Spatio-temporal mapping of street art using Instagram
First Monday

Spatio-temporal mapping of street art using Instagram by Christopher D.F. Honig and Lachlan MacDowall



Abstract
By scraping metadata from Instagram images tagged with #MelbourneStreetArt, we are able to create geographical and temporal maps of street art in Melbourne, mediated through the collective eye of Instagram. Apart from merely identifying the most popular images, geo-tagged metadata allows us to create spatial heat maps to identify physical locations of high-image production. Caption data beneath the images allows us to search for high frequency words, which we use to identify patterns within the online audience’s relationship to street art. Finally by simply plotting out the number of images produced each day and cross-referencing with the corresponding caption data, we are able to identify historically significant events within Melbourne’s street art culture. The analysis is easy to use, even for a researcher with minimal programming experience and can be used to project cultural trends (beneath any hashtag) or as a tool to navigate historical Big Data within a conservation context.

Contents

Introduction
Spatio-temporal mapping of image density
Structural linguistic analysis
Time sequence of data stream
Conclusion

 


 

Introduction

Although many possible historical narratives may be overlaid onto street art, a common origin point traces its roots to contemporary graffiti culture, in New York in the late 1970s (Deitch, et al., 2011). New York style subway graffiti paralleled the rise of hip-hop culture and in 1983, a series of subway artists transitioned into a gallery context, spawning the short-lived ‘post-graffiti’ movement (Thompson, 2009). Simultaneously artists including Jean-Michel Basquiat and Keith Haring (who had traditionally been connected to graffiti culture) achieved commercial success in fine art circles and are now highly regarded within the Western art cannon (Deitch, et al., 2011).

In the late 1990s, a new cultural practice again emerged from graffiti culture, now widely referred to as ‘street art’ (McCormick, et al., 2010). Where graffiti traditionally invoked ornamentalized letter forms, street art primarily used image-based works with central motifs drawn from mainstream cultural iconography (for example celebrity faces and retro gaming aesthetics) (CDH, 2013). Street art also departed from graffiti in its diversity of forms and materials (such as stencils and wheat paste posters) and new political motivations (for example, in response to the 2003 U.S. invasion of Iraq).

In 2006–2007, the cultural movement achieved major commercial and mainstream recognition after the record sales of works by the U.K. artist Banksy (Schacter, 2016; Young, 2016). In 2008, Banksy sold his work Keep it spotless for US$1.87 million, the highest price to date [1]. Street artist Shepherd Fairey developed the Obama ‘Hope’ poster for 2008 U.S. Presidential race and Banksy’s 2010 film Exit through the gift shop was nominated for Academy Award for best documentary feature [2].

Although beginning as a counter-cultural practice, the widespread popularity of street art meant it was soon co-opted into mainstream commercial culture (CDH, 2013; Young, 2016). The cultural practice has become associated with gentrification (Schacter, 2015) and new forms of advertising (hand-painted mural ads, youth-oriented advertising aesthetics). However the diversity of practitioners and motivations, the shifting political landscapes and the ever-changing urban context mean there is no singular meaning that may be ascribed to street art. Its inherent cultural complexity, its spatial specificity within the city, its broad and diverse audience and its organic growth outside of established art institutions, all make street art an interesting cultural phenomenon to study.

There are a number of conditions that encourage the documentation and sharing of street art in social media platforms: its broad popularity, its ephemerality within the public domain and its ocular-centrism. (MacDowall, 2008; Honig and MacDowall, 2016). In particular, the social media platform Instagram (a free social networking service launched in October 2010 and primarily designed for mobile devices) has proved a popular medium for street art. Beyond simply functioning as a container for the artworks, it has been argued that Instagram has actively shaped the production of graffiti and street art and arguably accelerated and amplified its reach (Avramidis and Tsilimpounidi, 2017).

As a tool for mapping cultural taste, repurposed data from Instagram has been particularly useful to academic researchers (Hochman and Manovich, 2013; Silva, et al., 2013; Li, et al., 2015; Shelton, et al., 2015). It offers both large numbers of users and low levels of user privacy, which make it easy to scrape large, detailed datasets. Both the architecture, ‘game-play’ and algorithmic environment of Instagram also reward users for frequent and prolific hash-tagging of their images.

However, the very technology that enables these forms of cultural research also produces a shift in the spatial significance of public-space practices such as street art. That is, at the very moment when we can access large amounts of data about the physical location of street art works chosen by audiences, the physical location is becoming less significant in the meaning of the work, as the practices enabled by Instagram detach the works from the physical location and often, their immediate spatial context (Avramidis and Tsilimpounidi, 2017).

