learningontheedge.com

While struggling to make sense of what may turn out to be an overeager attempt to visualize complex CCK08 network data, there are others working on a variety of visualization related things as well.  Matthias Melcher has an already famous visualization of the various aspects of the course posted a while ago on his blog   x28 .  He has also visualized the moodle threaded discussion “What is Connectivism” using a viz tool called deepamehta.  He has posted resulting jpg images on flickr.  Seeing how much info is involved in just a single discussion thread is causing me to rethink the scope of what I was intending to try.  Another interesting visualization is being done around twitter.  Tom Whyte is conducting a bit of a viz experiment around CCK08 participants twitter networks and the manyeyes tool.  His Blog has more information and links to the visualizations so far.  The more people that add to the information pool, the more interesting the graph gets.  I went through the process of gathering the information from my twitter network as indicated and sent it on to Tom, but I just had to dig a little deeper into manyeyes.  I couldn’t resist “borrowing” the data set and adding my own to it to see if I could get the tool I have been using to create a similar network graph.  After a bit of fiddling, I managed to create a cytoscape network and after a bit of playing around with layout produced the following graph that is fairly similar to the one produced by manyeyes.

This particular layout is formed using a spring-force algorithm.  Each node tries to get as far away from the others as possible and the edges act as springs pulling them back in.  In Cytoscape it is possible to play around with the edge weightings to factor in things like # of tweets to different members of your network and it allows for bi-directional edge creation to show up (mutual following) so I may continue (If it is okay with Tom) to play around with the growing data set in the background  One thing I noticed that I have to investigate a little further is the apparent thinning of the twitter network by the backup tool used to gather the list of friends (people you follow).  While browsing the network graph I noticed I was not connected to a few people I know I am following and found out that about 25 connections were not in the data set.  I checked some others and noticed that most were missing connections.  It is not a big deal, and it may relate to activity or based on when following commenced, or it may be a problem in the API,  Regardless, this is still very interesting and I have learned much working with the various tools and watching others do the same.

Now to the heart of the matter.  There are a number of others looking at this topic, so I will pour out my ideas related to the subject and then search out others and their ideas and continue to refine and perhaps work on this as a coordinated effort.  I locked in a network sample of about 450, whoever showed up as a participant on moodle yesterday.  I was able to creat a text file of names and format it in a way to allow it to be read by a program called cytoscape.  Currently I have a network consisting of 450 satellite nodes representing each individual and a central node CCK08 which more or less just means everyone is a participant.  This is where it is going to get interesting, and may take many attempts (have to save often so I can back up).  One of my goals is to find a way to represent each individuals participation in the network.  This can be done in many ways, but for my first attempt, I am going to try to do the following.  Surrounding each node representing the individual, I hope to create a separate node for each entity they produce related to CCK08.  Each blog post, Moodle forum entry/reply, blog comment, video, pod cast, diigo sticky, Facebook post, google groups, Twitter (this ones gonna hurt)   Each class of entity can be associated with a shape or color or other differentiating feature.  These can then be linked to other entities they are related to by an edge which will show up as a line connector.  For example if someone comments on a blog post, a line joining their comment node would link to the originators blog post node.  I am also thinking of creating nodes to represent the various “official” readings, recorded ustream sessions, elluminate sessions, forum threads, and other discussion threads etc.  That way a blog post or twitter or whatever that refers to one of those can link to the general node as well as a specific subnode.  Once some of the initial construction and linking is done, it will also be possible to go back and add a collection of attributes to the nodes and edges.  In fact, it should be possible to include things like tags, key words or even the entire (or abreviated form) of the original posting, in effect turning the node into a direct link to the raw information.  If I can figure out enough about how to operate cytoscape, it is possible to then analyze and restructure the network graph based on any number of algorithms to find clusters, subnetworks, related themes etc.   the list is endless.  It may be possible to do this manually for a while, but at some point, it would be nice to explore automating some parts of it.  Right now, I have a desktop aggregator pulling feeds tagged with CCK08 from several sources.  I doubt I will have time to explore this much this time around, but rss feeds are xml based, as is XGMML, one of the file formats cytoscape can use to represent network graph data.  It is forseeable that with enough skill (beyond what I currently have) it would be possible to parse some of the rss feeds, prepare the data in a way that would allow it to be appended to the appropriate xml file or entered into a database back end that is used by a script to dynamically generate a new updated XGMML file periodically that can be loaded into cytoscape or other XGMML aware visualization tool for viewing and navigating.  This could be done at regular intervals and each incarnation could be archived to provide raw materials for constructing a time sequence to show how the network grew and changed over time…     I will be happy if I can make any progress on the first few tasks I have set out, but threw the rest out as fodder for a next time event involving similar circumstances.  I imagine a lot of time could be saved initially by developing a process/tool set to capture some good information right at point of registration in a way that builds out the network graph from day one so learning as much about the overall process is still something worth investigating.

