Accuracy + Intuitive Impressions

An accurate depiction of the Blogoshpere (via Data Mining)

An inaccurate representation of the Social Media Network

Just left Data Mining a few minutes ago after having an interesting discussion in the comments area with Matthew Hurst, the author.  Matthew offers a valid criticism of "Visualizing The Social Media Network" in that it's not technically accurate.  From the post:

"The thing that worries me here is that this isn't what social networks (that is to say, the networks found in social media) look like when rendered as a graph. In other words, while this image may describe the fact that there are clusters (communities) and these communities have relationships with each other (via which information is exchanged) it does a poor job of visualizing those communities and the relationships between them. Consequently, I think this is one of the cases where the image has a negative impact and where some simple words would provide a more accurate description."

And Matthew is probably correct.  The visual may not be completely accurate as it's drawn from my own personal perspective as I was learning how the social network operated.  And personal perspective tends to distort as we all know.  However, it was helpful to me to visualize this for myself—and many of you picked this graphic up as well.

So, here is the "but".  But, even though my visual is technically inaccurate (I guess), does that strip the value away from it?  Matthew says that simple words would have provided a more accurate description.  I guess I could have just written about my experience...  But then Matthew went on to describe what I did here as an "intuitive impression".  Man, I love that term.

So please note that moving forward any visuals you come across here are to be taken as intuitive impressions vs. accurate depictions.  That's my new disclaimer.  Personally, I think my visual is a nice Yin to Matthew's data-driven Yang.  But that's just me.  :)

Comments

[Paul McEnany]

Honestly, I don't think Data Mining's depiction tells any story. It looks like just one big ball with some spikes, which is not really an accurate depiction either.

I could be wrong here, but I would guess that the closer in to the ball you got, the more it would start looking like your depiction, up to a point, anyway.

Of course, that is my intuitive impression, made up mostly of bullshit and guessing. :)

[DA]

LOL!

Paul, Bullshit is VERY intuitive.

[Matthew Hurst]

Dan, I appreciate your continuing with this discussion.

Paul, as for the graph telling a story, it is probably not the best choice wrt an analogy to what Dan has been talking about. Try this post for a more related view:

http://datamining.typepad.com/data_mining/2006/07/mapping_categor.html

As for the structure of individual communities:

http://datamining.typepad.com/data_mining/2006/09/tribes_of_the_b.html

Arguing that the big ball picture is not accurate - it is 100% driven by data. And in this case, the data is a very comprehensive set taken from a blog search engine's index. Given that it is data driven, I'm wondering where the inaccuracy comes in.

[DA]

Matthew, thanks for the additional links, I'll check them out. I think Paul's point was not that your data-driven map is inaccurate, but questions if it tells a story.

And Data isn't supposed to tell a story, at least not in a narrative fashion.

speaking of accuracy, I think you might be calling me Dan? It's David.

[Paul McEnany]

Thanks, DA. You took the words out of my mouth.

The thing I've learned most about data is that it, too, is as much a product of art as science. The more data there is, the easier it is to tell the story you want it to tell.

I'm not saying that your data is inaccurate, Matt, but data analysts have to rely on their gut sometimes, too.

[Roger von Oech]

Interesting post. Here's my two cents. When I look at these pictures, I just think "caricature." Like those of Bush (big ears) or Clinton (big nose) or Carter (big lips), caricatures emphasize certain information and ignore other info. They help me recognize the basic relationships (very helpful), but I don't take them literally. Thus, they're a form of visual metaphor.
http://blog.creativethink.com/

[Christopher Fahey]

I agree with those above who question the value of the "accurate" data-driven model from Data Mining. Just because it is based on accurate data doesn't mean that it communicates that data accurately. The design of the algorithm that transforms the raw numbers into a graphic that communicates meaning and insight to the viewer is the "art" Paul is talking about, and I think I have to agree that the Data Mining model lacks that "art" -- it really doesn't communicate much of anything at all -- and in fact, contrary to Matthew's speculation about David's diagram I think it is Matthew's diagram that risks miscommunication (for example, it looks like it's saying that all blogs are closely connected to all other ones except for the 25 or so around the edges). es David's diagram, while clearly pulled out of his ass, communicates one small thing very clearly.

A lot of data-based visualization has a tendency I think to be really cool to look at but largely ineffectual in communication because really good communicative algorithms are so hard to make, especially. Sometimes a human designer is better because they can adapt the design of the graph to the message being communicated.

Of course, the absolute best approach is for a human designer to interpret real data.

[DA]

"David's diagram, while clearly pulled out of his ass, communicates one small thing very clearly."

Christopher—you've given away my secret. IE, my ass.

;)