Although it offers an exciting new research methodology, there are also a range of ongoing practical limitations with the use of big data (boyd and Crawford, 2012) and Instagram in particular as a research archive (Highfield and Leaver, 2015): 1) Social-media data should be understood as simulacra; it does not directly correlate to the physical world it depicts. The data represents a convolution of users’ motivations for posting, artefacts of the social media platform itself as well as the subjective ‘real world’; 2) The data has been removed from the original context, which can create problems of interpretation (for example language algorithms struggle to detect sarcasm); 3) Claims to the objectivity of big data should not be overstated; apart from the biases introduced by the researchers themselves (choosing what questions to ask, how to interpret the information) data itself may be skewed (how participants have self-selected into the datasets, what functionality the platform allows users to engage in). In these contexts, more data is not always better data. There will always be a place for traditional ethnographic study and direct interactions with a culture. So big data offers a new tool, but not a substitute for traditional methods of inquiry.

Beyond technical challenges for big data, a mounting concern among researchers has become academic ethics in a rapidly expanding field (Richards and King, 2014). Perhaps the chief concern among a range of questions about academic ethical responsibilities, relate to the data privacy of social media users; although the data may be ‘public’ this does not equate to permission of use in new contexts. For example, a user may make information ‘public’ on social media without the expectation of the data being sold, used in advertising or used in academic research. Although a user may have ‘no problem’ sharing the information, academic researchers must still be accountable to the terms of when, how and why the individual shared the information (boyd and Crawford, 2012). Such concerns have led some researchers to rethink ‘privacy’ as an ‘accountability of presence’; rather than considering privacy as a matter of revealing/hiding specific information, it becomes more productive to consider privacy as relational, complex and shifting through a series of social accountabilities (Troshynski, et al., 2008). It is also worth noting that a specific suite of concerns have arisen over the use of location-based metadata privacy and its capacity for use in surveillance (Troshynski, et al., 2008), for example to enforce spatial restrictions on recently released prisoners.

As Foucault (1977) has argued, the omni-present threat of surveillance renders the use of direct power or violence unnecessary; the perception of panoptic surveillance creates a compliant, disciplined subject or docile body. User’s social media metadata may be technologically operationalized for surveillance, but big data academic researchers may also indirectly contribute to the perception of surveillance, which may alter users’ behaviour. For example, how may online political discourse be affected if users perceive that their online data can be mapped (and retrospectively so)?

New research fields inevitably raise new ethical boundaries. At the intersection of street art and big data, significant controversy arose when researchers used the location of artworks by street artist Banksy, to try to reveal Banksy’s identity (Hauge, et al., 2016). The paper was widely criticised in academic circles: for being empirically unsound, for creating a sense of mistrust towards academic researchers in communities involved in illicit practices (which may inhibit academic fieldworkers) and for identifying the name of Banksy only to garner broader mainstream publicity (Bengtsen, 2016). The work was also used as a case study in rethinking the ethical regulation of big data research as an extension of social sciences research (Metcalf and Crawford, 2016). In defence of their work, the authors argued that their research proposal had passed through independent ethics approval at their host institution (Bengtsen, 2016). Rather than resolving the criticisms, this perhaps only raises new questions. In our current work and previous research (Honig and MacDowall, 2016) our proposed research methodology also passed through an ethics approval process and was rated in the lowest category of risk (so no formal ethics process was required, beyond informal discussions with ethics advisers). Rather than merely allowing the research to proceed unquestioned, this perhaps highlights the need for ethics approval procedures to be modified to accommodate the new ethical boundaries raised by big data research in social sciences.

A previous study used Instagram to investigate the construction of genres of ‘street art’ and ‘graffiti bombing’ based on audience preferences and follower correlations between accounts (Honig and MacDowall, 2016). In essence this previous study tracked users (via @ tags). The current paper seeks to track images (via # tags) to infer new insights about Melbourne’s street art culture. The specific objectives of the paper are: 1) To map the spatio-temporal growth of street art in specific physical spaces in Melbourne by tracking geo-tagged metadata (thus viewing street art growth mediated through the lens of Instagram); 2) To perform a structural linguistics analysis to search for commonly recurring terms within caption data (what information do street art Instagram photographers deem important to specify in caption data and what does this convey about the online street art documentation culture?); and 3) to simply track the rate of image production as a function of time to identify major events within Melbourne’s street art history (as a tool for navigating historical archives). The paper also offers a range of other more minor insights, such as which images are most popular and what may be anecdotally inferred from these images.

The analysis could be completed in any city in the world and for any hashtag, but we select ‘Melbourne’ and ‘street art’ simply because we are based here and are more familiar with the cultural history of the local street art scene.

