Towards a Knowledge-Based Model for Instructional Design

[akivela]

Koceva, Frosina: Towards a Knowledge-Based Model for Instructional Design. Proceedings of the Second Doctoral Workshop in Artificial Intelligence (DWAI 2014). Available at http://ceur-ws.org/Vol-1334/idea2.pdf

[Anisorf]

Dear Wandora Team and Community
First of all I would like to thank you for the grate work you are doing, I must say even after some month of using wandora, there are functionality that I'm still discovering.

Recently I was checking the L-system. I would try to express what I'm trying to do and I would like to have your suggestions on which is the best way to go for it.
I've build an educational schema for representing a subject matter (I define type of topics mainly Primary_Notion_Topic and Secondary_Notion_Topic) and type of associations (is_requirement_of; is_item_of; is_suggested_link_of; is_related_to). The is_requirement_of relationship is the most important since it denotes that a topic T2 is a prerequisite for topic T1.

I'm mostly simplifying the wandora interface in order to be used by the teacher which will design his/her subject matter( e.g., fundamentals of computer science) by specifying the topics and relating them with the association that I've mentioned before.
Once the teacher has finished designing his/her course, than in the graph I should distinguish between the Primary_Notion_Topic(a topic without incoming is_requirement_of arcs) and the Learning_Outcome_Topic (a Secondary_Notion_Topic without outgoing is_requirement_of arcs).

Now this are my questions:
1. After a subject matter is represented one or more than one learning path is/are created, that connects the Primary_Notion_Topic with the Learning_Outcome_Topic. In order to create this learning paths I was thinking to implement a kind of topological ordering algorithm, but is there other way to do so ? Maybe using the L-system and validating the graph in respect of a grammar? .
2. In the graph view which is the best way to go in order to represent oriented associations. I saw the javascript d3graph solution that you suggested in some previous post, but I was wandering if there are some other ways to implemented for the graph view.

Any suggestions will be more than welcome.
Thanks again for the awesome work .

[akivela]

Hello Anisorf

Thanks for using Wandora. I am very happy that you have found the application useful.

About your first question.

Wandora's L-system generator [1] is used to generate topics and associations (or nodes and edges in graph terminology). In this sense the L-system generator can be used to create learning paths i.e. topics and associations connecting the Primary_Notion_Topic and Learning_Outcome_Topic. However, all topics and associations created by the L-system generator doesn't express any real-world concepts, they are kind of subjectless. L-system generated topics and associations are meaningless concepts and the user has to modify those later on to make them express something meaningfull, like learning paths. This extra work may ruin the usability of L-system generator. On the other hand, experimenting with the L-system generator and figuring out rules that geneate the learning paths may be very interesting as a research problem. For example, figuring out that certain type of learning paths are topologically similar and can be expressed with certain L-system rules. Indeed, very interesting research problem.

Thinking a practical solution to easily produce learning paths, I would probably use the Sketch grid [2] and it's "Make associations using Wandora layout" feature. I would probably make the associations in Excel (or OpenOffice Calc) first, and then copy/paste the cells into the Sketch grid, and finally create associations. Notice, you can paste the Wandora layout associations directly to Wandora [3] via system clipboard. Thus, Sketch grid is optional step.

Yet another practical solution is to use Wandora's Excel extractors [4], perhaps the "Excel adjacency matrix extractor". However, Excel extractor is little tedious as it uses default association type and role topics instead of type and role topics that would match better learning paths. Again, the user must change association type and role topics to finalize the work. Perhaps I could change the "Excel adjacency matrix extractor" to support user specified association type and role topics. What do you think?

About your second question.

I am afraid that current Wandora version views all associations without orientation, arrows at the end of edges. A kind of orientation is achieved with the Graph topic panel [5] as it can view role names over the edge. And skilled developer can indeed hack Wandora's D3 graph service module [6] as described in [7]. As the feature has now been requested several times, I'll add somekind of native support for arrowhead edge visualizations of associations to my working list. But I can't give you schedule for the oriented associations feature.

