ActiveMath FAQSearch |
What is the AI in ActiveMath?ActiveMath integrates many different AI techniques such as knowledge representation, learner modeling, or tutorial planning. Knowledge representation: ActiveMath’s basis is its semantic knowledge representation. It includes two markup formats. The first is extended OMDoc, a markup format for mathematical documents. The second is OpenMath, a markup format for mathematical formulae. Apart of this, ActiveMath incorporates an Ontology of Instructional Objects (in short: OIO), a representation of pedagogical rules used to describe learning scenarios, and a representation for the encoding of sophisticated exercises. ActiveMath’s database is a set of single atomic learning objects. each object carries additional metadata such as, difficulty, abstractness, typical learning time, context, educational level etc. Also, learning objects are connected with each other by relations. One of the frequently used relations expresses a pedagogical dependency between two learning objects. That is, “What do I have to know, in order to understand the current learning item?” One of ActiveMath’s key features is the inclusion of Pisa competencies into the knowledge representation. For each learning item and for each step in an exercise one or more competencies can be assigned to. Using this information, the learner model can derive whether a user has deficiencies regarding the Pisa competencies, and the course generator can generate specific courses that train a Pisa competency. OpenMath enables ActiveMath to address different Computer Algebra Systems, to search for learning items that comprise a sub formula, or to perform a drag-and-drop operation from the ActiveMath browser into the plotter or input editor. Knowledge representation is a clasic AI topic. References:
Learner Model: ActiveMath tracks a user’s actions including all steps in exercises. From this information it tries to assess a users’s performance. Input information going into the learner model is for instance: was he able to (partially) solve an exercise, if not what went wrong, if he gave up then in what step? User modeling is a classic AI topic. Tutorial Planning: The tutorial planner can generate courses that are adapted to the user’s goals, context, educational level, knowledge, and whished scenario. At it’s heart it uses a hierarchical task planner algorithm. Planning is a classic AI research topic. References:
Trackback URL for this post:http://eds.activemath.org/en/trackback/96
|
Modifiablity and Customizability of ActiveMath's Meta-tagging?
Thank you for the lucid exposition. Some reference links to detailed documentation would be welcome, i.g. OIO, Pisa Competencies, et cetera.
Could the pedagogic meta-tagging be adapted to cater for heuristic and problem-solving frameworks? I am thinking, among others of Pólya, Wagenschein, Weihnacht, Wittenberg, as well as methodologies, strategies and leading ideas developed for the training and preparation of IMO teams by Engel, Larson, Gardiner, Zeitz, et. al. ?
If so, how?
As this is a forum I’ll
As this is a forum I’ll reply in a primitive way. Personally I think meta-tagging is overrated. All content should be “meta-tag-prepared” for the future, but extensive 60-fields meta-tag overviews are not sensible as long there isn’t enough interactive, pedagogical and didactical correct content. This has also been a discussion in Holland with SCORM. I would aim for “Runtime” content packaging, but now millions have already been spent on “crap” content with nice meta-tagging.
Could you please explain?
Sorry, but I don’ t understand.
Sorry I’m unclear. What I
Sorry I’m unclear. What I mean is that meta-tagging data for me is of little importance, UNTIL more content is made available. Furthermore, I’m discovering that clicking the forumposts on the right of the homepage, brings me to the post, but not to the book article. In this case I reacted to ‘meta-tagging’, discovering later that the article had a broader context.