‘Dissatisfied with service delivery’

What do protestors say during South African ‘service delivery’ protests? Burning tires and violent actions make the news, but protesters’ perspectives are seldom heard in the media.  After the crowd has dispersed, what happens to the protesters and their demands? Unlike in other countries, where social media can be used to mobilise and bring issues to the attention of a wider public, in South Africa, social media are expensive and inaccessible to many. Police records describe the protests as ‘crowd control incidents’, they note whether the crowd was peaceful and lump a wide range of issues, grievances and campaigns together, categorising them as ‘Dissatisfied with service delivery’.

Police crowd control data - Map of protests in South Africa 01/01/2009-30/11/2012
Police crowd control data – Map of protests in South Africa 01/01/2009-30/11/2012

This visualisation project uses police data to represent the number of service delivery protest incidents in South Africa, during 2009-2012.

This visualisation shows how many protests are recorded in the police crowd control data for the period 01/01/2009-30/11/2012, The red circles indicate which areas have experienced more protests than others. We’ve also included links so that you can check Google to see which incidents received attention from South Africa’s media.

The project is work in progress by Marion Walton and UCT’s Interactive Media class. The class is taking their first steps in data journalism, and are learning about JSON data and the Google Maps API. We are currently cleaning the data and exploring visual techniques to show the frequency of protests and the nature and distribution of media coverage.

Code  adapted from Gabriel Svennerberg.

Getting started with social network analysis

I teach an MA course in Advanced Media Methodologies at the University of Cape Town. This  year I’m presenting an elective which introduces Media students to Social Network Analysis. I’m really looking forward to teaching the course and seeing how a conceptual grounding in social network analysis and the  techniques of visualisation will change the work my students are able to produce for their dissertations.

We don’t have much class time and there are so many new skills to be learned.  I decided to design the course around a series of exercises and readings that students can use to prepare before class.

Here is a first draft of the outline with the course readings and exercises. Any feedback welcome!

Analysing Social Media: Text, image, network

 

Early adopters (joined pre Dec 2007) in my own Twitter network
Early adopters (joined pre Dec 2007) in my own Twitter network

Week 1: Reading and exercise

Garton, L., Haythornthwaite, C., & Wellman, B. (2006). Studying Online Social Networks. Journal of Computer-Mediated Communication, 3(1).

  1. Create a blog (if you don’t have one already). You can use a free site such as wordpress.com. You’ll be posting your answers to the class assignments on the blog.
  2. After reading the Garton et al (2006) reading for this week, prepare and pilot a short interview. Your interview should explore a research participant’s use of social media to communicate with his/her strong ties and should be designed to yield both quantitative and qualitative data. Post a short rationale for the interview questions on your blog and bring the questions to class next week.
Spreadsheet listing connections in our class
Spreadsheet listing connections in our class
  • Complete the Connections spreadsheet We will use this to map social networks during class.
    1. Click through to the editable spreadsheet on Google Drive
    2. Add your details to the final line of the spreadsheet.
    3. I have already added my details and the fact that I know all of you.
    4. Add your details by putting your name below the final line of data in the first column. In the second column, (next to your name), add the name of any other student you already know in the class, one per line. (I have already added the connections between the Interactive Media production students.
    5. In the third column, indicate from which class you already know that student.
    6. If you know the student from more than one class, add another line with your name, the student’s name and the name of the additional class.

Week 2: Readings and exercises

Hansen, D., Shneiderman, B., & Smith, M. A. (2010). Analyzing Social Media Networks with NodeXL.  Morgan Kaufmann. (Chapter 3) Chapter 10)

Bruns, A., & Burgess, J. (2012). Researching News Discussion on Twitter. Journalism Studies, 13(5-6), 801–814.

1. As shown to you in class, and using the vertex data from the Connections spreadsheet:

  • Download NodeXL and follow the installation instructions. You will need a Windows PC with Excel (or Windows and Excel installed on your Mac). You will also need internet access on the machine. NodeXL will not work on the UCT network behind the firewall.
  • Work through the NodeXL tutorial
  • Create a NodeXL sociogram to depict the relationships recorded in the Connections spreadsheet
  • Calculate the graph metrics. What are the various centrality measures? What do these numbers mean? What does this suggest to you?
  • Are there any clusters? What do you notice about them? What does this mean?
  • What is the graph density? What does this tell us?
  • How can you make the graph more readable?
  • Create a matrix to depict the relationships..
  •  How would you go about showing how everyone in the class communicates with fellow students and tutors about the social media assignments?
  • Do you have any criticism of the data we collected or how NodeXL represents it? How could we improve the data in the graph?

