Rewiring the Spy
This visualization was created in collaboration with Lisa Strausfeld and Pentagram as cover art for the NY Times Magazine. The cover story discussed the impact of social media and open data on the intelligence agencies.
The art needed to be data driven and it needed to be delivered in a high quality print-ready format. We were interested in the correlations between the various components of a terrorist activity: actors, weapons and targets. We didn't have any data, but we did have an idea.
When you make a search on Google, you're usually just concerned with the list or results. However, you're also presented with an estimation of how many matches Google has in its system. For a simple query, the number can be astounding, but we can use this aggregation to our advantage.
"Oklahoma City" returns about 86 million results. "New York City" returns about 358 million. "Timothy McVeigh" returns 439,000. Now, when you look at the combinations, the numbers get more useful:
"Timothy McVeigh" "New York City": 88,000
"Timothy McVeigh" "Oklahoma City": 281,000
There are over three times as many results in combination with Oklahoma City. When you take into account the fact that Oklahoma City also only has about one-quarter the search results as NYC overall, the correlation becomes even more clear. Seems like we're onto something.
Of course, everyone knows Timothy McVeigh is more connected to Oklahoma City than NYC. The point though is to illustrate the idea that by finding correlations in the data you can find possible connections. Correlation is not causation, but it can point you in the right direction for further inquiry.
Once we realized that this approach could highlight connections, we compiled lists of known events and their actors, weapons and targets. We pulled numbers for them individually as well as their combinations.
With the dataset complete we chose a fairly literal representation of the data in a three dimensional point cloud. Each component (actor, weapon, target) is attached to a virtual node. Each node is connected to each other node with a virtual spring. The spring wants that particular pair of nodes to be exactly as far apart as their Google correlation specifies.
Of course, with more than a few nodes, not every spring can be exactly the desired length. The iterative physics algorithm lets each spring push and pull until they find an equilibrium. The specific positions are dependent upon the initial positions of the model, but they tend to fall in arrangements that satisfy the requirements relatively closely.
The interactive tool enables the user to navigate the 3D space, looking at the cloud from any angle. It allows highlighting a given set of nodes and also "plucking" the model, essentialy reshuffling it and letting it resettle into a different steady state.
For print we used the interactive tool, virtually flying around a three-dimensional space filled with bad actors, deadly weapons, and terrorist targets, taking vector-based snapshots when we found interesting and attractive angles. The vector files are compatible with Adobe Illustrator and other industry standard tools enabling the required post-processing and embedding for print.
Think we can help you see your data from a new perpective? Let's talk.