An important ethical distinction to emphasise in the current work is that we are primarily studying locations (rather than users or artists); we primarily look how many images are generated at a particular location, rather than what images are being generated, why or by whom. Users’ data is anonymized into large maps. So the research has a particularly low ethical risk.

The current paper follows the procedure of Honig and Macdowal (2016) and so data is accessed from Instagram using Magimetrics, a third-party Web crawler. Magimetrics has an API agreement with Instagram and conforms to their terms of use. We collect information from publically available accounts only, under a given hashtag. The data can then be processed in simple packages such as Microsoft Excel or Python. Here the data processing does not require a sophisticated mathematical operation, only reorganisation into more convenient outputs. Our geographic heat maps of image density are generated in the online software eSpatial.

Hash-tagging on Instagram follows a complex grammar. Images of street art from Melbourne are posted on Instagram under a variety of hashtags: under a single hashtag (#melbournestreets, #melbstreetart #streetartmelbourne, #melbournegraffiti, #melbgraf), in a series of separate hashtags (#streetart #Melbourne) or, where the content is deemed to be implicit, simply the general location of the artwork.

However, this hashtagging grammar is shaped by the Instagram platform. At various points the owners of the Instagram platform decided to disable the use of certain hashtags, either because they were too generic (#photograph) or made reference to explicit content that is banned under Instagram’s terms of use. Later versions of the Instagram app detect common hashtags (and those previously used by the account), adding to a hashtag’s connectivity but also helping shape a consolidation in the dominant tags for any subject.

The dominant tag for street art in Melboourne is #MelbourneStreetArt with over 130K posts (compared to 45K for #SydneyStreetArt) and comparable to English language hashtags in other major cities that are well known for street art such as #NYCStreetArt (182K), #LondonStreetArt (180K), #ParisStreetArt (100K) and #BerlinStreetArt (67K).

This study collected the data for the hashtag #MelbourneStreetArt from Instagram, returning over 79K publically available images (as of September 2015). We are primarily interested in the image metadata and so we collect the following information: the date an image was posted, the complete caption, the user name and ID who posted it, the number of likes and comments (but not comment text), the file type (image or video) as well as geotagged coordinates and location name (if provided by the user). The data can be sorted in a variety of ways to trace the growth of street art and user interaction with street art mediated through Instagram. For example, a simple analysis is to search for the most ‘liked’ image within the data for a particular hashtag. The image with the most likes will reflect a convolution of both the popularity of the image itself and the number of followers of the original poster at the time of posting (who is able to reach a wider audience).

In the case of the hashtag #MelbourneStreetArt the most ‘liked’ image was posted by @DigitalDoes, a graffiti writer originally from the Netherlands who visited Melbourne in 2013 (Figure 1). @DigitalDoes has over 99K followers (as of June 2016).

Figure 1 itself offers a number of immediate insights about street art culture. Firstly, it is a photograph of a page from the book Street art now, a catalogue of images of Melbourne street art from 2012–2014 collated by Instagram user @DeanSunshine. The book contains minimal text and instead primarily functions as a print equivalent to an Instagram feed. @DigitalDoes’ inclusion of the image in his own Instagram feed emphasises a value ascribed to the physical image over the digital image. A screen shot of the original art work that appears in @DeanSunshine’s curated Instagram feed is not shared, but the photograph of the curated book is worthy of sharing (and indeed garners a huge number of likes). The physical image is seen to convey a higher status than the digital image.

 

The most liked Instagram image within the hashtag #MelbourneStreetArt posted by @DigitalDoes
 
Figure 1: The most ‘liked’ Instagram image within the hashtag #MelbourneStreetArt (as of January 2016) posted by @DigitalDoes.
Note: Larger version of figure available here.

 

Figure 1 also emphasises the recirculation of images within the online space. The original mural has been photographed by @DeanSunshine, printed and bound in a book, photographed by @DigitalDoes and shared on Instagram and now screen shot and shared in this paper. The original mural has three additional tiers of contextual reproduction beyond the original physical mural (please feel free to print it out, photograph it, put it on Instagram and tag @StreetArtResearchBot to go truly meta). This process of reappropriating simulacrum highlights a digital cut and mix culture reminiscent of the physical assemblage and appropriation typical in street art DIY methodology.

Finally @DigitalDoes’ inclusion of the image and the text within the caption function as an advertisement of the book Street art now, to his Instagram audience. This mirrors discourses around the commodification and recuperation of street art (CDH, 2013; Schacter, 2015); the most ‘liked’ image of #MelbourneStreetArt is in fact not an image of street art, but an advertisement for a book.