2.      Advanced (for students who want to use social network data for creative projects)

Week 3: Readings and exercises

Hansen, D., Shneiderman, B., & Smith, M. A. (2010). Analyzing Social Media Networks with NodeXL.  Morgan Kaufmann. (Chapter 10)

  • Papacharissi, Z., de Fatima Oliveira, M.: Affective News and Networked Publics: The Rhythms of News Storytelling on #Egypt. Journal of Communication. 62, 2, 266–282 (2012).
  • Lewis, S.C. et al.: Content Analysis in an Era of Big Data: A Hybrid Approach to Computational and Manual Methods. Journal of Broadcasting & Electronic Media. 57, 1, 34–52 (2013).
  1. Read Hansen et al. Chapter 10 and download your own set of Twitter data to explore and graph your own personal network on Twitter.
  2. Download Twitter search data for a keyword that interests you.

Optional (for creative projects):

  1. Read Chapters 6-11 Stanton, J. (2013). Introduction to Data Science.
    http://jsresearch.net/groups/teachdatascience/
  2. Conduct your own popularity contest to compare and graph Twitter activity around two words or phrases which are in the news right now.

Anyone using the new version of Mapstraction?

When teaching my students at UCT’s Centre for Film and Media Studies how to visualise geographical data, I’ve previously used Mapstraction, along with a good textbook by Adam DuVander and his excellent Twitter geosearch example.

The Mapstraction lesson only required a couple of small updates to form the basis of an assignment where the students created their own custom version of a visualisation of geocoded tweets. It worked well, providing an excellent example of how to mashup social media data with a map. I also like the tutorial because it provides a relatively simple research tool for my postgrad students (who are usually not web developers). For example, I used it recently to research the applications used by journalists and delegates at the ANC’s December 2012 conference in Mangaung. It was also helpful as a way of showing students what a tiny proportion of twitter data is geocoded (usually lower than 1%), which smartphones are in use in various countries, and (perhaps most important) the dangers of assuming that the comments and activities of Twitter users in South Africa reflect the preoccupations of the population as a whole. As one of the delegates to the Mangaung conference tweeted ‘ANC’s masses are not your Twitter people. So Social Media Hype will mislead you’.

geocoding_mangaung
Geocoded tweets at ANC’s Mangaung conference in December 2012.
Applications used to post geolocated tweets from ANC Conference, Mangaung, 2012
Applications used to post geolocated tweets from ANC Conference, Mangaung, 2012

Mapstraction always appealed to me because of the ability to use it for open data providers, and the ease it promises if you want to switch from one map provider to another.

Unfortunately, it’s the nature of the game in this field that you have to keep running just to stand still. Preparing the new version of my course, I realised that I needed to update the tutorial as my exercise and the textbook builds on v2 of the Google Maps Javascript API. This is now deprecated and apparently won’t be available for much longer (until May 19 2013 to be precise). Given that I’d used Mapstraction, I didn’t think that it would be too difficult to make the switch, but sadly this was NOT the case.

geocoding_beach_durban_gv3
Geocoded tweets about the beach posted near Durban, Janauary 2012, now using Google Maps API v3

The new version of the mashup (you can try it out here) allows you to search Twitter for geocoded tweets, and after searching you can summarise and view the twitter data.

I’m now using Gabriel Svennerberg’s textbook, and his Google Maps API v 3 JSON tutorial.

I still use the Twitter API and most of DuVander’s code for mining the JSON data, but I (reluctantly) abandoned Mapstraction, as I struggled a bit to get it to work, and I’m a bit short of time. I now display the data directly via Google Maps Javascript API v 3 instead of using the Mapstraction layer.

