Sunday, January 3, 2010

Hierarchies and Nets

KV asked how Luhmann avoided a hierarchical structure in his Zettelkasten. The answer depends to some extent on how strictly you interpret the word "hierarchy." In some sense, any ordering that has 1 followed by 1.1, 1.1.1, etc. may be interpreted as hierarchical, as "1.1.1" can be viewed as depending on 1. But it does not have to be so interpreted. It could simply indicate a sequence.

However, this is not the kind of hierarchy Luhmann wanted to avoid. What he had in mind is a more substantive hierarchy, where one concept contains other concepts or where all the concepts in a certain domain depend ultimately on one basic concept. Trees represent this kind of hierarchy:

His Zettelkasten does not have this structure of an acyclic graph, but is rather a flatter graph, in which there is not one central node or concept, but in which everything can refer to everything. Like this:

It's more like a network. Luhmann himself used the metaphor of "spider web."

A tree is in fact only one (very specialized) form of a graph, as you can see from this:

Outliners basically follow the tree model and have difficulties accommodating others.

Luhmann's system is not particularly tied to any of these and can accommodate all of them to some extent. Therefore, again, he does not have to avoid any hint of hierarchy.

See also Outlines and Meshes and Outlines and Hypertext. Brainstorm may be viewed as a non-hierarchical outliner.

1 comment:

KV said...

Thanks MK for the explanation. I think perhaps I have attached too much meaning to Luhman's use of the numerical ID's on his cards. As you have pointed out in computer based systems this is not necessary.

However, I have taken some inspiration from the concept and have extended it for several months now to the simple case of creating and naming computer files. My problem has been with hindsight I often become dissatisfied with the name I have assigned to a given file and therefore change it. This results in the inability to locate that file when running across old references to it.

I now create files with names like, "FileName 20100110T2121.pdf" where the ending is a date and time stamp which is unique. Then whenever I run across a reference to this file name in my notes, I can copy just the date stamp portion and paste this into OS X's Spotlight or Windows Search where it is sure to be a top hit. As long as I never alter this date stamp when renaming a file, it can be found. An added benefit of the Spotlight and Windows search feature is that besides showing just the file itself they will also show other documents that reference it. This makes a bidirectional hard link of sorts.

I think a loose version of Zettelkasten could be created using text files (and other files) with a unique ID embedded in their file names that allows Spotlight, Windows Search, or grep to tie all the links together.

Thanks MK and please continue writing these most interesting blogs!