Who produces the data? We strip the metadata from 79K images on Instagram with the hashtag #MelbourneStreetArt. But what does this data represent and whose perspective is being surveyed? Not all users post an equal number of images and so although the data stream has been produced by 6.2K unique users, more than half of these users posted only one image with the #MelbourneStreetArt tag (3.4K users, 54.8 percent of users). At the other end, the most prolific user (@MelJewell) alone created 3.2K images. The top 10 most active accounts produce almost a quarter of the content (18.6K images, 23.5 percent of images) and 133 accounts (2.2 percent of all users) create over two thirds of the total content (68.3 percent of images). This finding is consistent with previous academic work which report that small percentages of users in online communities typically generate the majority of the content (Lampe, et al., 2010). So for the purposes of the current analysis, it is more reasonable to consider this data to be the collated opinion of approximately 200 people, who have all dedicated hundreds of hours to documenting and collating images of Melbourne street art, rather than the equally weighted opinion of 6.2K distinct users.

The role of online documenters (sometimes called ‘clickers’) is now well established in street art culture. As the distribution in Table 1 suggests, clickers can range from those informally sharing a small number of images to more sustained and in-depth practices of documentation. Major documenters often collaborate with artists and may also produce blogs or books which exhibit street artists to wider audiences. Far from simply reproducing standard images of street art, many documenters develop photographic strategies to represent and stage the work in context, or chart the changing nature of key sites (Avramidis and Tsilimpounidi, 2017). The work of these sustained and specialist documenters are over-represented in our sample (Table 1).

 

Table 1: Breakdown of images created by users under the tag #MelbourneStreetArt. For example, 2,396 users created between one to 20 images each, which represents a total of 11,204 total images (14.1 percent of all images created beneath the hashtag).
Images shared by single userTotal users>Total images
NumberPercentageNumberPercentage
13,37454.83,3744.3
2–202,39638.911,20414.1
21–1002444.110,57513.3
101–1,0001232.035,42844.7
≥1,001100.218,65223.5
Total6,15710079,233100

 

 

++++++++++

Spatio-temporal mapping of image density

Within the set of images tagged #MelbourneStreetArt, 14.4 percent are ascribed a ‘public’ geotag (12.8K images). We break this list of images into annual blocks, beginning in 2011 (237 images) and ending in 2015 (5,659 images). The growth of Instagram (and the increasing popularity of street art) means that each block of time does not contain the same number of images. We plot the spatial distribution of images each year overlaid with a map of Melbourne (Figure 2). The maps should be seen as a combination of both the growth of street art and the growth of Instagram and serves as a tool to investigate where and when the Instagram audience is encountering street art (and choosing to post it online).

 

Image density heat map of Instagram media tagged with #MelbourneStreetArt, 2011
Note: Larger version of figure available here.
 
Image density heat map of Instagram media tagged with #MelbourneStreetArt, 2012
Note: Larger version of figure available here.
 
Image density heat map of Instagram media tagged with #MelbourneStreetArt, 2013
Note: Larger version of figure available here.
 
Image density heat map of Instagram media tagged with #MelbourneStreetArt, 2014
Note: Larger version of figure available here.
 
Image density heat map of Instagram media tagged with #MelbourneStreetArt, 2015
Note: Larger version of figure available here.
 
Figure 2: Image density heat map of Instagram media tagged with #MelbourneStreetArt by year from 2011. Some of the hot spots have been identified and labelled.

 

We note that prior to 2012, some of the geotag labels have been given improper coordinates. For example, images tagged as the suburb ‘Fitzroy’ are actually ascribed coordinates in a neighbouring suburb of Parkville in plots from 2011 and 2012.

The highest density of images consistently comes from Hosier Lane in Melbourne’s CBD. The laneway was formerly the location of street artist’s studios and Andy MacDonald’s Until Never gallery and the Citylights project (back-lit signal boxes displaying artworks). Over the last decade it has become a landmark site for street art in Melbourne (Honig, 2017), attracting thousands of tourists, featuring prominently in city branding and supported by the City of Melbourne and several tour companies, who guide paid walks through the space (other local sites including Union Lane, Duckboard Place and Blender Lane).

Our analysis presents a combination of street art at the physical site and the Instagram users at the site (to photograph and share the works). So the high density of images generated within Hosier Lane may be attributed to both its proximity to the Melbourne CBD (with a correspondingly high daily foot traffic) and the quantity (and turnover) of street art at the site.

Bourdieu (1984) has argued that cultural taste correlates highly with an individual’s level of education and then social class. It has further been argued that the rising creative class (Florida, 2004) may have a cultural preference for street art (Schacter, 2015) and as young knowledge-based professionals, they may also have a higher uptake of emergent social media platforms, such as Instagram. So the growth of street art images on Instagram in a particular location may be useful as an indirect metric of gentrification pressure from the creative class in that location; increased street art and the increased online documentation of street art may suggest higher populations of creative class professionals within the local community. The analysis could be performed with a range of alternative hashtags that might suggest creative class professionals.

