My guess is the one on the left is Al Qaeda due to the multiple connections between cells, few inter-cell connections, and larger central hub.

In industry, cells/departments have specific patterns:
- Outlying clusters usually have 1 connection since there is only one manager or one external contact point.
- Leaves of a cell may connect to leaves of other cells. E.g., a developer may call a hardware tech for questions. This bypasses the central hub.
- The core is tight -- few external contacts. This is due to a core management team (e.g., CEO + VPs).

The graph on the right shows all of these features.

For terrorist networks:
- The main hub has many cross-connections. This prevents a single loss (capture/kill) from breaking the entire network. (Industry does not worry about this since subordinates are documented. Terrorists usually do not document structures since documents could compromise their network. Redundancy is used in lieu of documentation.)
- Cells may have cross-communication with parents, but are isolated from other cells. The multiple connections to parents show the communication redundancy. There is no intercommunication between cells because they do not know each other exist.

This matches the graph on the left.

Then again, I could be wrong.
A lot of this depends on the source of the data, duration of the collection, and scope of the graph. Is this a single Fortune-500 company or a department. Do the graphs span a week or a year?
Are the graphs from phone trees, network connections, IRC, IM, or something else?

Posted by: Dr. Neal Krawetz at May 17, 2006 8:59 AM

Could someone please post those graphs in the form of an adjacency matrix, it would make the analysis a little easier than trying to do it by eye. Also, there is something very dubious about these graphs, which is to suggest that all of these interactions are somehow equal in weight.
One immediate comment, is that the graph on the right at least appears to my eye to have a generally higher average degree than the graph on the left (i.e. greater average number of neighbors). At the same time the graph on the right has many more nodes of degree one.
Is it obvious that "splinter cells" have project related interactions with other cells at all? It seems that identifying the graph based on these graphs depends critically on exactly what is the definition of what creats an adjacency (a link).
Just my thoughts. It would be my opinion that there is not enough information here to make the determination without at least substantial prior knowledge about the nature of graphs of known entities (other fortune 500 companies, other clandestine networks).

Posted by: Aaron at May 16, 2006 2:57 PM

Its the cell on the left of course:)

Seriously. the cell on the left has spinter groups that are not connected and that is how I would picture a terrorist group that want to avoid all members knowing each other. The one on the right has everyone talking ot anyone - this is not secret enough.

Companies and normal social groups don't try to hide there alliances and give false trails where a secret organization would.

Posted by: J at May 16, 2006 10:28 AM

Ive been thinking about it.
It ´s strange how social webs looks like terrorist organizations, rigth it´s the correct because all the cells have been organizated with almost three persons and the staff departament of fortune only have two.
But it seems like a normal democracy social web design, it´s like a signal of how all the dinamics moves of the man , like democracy or the ancient regime, make another dinamic in the same way but in diferent direction, like terrorist or revolucionary strikes like the french revolution.
We have to take care in what are we working to stop Al quaeda dimension of operative cells, if we make another dinamics, like make a stronger state with non legal moves( "non legal" because "Auctoritas, non Veritas") because we are working at the same time in oppositte direction.

Posted by: Negroi at May 15, 2006 10:33 AM

Given that the fundamental basis of criminal or terrorist cells is that there needs to be minimal knowledge of the rest of the organisation within each cell, neither of the diagrams presented fits the model. Parts of each one do, but they are not consistent. A terrorist cell model would suggest a web of concentric circles, with very few connections between the circles, the point being that people in the outer rings can't identify the people further in.

Posted by: Tony James at May 15, 2006 8:52 AM