Well, it is just about the end of week one of CCK08 and I should begin writing more frequently here to clarify my own thinking.  It has been very interesting watching the interactions, the comments, the twitters, the blogs, the forums etc. that have all been related to CCK08 this week.  I have not been as active as I would like, but have contributed in a few places and have spent more time focusing on the one big problem that I would like to work on.  There are a few others now looking into how to visualize the network activity and I will make contact with them over the next few days to see if there is some common ground.  I have refrained from getting too involved with discussing aspects of connectivism and comparing it to constructivism and all the theoretical stuff.  I am not overly concerned with convincing anyone that patterns of interactions in networks can represent a type of knowledge (on some level) or that a form of  learning can be thought of as varied types of interactions or engagement with an ever changing jumble of networks.  I am not even overly concerned with convincing anyone that this reality seems to be moving into many areas of society, not the least of which is the corporate  realm.  There has been a lot of “talk” in the last few years about how it will be important for successful companies and organizations to understand and utilize some of the concepts surrounding networks, (at many levels,)  but not much has really changed.  The best part about this is that there will be no need to convince anyone of anything.  Companies that figure out how to truly utilize networks for competitive advantage will still be here a few years from now, those who don’t won’t.  I would rather leave all the deep philosophical ponderings to those more equipped, I can sense it and see the patterns emerging all over the place….that’s good enough for me for now.

Going back to what I would like to focus on is the representation and visualization of various aspects of the network that has formed around CCK08.  Despite those who have suggested that visualizing the network, seeing the patterns of interaction, the clusters, network growth patterns indicative of feedback cycles etc. is not that important, I am interested in seeing them anyway.  The question of Content vs Plumbing came up at least once.  The assumption that it is the “content” in the network that is most important and that the plumbing or configuration of piping is insignificant is not true in all cases and the two do not have to be distinct.  First of all, there are many examples where the information related to the pattern of interaction in the network itself does tell a story, and in fact may tell the better part of the story or the whole story.  In the last few weeks there have been a series of tropical storms causing problems in the atlantic-gulf region.  A content rich way of analyzing these storms comes in the form of very complicated fluid dynamics equations that can be set up, fed information and solved…sort of with the right assumptions, simplifications and access to a hi-performance computing grid.  This can be augmented with a variety of satellite images and put together to provide a pretty good understanding of the storm and where it is likely to head.  Another way to approach the same problem is to use a network of fairly simple devices that can continuously monitor a few simple states and communicate them to the other nodes and ideally to a central location.  As a storm interacts with the nodes in the network, the changing states are communicated through the network and can be used to develop a pretty good understanding of the storm and over time, after several storms, with proper visualization of the data, may show patterns that can be used to even better predict future behavior.  Techniques like this are already being used in many fields, and the overall knowledge/understanding/learning that takes place in scenarios like this still happen even though the individual bits of content being transfered through the network pipes are small and seemingly insignificant (even binary in some cases I’m on on on on ..now I’m off).    The biggest problem in most such situations is finding ways to represent all the interactions in some form and present them visually. That is then followed closely by the task of somehow tracking important interactions continuously over time.  I won’t even try for 1900.  I have notice over the last day or two that the number of participants showing up in a few central locations seems to be slowing down.  I have selected about 450 from the moodle group and am going to begin working with that subset.  This should give me enough to worry about.