Beyond the Melbourne CBD, we find an established hot spot in the suburbs of Fitzroy and developing hot spots in Brunswick and Footscray. All three of these suburbs are located in the inner city and since the 1990s have undergone shifts away from traditionally working-class suburbs with local manufacturing bases, towards start-up creative industries and newly renovated inner-city living spaces.

A final observation is that a number of localized hot spots using the tag #MelbourneStreetArt are commercial establishments, such as cafes and bars. Again, this provides supportive evidence for the broader claim of a relationship between street art and gentrification, in which street art may be used to draw higher income creative professionals to existing commercial sites.

 

++++++++++

Structural linguistic analysis

Beyond the metadata from the images, we can also scrape the captions attached to each image. The average image caption in our data set is just 14.2 words long, but this gives a total of 1.14 million words to process (we identify words by searching for letters separated by a space). Among these total words, 74.7K are unique. For practical reasons, all emojis are excluded from our data but we note that this is an increasingly important form of online discourse; an internal study conducted by Instagram in 2015 suggested that emojis are included in over 40 percent of user generated text (Dimson, 2015; Jackson, 2015).

The first way we process the data is to simply order the unique words by their frequency. The 10 most common words within the image captions beneath the #MelbourneStreetArt tag are (in order of frequency): #streetart, #Melbourne, #graffiti, the, #art, #MelbourneGraffiti, #urbanart, #streetartmelbourne, in, #igersmelbourne. Two of these words are common prepositions, the other eight are all hashtags used within the caption.

Processing the data in this way treats the language within a structuralist framework. It considers all the words as having singular, literal and definite meaning and removes the word from its grammatical context. For example, the word ‘art’ can be used literally (a canvas painting in a gallery), ironically (a pile of garbage as a criticism of contemporary art) or as a proper noun (Art Garfunkel). Our analysis does not detect contextual meaning as the words are removed from their original syntax. We have not considered bigrams or higher order combinations of words. So for example the words ‘happy’ and ‘not happy’ denote opposite meaning, but our analysis would only search for the frequency of the word ‘happy’.

It is also worth noting that generalist words appear more frequently. For example, it is highly common for photographers to identify the artist’s name in the image caption, but this splits the ‘artist’ signifier across hundreds of separate words (all the different artist names). By comparison, the word ‘mural’ has only a few variant synonyms (piece, painting, wall art etc.) and so may appear much higher in the frequency list than any individual artist, but not necessarily higher than the total number of captions in which an artist is identified.

The descriptive signifiers selected by the users may reflect a range of aesthetic and cultural motivations: Users may select tags to cross-reference the image for other users to find and follow; language can be used to signal belonging within an online culture where users may never meet; the selected terms may articulate the key frames of reference to the medium (in essence, what features are considered most important to the users). Similar to the images within a user’s feed, comments can be attached to cultivate an online persona, for example in the common Instagram practice of attached sentence-long tags which will likely be unique and have no connective value (#thisisalongsentenceasasinglehashtag).

Captions may also be used to provide a description of the image, a context to the image, a taxonomical accounting of the image’s genre or a range of other types of information. Within the captions, we have identified a number of recurring thematics:

  1. Location. Captions rarely provide the specific address of an artwork (for others to find in physical space) but do commonly reference the community, suburb and city the work is located in. Common explicit references to Melbourne include: #Melbourne (2nd most frequent word), #BurnCity (17th) [3], #Melburn (22nd), #StreetsOfMelbourne (24th) and Melbourne (52nd). Inner city suburbs known for street art are also highly common in the caption data: #Fitzroy (27th most frequent word), #Brunswick (60th) and #Collingwood (64th). In essence the users regard spatial context as important to describe the works, ahead of other contextual signifiers such as surface the work is painted on, date the work was painted or biographical information about the artist.

    Although references to suburb and city are highly common, there are comparatively few references to national identity: #Australia (67th most frequent word), #wow_australia (143rd), #ig_australia (158th), #StreetArtAustralia (199th) and #OzStreetArt (853rd). The emphasis on city and suburb ahead of nation implies a tight suburban regionalism within street art culture that prioritizes identification and an affinity with the immediate community ahead of a national identity. So the users associate ‘street art’ with ‘inner city Melbourne’ or ‘my immediate community’ as opposed to the broader signifier of ‘Australia’ or other geographical categories.

    Users sometimes provide an immediate spatial context to the image: #Urban (47th most frequent word), #laneways (76) and #laneway (115). This can help establish the staging of the work in the mind of the audience and help to convey the imagined experience of encountering the work in physical space (Young, 2014).

