Tuesday, October 20, 2009

Confession Time: Analyzing the Structure of Emergent Networks in the PostSecret Community

What can social network mapping show us?

As described by Marin and Wellman (2009), social network maps explain behavior by providing insight into interpersonal relations instead of individual attributes. For this reason, it is logical to use these conversational maps to analyze differences between neutral online forum discussions and those that touch upon personally sensitive, confidential, or taboo topics. Individual willingness or hesitation to contribute is determined by the environment fostered by the nature of the conversation, and social network mapping provides a concrete representation of this interdependent relationship.

Will Postsecret networks be different? If so, why?

We predict that conversations within forums that require self disclosure and revolve around socially taboo topics will solicit unique patterns of interaction form those centered around neutral discussions. Our reason for believing this is twofold. Firstly, the highly individualized nature of the topics discussed may prevent the development of extensive two-way conversational exchanges. Though individuals may inject opinions and personal stories into the feed, it's possible that a lack of mutual experience or opinion will stunt direct responses. Additionally, the level of self disclosure that characterizes these threads render contribution a socially high-risk activity. The desire to undertake high risk activity, however, is fueled by the level of trust and commitment within the group - feelings that prove difficult to foster through online environments due to their perceived anonymity and social ambiguity (Passy and Guigini 2001, Wellman 2003, Benkler 2006, Friedkin 2004).

Lets find out...

In order to test this hypothesis, data was codified and mapped from three Postsecret forums, each of a different topic. One discussed recovery and addiction, another spirituality, and another the LGBT community. These maps were then compared to those centered around neutral discussion topics from the online community Usenet. Similarities and differences were then observed between the structures of the two. Due to the confessional nature of the conversations selected, we expect to see a diffuse, centralized structure emerge around the original poster, with single contributions from most users and few ties between peripheral nodes.

Postsecret Data

Secrets of Addiction and recovery: Explaining Scars

This conversational map clearly reflects the confessional-style network we initially predicted. The original poster issued the question, in reference to confessing to cutting habits: "so when you did, if you did, how did you explain yours? and how did whoever you told take it?" Most nodes issued only one response, directed at the group as a whole. The original poster steps in as a moderator and occasionally issues direct replies to other nodes.

Secrets of the LGBT community: National Equality March in Washington DC

The content represented in the data map above was largely a call to action regarding a pivotal event within the LGBT community. The originator of the thread assumed the role of moderator, with frequent directional comments aimed at stimulating and maintaining pertinent conversation. The majority of comments made by peripheral nodes address the network as a whole (i.e event specifics, attendance announcements) or address emerging subtopics.

Secrets of spirituality: Religious minorities have a harder time

This post is less of a confessional post and more of a rant post. It debates the idea that Atheists are at a disadvantage because they have less of a community network. This is less of a “post your own story,” than it is lets discuss ideas of religious communities and the social networks they provide. While many people are responding to the original poster, there is vigorous discussion between other posters as well.

Looking at Neutral discussions: Usenet Data maps

Phish Fans

The variance in network connectivity and interdependence exhibited above, represent the mixed content of the “Phish Fans” network. A network of this type most likely contains both neutral content threads (i.e. concert date announcements) and conversational threads (i.e. opinion based lyric interpretations). The structure of this network is far less rigid than the clearly observable differences between a conversational based network and a neutral content based network.

Tech help

This set of data maps demonstrates the rigidity of neutral content based social networks. The objective focal points of these networks require little conversational interaction between network members. The originator is highly centralized and peripheral nodes have limited to no interconnectivity. Essentially all statements are directed towards the network as whole, not to specific nodes.

Support groups

These images again show the evident differentiation between networks of conversational nature and those networks of neutral based content. The roles of centralized nodes and peripheral nodes are significantly less distinguishable than those seen in the Technical Support forum above.

What can we conclude about the structure of Postsecret groups? Do they differ? How so?

There are several conclusions that we can draw from our analysis:

1) The three threads we studied had a generally centralized network system with many people speaking either to the thread initiator or everyone in general. While we did see some intra-thread conversation going on, it was in the minority.

2) While our hypothesis sought a confessional style forum, only one of the three threads we choose was a confession. The second was a discussion of religious communities and the third was a discussion about a LGBT march. Thus, as our premise was based on the assumption that the PostSecret forums would contain mostly confessional posts, it is difficult to have conclusive findings.

3) To further study the idea that confessional forum user networks would be different from other types of network structures we would recommend, first choosing threads with more than 45 posts, perhaps somewhere in the 60-100 range. Secondly, we would recommend studying more than three threads, though we cannot say how many more would be necessary.

4) The comparison to the Usenet data is also inconclusive; because we did not find more confessional type threads. If we had to draw a comparison between our composite network and the three different Usenet network groups, we would say that our network structure is closest to the Tech Support group, there is a highly centralized network and the peripheral nodes have little contact.

5) To further study the difference between Usenet network structures and forum threads, our group would recommend studying more than three forum threads, this would allow for a better average of the threads in a forum.


  1. I find it very interesting how the support groups are organized and how the nodes are centralized. Compared to the support groups, tech help shows nodes that spread out evenly through out.

  2. Cool project! I think your new network visualizations might be more interpretable if you layed them out again with a different setting. let me know if you have questions.