Kind Regards,
Aki / Wandora Team

[1] http://wandora.org/wiki/L-system_generator
[2] http://wandora.org/wiki/Sketch_grid
[3] http://wandora.org/wiki/Transferring_data_with_clipboard
[4] http://wandora.org/wiki/Excel_extractors
[5] http://wandora.org/wiki/Graph_topic_panel
[6] http://wandora.org/wiki/D3_graph_service_module
[7] http://wandora.org/forum/viewtopic.php?f=5&t=136&sid=6fbce12deab64e91cc027316f1f69990

[Anisorf]

Dear Akivela, thanks for the fast replay and the suggestions.

1. What I'm trying to do is to automatically generate the learning paths, i.e. all of the possible topological sort outcomes, from the given topic map representing the subject matter. I saw also the Linear List Generator which if I'm not wrong generates a new list of topics and associations of a chosen type. I could think to try to implement here a modified topological sort algorithm. I'm saying modified because I need all the possible learning paths (permutation consistent with the partial order of the topics) from a given Primary_Notion_Topic to a Learning_Outcome_Topic. Or maybe I should generate the learning paths as new layers. What do you think?

2. I've hacked little bit the D3 graph in order to get oriented association. I'll try to implement oriented association also for the graph panel, I'll share the code.

Thanks again.
Best Regars,
Frosina

[akivela]

Hello Frosina

Sorry for the delay. Had a short vacation.

What do you consider a learning path? It connects the Primary_Notion_Topic and Learning_Outcome_Topic but it is a bit unclear to me if the learning path is an association or a topic or a collection of topics and associations? Can you provide a simple example of a Primary_Notion_Topic and a Learning_Outcome_Topic and learning paths that connects these. It would help a lot.

Linear List Generator [1] can be used to create topics and associations. Created topics are associated as topic chains resembling linear lists. The most recent Wandora release allows the user to change default association type and role topics. By default the association type and role topics are subjectless i.e. they don't represent anything special, just like the topics and associations the L-system generator outputs. At the moment I am a bit unsure if the Linear List Generator can solve the problem of creating learning paths.

If you create useful changes or/and additions to Wandora, you can always commit those to GitHub [2] and send us a pull request. I warmly welcome all kind of commits, although can't promise we use/integrate all of them.

Kind Regards,
Aki / Wandora Team

[1] http://wandora.org/wiki/Linear_list_graph_generator
[2] https://github.com/wandora-team/wandora

[akivela]

By the way, have you looked at the Fully Connected Graph Generator [1]. It creates topics and connects all of them to every other topic, giving a complete graph [2]. Is complete graph something you are looking for to represent the learning paths?

[1] http://www.wandora.org/wiki/Fully_conne ... _generator
[2] https://en.wikipedia.org/wiki/Complete_graph

[Anisorf]

Dear Akivela,thanks again for the big support you are giving me.
I would like to apologies for my unclear explanation and poor English.

Let say that the teacher is creating the map that represents part of the Computer Science domain (see picture bellow). In this map for simplicity I did not expand the Occurrence nodes which are the learning material. At this point the teacher decides that the Primary_Notion is Information and that the Learning_Outcome is Language. Than a topological ordering algorithm generates a subset(a sequence) of topics and properties from the map. One of the possible subsets is:

{Physical World, Digital World, Analog Signal, Digital Signal, [Digitization], Analog Information, Digital Information, Representation of Information, Encoding, [Binary|Hexadecimal|Property|Decimal], Language}