  2. Formal material qualities. Common recurring words in the caption data include descriptions of the tools and materiality of the artwork. For example: #paint (40th most frequent word), #AerosolArt (41st), #Pasteup (45th), #SprayArt (50th), #Mural (53rd), #SprayPaint (55th), #Aerosol (59th), #Stencil (89th). The medium may already be shown in the image, but articulating it explicitly allows for a cross-reference to other works, it can display the cultural literacy of the photographer and it can function instructively to followers new to street art.

  3. Self-reference. Users often make explicit reference to Instagram and the imagined online community. Direct references include: #IGersMelbourne (10th most frequent word), #InstaGraffiti (26th), #InstaGrafite (37th), #InstaMelbourne (39th), #InstaGraff (49th), #InstagramHub (51st), #IGDaily (54th), #InstaDaily (62nd), #GraffitiIGers (67th), #InstaGood (92nd), #IGers (105th). In online communities, where traditional social cues and shared artefacts may be missing, language is often used to communicate a sense of oneness (Fayard and DeSanctis, 2010). Here we find this in the construction and maintenance of internal references to these online sub-cultural communities. For example the tag ‘IGersMelbourne’ creates a shared intersection of interest amongst a community of Melbournians and Instagram users.

    Self-reference can also include references to photography or techniques used to capture the image. These can include: #iPhoneOnly (30th most frequent word), #PhotoOfTheDay (35th), #DopeShotBro (57th), #StreetPhotography (61st), #PicOfTheDay (90th), #All_WallShots (91st), #StreetArtPhotography (100th), #NoFilter (291st). Some of these tags provide a context to the image (the use of image enhancement or not) but they also articulate the personal stories of the photographer. This language reminds us that the photographer is as much a cultural producer as the mural painter.

Another way to process the data is to look at specific pairs of antonyms. For example the word ‘beautiful’ appears 1,164 times where ‘ugly’ appears only 14. The comparison implies that as an historical archive for recording Melbourne street art, user generated records on Instagram appear to significantly favour works that the users describe as visually pleasing. The ephemerality of street art means that confronting or visually displeasing works may over time be diminished or removed from the user-generated online archive.

The words ‘girl’, ‘#girl’, ‘woman’ and ‘#woman’ appear a total of 1,053 times. By comparison the words ‘boy’, ‘#boy’, ‘man’ and ‘#man’ appear just 599 times. The comparatively high frequency of female signifiers in online image captions may reflect both the commonality of the ‘pretty girl’ motif in street art muralism (CDH, 2013) as well as the selection preferences of the audience of photographers and the broader asymmetrical ascription of gender to women rather than men.

 

++++++++++

Time sequence of data stream

A final and simple way to process the image data is simply to plot the number of images produced each day as a continuous data stream. Spikes in the rate of image production can often indicate a major event (Petrović, et al., 2010). For example in the figure below, peaks in image production represent major street art events such as the Patterson Project (a multi-storey building scheduled for demolition, painted by artists and opened to the public) and All Your Walls (a council-sponsored painting event at Hosier Lane, a focal point for Melbourne street art). There is also a spike to coincide with the exhibition of Melbourne street artist Be Free. However the data does not only highlight major art events; it is also subject to static error from the peculiarities of individual users. For example, the largest spike in image production came on a day when a single user uploaded 146 images. So our data size is not large enough to entirely suppress the background static caused by the idiosyncrasies of individual users.

 

Volume of images produced per day, tagged with #MelbourneStreetArt since the launch of Instagram
 
Figure 3: Volume of images produced per day, tagged with #MelbourneStreetArt since the launch of Instagram. Peaks in the data may be used to identify major cultural events. The inset shows a fast Fourier transform of the time signal.
Note: Larger version of figure available here.

 

The inset in Figure 3 shows a fast Fourier transform of the time sequence. For readers unfamiliar with the mathematics of signal processing, a Fourier transform converts the signal from a time domain to a frequency domain. Spikes in the Fourier transform represent repeating events. There are minor peaks at the periodicity of weekly and bi-weekly (although we note these are barely above the noise). The peaks imply that the community of Instagram street art photographers may regularly recycle through the city on a weekly or bi-weekly schedule, looking for new works. The most common day to post images is a Sunday (16.4 percent of images) and the least common is a Friday (13.3 percent of images).

