Street Art or Stream Art? Following #Banksy on the streets of NYC

What does the famous street artist Banksy has to tell us about the relation between physical places and their social media representations?

Radial visualization of 16,164 Instagram photos geo-tagged to NYC area between October 1 and November 20,2013. The photos are organized by location (perimeter) and upload date and time (angle).
Radial visualization of 16,164 Instagram photos geo-tagged to NYC area between October 1 and November 20,2013. The photos are organized by location (perimeter) and upload date and time (angle).

In this paper we define hyper-locality on social media by looking at the relations between physical places and their social media representations. To do this, we use a particular case study – social media photos tagged and shared on Instagram during the anonymous British graffiti street artist Banksy’s month-long residency in New York, October 2013.

Banksy installed a new work nearly every day in different locations in the city, and announced them by posting photos of the works on Instagram and on his personal dedicated website (banksyny.com). He then asked his followers to post their own photos of the works with the hashtag #banksyny. In many cases, the only way to detect the location of the works was to search for their earlier representations online, tagged to #banksyny.

Joining Banksy’s artistic experiment, thousands of people flocked around the city in an effort to catch a glimpse of the works and photograph them before they disappeared, defaced or painted over. At the same time, the social media outcome of this experiment was a unique dataset of photos and metadata that exemplifies various spatial, temporal and content patterns in each of the locations.

Using computer vision techniques we clustered the dataset into groups of photos that document each of the artworks. We then used these clusters to analyze and visualize the dataset in various ways. For example, by plotting the clusters on a map we showed the difference in geographical spread of each of these specific works:

A map of locations of all photos of 7 clusters (Only NYC area is shown). Each cluster is colored in order to represent the spread of photos of the same artwork.
A map of locations of all photos of 7 clusters (Only NYC area is shown). Each cluster is colored in order to represent the spread of photos of the same artwork.

In another example, we compared the visual attributes of Banksy’s photo to those of his followers to measure the distance between the “original” image and its “copies”.

A matrix image plot visualization of 6 clusters. In each cluster, (X) - brightness mean, (Y) - hue mean. A red square represents the original photo of an artwork posted to Instagram by Banksy himself.
A matrix image plot visualization of 6 clusters. In each cluster, (X) – brightness mean, (Y) – hue mean. A red square represents the original photo of an artwork posted to Instagram by Banksy himself.

Our paper combines quantitative and qualitative methods, and employs for the first time perspectives from the fields of Digital Humanities and Art History in social media research. Based on our theoretical and historical analysis of Banksy’s case study, we offer to characterize hyper-locality on social media as: fragmented, temporalized, and nomadic.

Want to learn more?
Check out our ICWSM presentation (On Hyper-locality: Performances of Place in Social Media) or contact us at: nah61@pitt.edu

Nadav Hochman, History of Art and Architecture, University of Pittsburgh
Lev Manovich, Computer Science, The Graduate Center, CUNY
Mehrdad Yazdani, California Institute for Telecommunication and Information

About the author

Nadav Hochman

Nadav Hochman is a doctoral student in the History of Art and Architecture department at the University of Pittsburgh, and a visiting scholar at the Software Studies Initiative (The Graduate Center, City University of New York). Most recently Nadav was a visiting researcher at the MoMA NYC.

His research intertwines media theory and computational methods for the analysis of large online visual cultural data sets. Using data visualization techniques from a digital humanities perspective, he examines how and what can we learn about local and global cultural patterns and trends by aggregating large amounts of user-generated visual materials.

His work was featured in various media outlets such as Wired, Guardian, Fast Company and The Atlantic.

View all posts