Can network drawing tools detect actual communities, distilling and distinguishing the data in terms of shared topical and genre characteristics?
The short answer is: Yes, at least sometimes.
Consider the blog of Ze Dirceu, the one-time Workers Party rainmaker recently sentenced to prison. One hand, it contains technical elements — social platforms or “commons” through which the content of the network is processed — and the content of the network itself.
You can think of this view as “The Big Brain” or “Phrenology” model.
In this analysis of the 21212 Demo Day held back in October, for instance, the network roster discovered was almost equally divided between a list of candidates for venture capital on one hand, and bilingual, neoliberal blog farms and think tanks, on the other. Two major news and media organizations — Globo and iG — were also present. Globo, as we know, has developed an ad campaign for PEGNs — “small ventures, big companies” — into a regular TV show and a print magazine.
Also playing a role is Totvs, the up and coming, extremely acquisitive software house with the goal of becoming something of a Brazilian Oracle. Movile and 21212 — new names, to me — did the digital marketing.
I take an unfiltered, directed network, subject it to the Fruchterman-Reingold transformation, assign the nodes to partitions determined by the modularity score, and then scan it visually to see whether any familiar relationships appeared — I know that CartaCapital magazine is feuding with Veja magazine and lo and behold, structural holes may appear in the sociometric data which can be explained in terms of this antagonism.
The mountain of data on NGOs I have collected over the last year or so has familiarized me well enough with the major philanthropic networks that exploring this topic makes for a good benchmark.
These, after all, reflect digital strategies devised and executed by digital strategists. It should not surprise us to find aspects of the VoBE — Virtual Organization Breeding Environment — in action. It is an exercise in reverse engineering.
Consider the drawing, above. It depicts the ego network of a very ambitious public-private startup network recently introduced to get entrepreneurs together with VCs.
Open the cluster of Startups iG and you get an extensive roster of struggling artists and coders, which might be valuable in itself.
Open other clusters to find discover angel and VC candidates under the aegis of the government and private sector.
In fact, the graph shows three what look like three paths open to those seeking venture financing: the academic — INSPER, ESPM, USP — the state-sponsored — FINEP, BNDES and the Ministry of Development — and the attraction of private capital by the first two, a task that is left for APEX, the export promotion agency. Publicity is managed by Ernst & Young and Movile, along with 21212.
The next step to be taken is to inspect the boundaries between the modules.
The blue partition is mixed, but there may be some crosswiring with English-speaking venture people. A centrality test might be in order here. How strongly are the linked in to Linked In and who do they link to?
Globo publications and incubator projects.
This view seems less coherent.
Events Events Events
As to Endeavor:
Overlooking the value of lines, the beating heart of 21212 Demo Day was Endeavor, a wealthy and powerful entrerpreurship program with ties to the Kennedy School at Harvard and an international presence elsewhere.
Endeavor‘s dot.org in Brazil is distinct from its .com.br, oddly, but they represent the same program, which has worked very closely co-sponsoring events such as TED and Campus-Party with Instituto Millenium, a Cato-style think dominated by media owners and their supporters.
An egocentric subnetwork of 2,000 nodes is extracted from crawled data. The data is filtered and modularity is applied to it. Adjacent groupings are found, and we ask ourselves, what do these have in common that allow us to call them communities, if anything?
It is very common to find large, highly coherent clusters around the social networking sites and portals — Linked In in particular creates a lovely diagram. These are often positioned as hubs, directing attention toward some authority.
A recent event — Semana Global, a week’s worth of entrepreneurship training — illustrates the modular, plug and play of such events.
Conaje is the National Federation of Young Entrepreneurs. Facisc is the “federation of business associations” of the southern state of Santa Catarina.
With time on my hands, I decided to conduct an experiment to see if I could learn the Gephi filtering tools a bit better. It is very simple: Compute the Big Brain, dissect it in terms of modularity, and inspect the resulting modules for their structural properties and content.
Brazilian bloggers, most of which identify with the “progressive” label.
Reflects the origins of the crawl in the conservative reader Arts & Letters Daily.
Only factor in common: hosted by WordPress.
Filed under: Brazil