 

++++++++++

Conclusion

We have used metadata from Instagram images tagged with #MelbourneStreetArt for spatial and temporal mapping of street art image production. Apart from simply searching for the most liked or most commented image, we have been able to create spatial heat maps of image production, to identify locations where street art is most photographed. By removing words from their original syntax, we have identified a number of trends in user descriptions. Finally we plot the number of images produced per day to identify historically significant events in Melbourne street art culture. The analysis uses established data packages and so it is easy to use, even for a researcher with minimal programming experience. The analysis presented here may be used to project cultural trends (with any hashtag) or as a tool to navigate big data in an historical context, for example as a user generated online documentation of cultural movements. End of article

 

About the authors

Christopher D.F. Honig is a practicing street artist and a lecturer in the Department of Chemical and Biomolecular Engineering in the Faculty of Engineering at the University of Melbourne.
Corresponding author: christopher [dot] honig [at] unimelb [dot] edu [dot] au

Lachlan MacDowall is an artist and research associate at the Research Unit for Public Culture in the Faculty of Arts at the University of Melbourne.

 

Notes

1. See, for example, James Tarmy, 2015. “You can own a Banksy for only $5,317,” Bloomberg (28 January), at https://www.bloomberg.com/news/articles/2015-01-28/making-banksy-collectors-cash-in-at-bonham-s-with-street-art-sale, accessed 10 February 2017.

2. https://en.wikipedia.org/wiki/Exit_Through_the_Gift_Shop, accessed 10 February 2017.

3. ‘Burn’ means to produce high-quality paintings in graffiti culture and has a phonetic similarity to the second syllable in Melbourne in the Australian vernacular.

 

References

K. Avramidis and M. Tsilimpounidi (editors), 2017. Graffiti and street art: Reading, writing and representing the city. London: Routledge.

P. Bengtsen, 2016. “Hijacking Banksy: Using a contemporary art mystery to increase academic readership,” Street Art & Urban Creativity Scientific Journal, volume 2, number 1, pp. 60–66, and at http://www.urbancreativity.org/uploads/1/0/7/2/10727553/p.bengtsen_sauc_vol2_n1_r3.pdf, accessed 11 February 2017.

P. Bourdieu, 1984. Distinction: A social critique of the judgment of taste. Translated by R. Nice. Cambridge, Mass.: Harvard University Press.

d. boyd and K. Crawford, 2012. “Critical questions for big data: Provocations for a cultural, technological, and scholarly phenomenon,” Information, Communication & Society, volume 15, number 5, pp. 662–679.
doi: http://dx.doi.org/10.1080/1369118X.2012.678878, accessed 11 February 2017.

CDH, 2013. “Notes on the commodification of street art,” Art Monthly Australia, number 263, pp. 42–44; version at http://www.cdh-art.com, accessed 11 February 2017.

J. Deitch with R. Gastman and A. Rose (organisers), 2011. Art in the streets. New York: Skira Rizzoli.

T. Dimson, 2015. “Emojineering part 1: Machine learning for emoji trends,” Instagram Engineering Blog (30 April), at https://engineering.instagram.com/emojineering-part-1-machine-learning-for-emoji-trendsmachine-learning-for-emoji-trends-7f5f9cb979ad#.yhnxljn1c, accessed 11 February 2017.

A.-L. Fayard and G. DeSanctis, 2010. “Enacting language games: The development of a sense of ‘we–ness’ in online forums,” Information Systems Journal, volume 20, number 4, pp. 383–416.
doi: http://dx.doi.org/10.1111/j.1365-2575.2009.00335.x, accessed 11 February 2017.

R. Florida, 2004. The rise of the creative class and how it’s transforming work, leisure, community and everyday life. New York: Basic Books.

M. Foucault, 1977. Discipline and punish: The birth of the prison. Translated by A. Sheridan. New York: Pantheon Books.

M.V. Hauge, M.D. Stevenson, D.K. Rossmo and S.C. Le Comber, 2016. “Tagging Banksy: Using geographic profiling to investigate a modern art mystery,” Journal of Spatial Science, volume 61, number 1, pp. 185–190.
doi: http://dx.doi.org/10.1080/14498596.2016.1138246, accessed 11 February 2017.

T. Highfield and T. Leaver, 2015. “A methodology for mapping Instagram hashtags,” First Monday, volume 20, number 1, at http://firstmonday.org/article/view/5563/4195, accessed 11 February 2017.
doi: http://dx.doi.org/10.5210/fm.v20i1.5563, accessed 11 February 2017.

N. Hochman and L. Manovich, 2013. “Zooming into an Instagram City: Reading the local through social media,” First Monday, volume 18, number 7, at http://firstmonday.org/article/view/4711/3698, accessed 11 February 2017.
doi: http://dx.doi.org/10.5210/fm.v18i7.4711, accessed 11 February 2017.

C.D.F. Honig, 2017. “Some rough sleepers are attracting tourists with their street art,” The Conversation (30 January), at http://theconversation.com/some-rough-sleepers-are-attracting-tourists-with-their-street-art-71983, accessed 11 February 2017.

