the student. An alternative mode of interaction provides for the student to choose via menu a topic he or she wishes  to pursue. This  choice  is conveyed  to the assessment  and planning routines,  which abandon their current  planning goals in favor of supporting  the express goal of the student. Based on  the new goal, a new plan is generated. A consequence  of this “mixed-initiative” approach to training is that the student must become more active in generating goals once the student has abandoned the IDT curriculum plan.  Support for mixed initiative is an ongoing area of development for MindWalk.

Presentation Control System / Student Interface

The  Presentation  Control System  and Student Interface  module support the presentation  of domain independent information and  tests to the student, grading of these  tests, and reporting results of the presentation or testing to other system components. This module uses off the shelf components for multimedia presentation of course material. Student’s control and participation is interactive, rather than passive.  This module presents profiling questions, the results of which are passed to other system components for adaptation of the material. Used in combination with other system components, the resulting training is a personalized and individualized experience.

Tests that are presented to the student are intended to test their proficiencies and weaknesses in a  particular domain, and test students against mastery criteria. As appropriate, tests are randomized to prevent memorization of previous interactions. Students must demonstrate competency in a particular concept before proceeding to dependent concepts.

Presentation of course material often includes simulations, exercises, and coaching / tutoring on the  material. Multiple communication methods, including animation, sound and video, life-like actors,  and hands-on practice are used to present the material. The interface is a simple and natural one, matching the experience of the intended training audience.

The  module  allows students  to edit voluntarily  at any point, and bookmarks  a student’s instructional and associated data files upon exit. The bookmark is used to restore the environment exactly as the student left it once the student re-enters the course.

API

The  API defines  the interface  (communication method)  between the Presentation  Control System / Student Interface and the Curriculum Planner. It provides methods for initialization and restarting,  loading a particular Curriculum Plan into the planner, loading a particular student information file into the  planner, running the planner for a controlled or uncontrolled number of steps, inserting new knowledge (facts) into  the planner knowledge base (e.g., the results of a student test), and scanning the knowledge base for pertinent facts or knowledge (e.g., what is the next item to present to the student). The API  allows the  Curriculum Planner  to be used with a  variety of off-the-shelf  tools for the implementation of the Presentation Control System / Student Interface.

Curriculum Planner

The   Curriculum   Planner operates   upon a network of   domain-specific concepts   or topics (curriculum graph), which is annotated with the system’s assessment of the student’s knowledge or skill level for each concept.   The result is to derive a plan and recommend instructional and testing material for the student. In addition to the curriculum graph and student assessment, the planner also uses the student’s preferences for media types and learning style to choose among alternative instructional approaches.

The Curriculum Planner itself is a domain-independent engine, and can operate on any properly formatted  curriculum graph. The curriculum graph content is, of course, necessarily domain- dependent, and  must be designed by the curriculum author. The graph specifies alternative instructional materials,  whereby the planning engine adapts to the student by selecting over available teaching methods for a particular lesson.   The planner can alter the plan at any time, based on the student’s demonstrated proficiencies and weaknesses in the domain.

The  Curriculum  Planner has  two overall operational  modes. The first is an assessment,  or “qualification” mode, in which the planer attempts to test the student to determine a general level of  competence  in the domain. The  second is a teaching  mode, wherein the planner  generates paths of instruction, to take the student from competence demonstrated upon first encounter to full mastery of the material.

The curriculum graph representation is not limited to a narrow domain category or instructional strategy (pedagogy). The author of a new course can alter or extend the number and types of links in the graph to reflect preferred strategies or theories.   The curriculum planner will consider numeric weightings that represent the student’s preferences for these strategies, along with outcomes resulting from their application.

As  plans  are developed  and portions of  the plan carried out,  the curriculum planner is  able to recognize when the current plan should be altered or abandoned in favor of some other valid plan. The  curriculum author can influence how tenaciously the planner adheres to its plan before attempting another route to the goal.

Visual Curriculum Graph Editor

To ease the burden of creating concept maps for each new domain, a visual editor is being developed to  allow a course author to visually lay out the curriculum map. This consists of the topics (concepts) for the domain, along with instructional interrelationships (links) between the concepts. The author can draw the concept-nodes, set links between nodes to specify instructional materials or   prerequisite relationships, express new instructional strategy parameters, reference instructional and testing presentation pieces, and declare a sequence of expertise levels to which concepts may be assigned. After completing the course, the author can then generate the appropriate  files to be used by the other components of the system, such as the curriculum planner and the student interface.