where in square brackets we put the topics that are optional in our learning path (since connected with the association without prerequisite constrains: is_item_of; is_suggested_link_of; is_related_to).
Formally, the is_requirement_of (i.e.,is_req) association order the topics T of the lesson according to the propaedeutics rules, therefore in the graph G=(T, E) there cannot be loops, thus obtaining a Direct Acyclic Graph, where T are nodes and E arcs, with: (Ti, Tj) ∈ E ↔ is_req(Ti, Tj). In this context, a Topological Order on a graph is a sequence S = {s1, s2, … s|T|} where each element T appears only once and cannot be preceded by any of its successors; given pair of nodes (Ti,Tj) in S if there exists an arc from Ti to Tj of type is_req, it follows that the node Ti is before the node Tj in the list: ∀(Ti,Tj) ∈ S: (Ti,Tj) ∈ E → i < j.
Tecnicly speaking I was thinking to generate the sequence of the learning path by defining the partial_order association between Ti and Tj. Thus the lesson plan is composed of topics, occurrence and the partial_order association. But for the partial_order association I'm not yet sure.

The idea is than to use this topological order for:
- editorial purposes the topological order can be exported in a XML format, which maintains the structure of the lesson plan and makes it possible to be imported in a text editor for further adaptations.
- the XML format can be imported in LCMS, giving an initial shape of the course and the learning materials.
- html pages can be generated transforming the propaedeutic order in navigational.

After checking the suggestions I think I should implement the generation of the topological order as a new tool for wandora. For the first prototype version of my tool I will also implement the oriented associations directly modifying the VEdge.

Computer Science ECM.png (187.27 KiB) Viewed 762879 times

Thanks again.
Best Regars,
Frosina

[akivela]

Hello Frosina

Thank you for explaining the concept of learning path. Now I understand you better. Learning path is a graph path and required techniques include pathfinding [1] and graph travelsal [2]. You don't need to create any new topics nor associations as the topic map already contains everything. Thus, you don't need any generators of Wandora. Generators generate new topics and associations algorithmically and that is something you don't want to do while finding the learning paths.

Now, thinking of Wandora's features that resemble pathfinding even in a distant manner, one specific tool comes to my mind. Wandora can copy a graph path i.e. chain of player topics in associations. This feature is labeled as 'Copy path to the edge of associations' and you'll find it by right mouse clicking in an association table. The idea is that Wandora traverses associations of certain type and builds a list of topic names faced during the travelsal. The feature was originally programmed to copy a path to the tree root node but it can be adjusted to copy more general paths too, I believe. The feature is implemented in Java class

org.wandora.application.tools.associations.CopyEdgePath [3]

I suggest you don't modify the CopyEdgePath class but create a new tool class for your copy-learning-paths or export-learning-paths feature and copy useful parts of the CopyEdgePath class. One obvious limitation of the CopyEdgePath is that it doesn't copy all possible paths but stops happily after the first path.

Happy hacking.

Kind Regards,
Aki / Wandora Team

[1] https://en.wikipedia.org/wiki/Pathfinding
[2] https://en.wikipedia.org/wiki/Graph_traversal
[3] https://github.com/wandora-team/wandora ... ePath.java

[Anisorf]

Dear Akivela,

I'm going on with the work on hacking Wandora, and at the start of 2016 there will be the pre-alpha wandora customized version for design of educational concept maps and generations of lesson plans.

One of the next things I would like to represent is the Topic Effort , which is a weight added to every occurrence. Since in my model the topics are concepts representing the subject matter and the occurrence are learning materials, thus the topic effort is the time needed to study the occurrence. I'd like to add this additional information (which is a number) to the occurrence edge.
I was thinking is it possible to model it with a kind of tri-ary occurrence edge (topic, effort, occurrence) ? Or should I hack the occurrence class ? I'm not sure how to store this additional information.

Thanks in advance for any advice.

Kind Regards,
Frosina

[akivela]

Hello Frosina

Great news you have made progress.

I think it would be difficult to add the Topic Effort value into the same occurrence structure as the learning material. Also, I would think twice before hacking the occurrence class. Hacking the occurrence class would probably cause some other issues.

How about using parallel occurrences for the Topic Effort. By parallel occurrences I mean the Topic Effort occurrences are just separate occurrences hanging on the same topic as the learning material occurrence.

Kind Regards,
Aki / Wandora Team