C.D.F. Honig and L. MacDowall, 2016. “Audience constructed genre with Instagram: Street art and graffiti,” First Monday, volume 21, number 8, at http://firstmonday.org/article/view/6810/5600, accessed 11 February 2017.
doi: http://dx.doi.org/10.5210/fm.v21i8.6810, accessed 11 February 2017.

J. Jackson, 2015. “Instagram engineer delves into emoji madness,” PC World (7 May), at http://www.pcworld.idg.com.au/article/574410/instagram-engineer-delves-into-emoji-madness/, accessed 11 February 2017.

C. Lampe, R. Wash, A. Velasquez and E. Ozkaya, 2010. “Motivations to participate in online communities,” CHI ’10: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 1,927–1,936.
doi: http://dx.doi.org/10.1145/1753326.1753616, accessed 11 February 2017.

X. Li, T.-A.N. Pham, G. Cong, Q. Yuan, X.-L. Li and S. Krishnaswamy, 2015. “Where you Instagram? Associating your Instagram photos with points of interest,” CIKM ’15: Proceedings of the 24th ACM International on Conference on Information and Knowledge Management, pp. 1,231–1,240.
doi: http://dx.doi.org/10.1145/2806416.2806463, accessed 11 February 2017.

L. MacDowall, 2008. “The graffiti archive and the digital city,” In: D. Butt, J. Bywater and N. Paul (editord). Place: Location and belonging in new media contexts. Cambridge: Cambridge Scholars Publishing, pp. 134–147.

C. McCormick, in collaboration with Wooster Collective’s Marc and S. Schiller, 2010. Trespass: A history of uncommissioned urban art. New York: Taschen.

J. Metcalf and K. Crawford, 2016. “Where are human subjects in big data research? The emerging ethics divide,” Big Data & Society, volume 3, number 1.
doi: http://dx.doi.org/10.1177/2053951716650211, accessed 11 February 2017.

S. Petrović, M. Osborne and V. Lavrenko, 2010. “Streaming first story detection with application to Twitter,” HLT ’10: Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pp. 181–189.

N.M. Richards and J.H. King, 2014. “Big data ethics,” Wake Forest Law Review, volume 49, pp. 393–432.

R. Schacter, 2016. “Street art is a period. Period. Or the emergence of intermural art,” Hyperallergic (16 July), athttp://hyperallergic.com/310616/street-art-is-a-period-period-or-the-emergence-of-intermural-art/, accessed 11 February 2017.

R. Schacter, 2015. “From dissent to decorative: Why street art sold out and gentrified our cities,” The Conversation (9 November), at http://theconversation.com/from-dissident-to-decorative-why-street-art-sold-out-and-gentrified-our-cities-46030, accessed 11 February 2017.

T. Shelton, A. Poorthuis and M. Zook, 2015. “Social media and the city: Rethinking urban socio-spatial inequality using user-generated geographic information,” Landscape and Urban Planning, volume 142, pp. 198–211.
doi: http://dx.doi.org/10.1016/j.landurbplan.2015.02.020, accessed 11 February 2017.

T.H. Silva, P.O.S. Vaz de Melo, J.M. Almeida, J. Salles and A.A. Loureiro, 2013. “A comparison of Foursquare and Instagram to the study of city dynamics and urban social behavior,” UrbComp ’13: Proceedings of the Second ACM SIGKDD International Workshop on Urban Computing, article number 4.
doi: http://dx.doi.org/10.1145/2505821.2505836, accessed 11 February 2017.

M. Thompson, 2009. American graffiti. London: Parkstone.

E. Troshynski, C. Lee and P. Dourish, 2008. ”Accountabilities of presence: Reframing location-based systems,“ CHI ’08: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, pp. 487–496.
doi: http://dx.doi.org/10.1145/1357054.1357133, accessed 11 February 2017.

A. Young, 2016. Street art world. London: Reaktion Books.

A. Young, 2014. Street art, public city: Law, crime and the urban imagination. Abingdon, Oxfordshire: Routledge.

 


Editorial history

Received 25 October 2016; revised 2 February 2017; accepted 9 February 2017.


Copyright © 2017, Christopher D.F. Honig and Lachlan MacDowall.

Spatio-temporal mapping of street art using Instagram
Christopher D.F. Honig and Lachlan MacDowall.
First Monday, Volume 22, Number 3 - 6 March 2017
http://www.firstmonday.dk/ojs/index.php/fm/article/view/7072/5921
doi: http://dx.doi.org/10.5210/fm.v22i13.7072





A Great Cities Initiative of the University of Illinois at Chicago University Library.

© First Monday, 1995-2017. ISSN 1396-0466.