I’ve added a couple of features which I believe may be useful to researchers, and which I hope will spur my students to engage with the Twitter data in a more focused way. (I’ve found students like to decorate the maps but are overly cautious when it comes to making use of the additional data available from the tweets). The new version is just a start, but it provides a list of geocoded tweets, allows the user to see all the query results in JSON format or download the data as a JSON text file (although this requires browser popup windows to be enabled).

I tried updating the example to use the new version of Mapstraction and the Google Maps Javascript API v3 but ran out of time. I wondered whether anyone else has seen working examples which use Mapstraction together with Google API v3 and Twitter data? Please do let me know! Comments on the mashup example are welcome, though it is very much work in progress.

 

Affording images: Digital imaging and media-sharing practices in a corpus of young people’s cameraphone images

Paper presented at Multimodality in Education colloquium held at Mont Fleur, Stellenbosch on 10 August, 2011 by Marion Walton and Silke Hassreiter, Centre for Film and Media Studies. University of Cape Town

The affordances of mobile phones as devices for creating, publishing and distributing images means that they are often seen as a threat to young people’s safety or to public morality. Alternatively, they are celebrated as having immense potential for supporting an individualised and highly networked mode of mobile learning or ‘m-learning’. These issues are particularly significant in the global South, where photographic practices and digital imaging are being adopted rapidly, as mobile networks reach over a billion people and feature phones with cameras become increasingly accessible.

This paper documents the image-sharing and photographic practices of fourteen young people who participated in a mobile video-making project over four months in July-November 2010 in Makhaza, Khayelitsha. We analyse the corpus of images which they shared with us as researchers. We explore distinct communicative genres which, in this context, are associated with (i) personal photographs, (ii) photographic composites (iii) downloaded images from popular culture (iv) multimodal image messaging. In this paper, our focus is specifically on interpersonal meanings and the representation of interpersonal meanings and social distance.

We argue that the social practices of young people and the marginal contexts of this appropriation play key roles in their domestication of mobile photography. Consequently, it is a mistake to assume that new genres and practices can simply be ‘read off’ by listing the features or affordances of the new generations of smart phones. Instead, it is necessary to consider a wider range of contexts and uses before the ‘affordances’ of the new medium can start to be understood. In particular, the differences associated with the specific contextual meanings of artefacts such as mobile phones, local genres of communication and interaction, and broader issues of access to communication infrastructure and mobility need to be considered. We argue that a contextualised study such as this should be conducted before embarking on the development of new curricula for learning or self-expression for young people.

Mobile republic: Visual approaches to discourse in South African mobile social networks

Yesterday I presented some of my work on visualising mobile messaging discourse at ISEA2010 Ruhr. Here are the slides from my presentation:

Social distance in images from Flickr and Guguletu
Social distance in images from Flickr and Guguletu

Walton, Marion. 2010. Mobile republic: Visual approaches to discourse in South African mobile social networks. Paper to be presented at ISEA 2010, Ruhr, Germany in August. Prepublication draft

A new generation of South African Internet users network online via net-enabled phones. Despite limitations, mobile-centric internet allows connections with broader mediated publics. Mobile networking (both public and intimate) has the potential to reshape South African public discourse and change the social fabric, but social and economic divisions mean that mobile social interactions are currently almost entirely digitally invisible. Visualisations of social networks and the mobile Internet are presented to suggest some of the mediated conversations and networking taking place in the social networks of the majority.

Prepublication draft

Social distance and mobile photography

Do we feel impersonal distance, a sense of personal contact or intimacy in relation to the people we see in images? Chances are that the scale of the shot (whether a photograph is a close-up or taken at a distance, or somewhere inbetween) has something to do with that feeling. I’m interested in looking at situations where, in any given body of images, the represented distance between the camera and the subjects of the shot is used to generate an overall sense of closeness or distance towards particular groups of participants.

I wanted to be able to quantify an aspect of social distance. My motivation was that I wanted to explore some of the overall differences I’ve picked up between geolocated images posted to two social websites, mobile site The Grid and conventional photosharing site, Flickr (discussed here and here). Social distance is influenced by a number of factors, including the vertical and horizontal orientation of the camera, but shot scale seemed to be the easiest thing to measure and quantify, or a lot easier than camera orientation at any rate. So I measured the height of faces depicted in the photos I collected and graphed the results.

Below are some of the visualizations I developed using Processing. I found a particularly useful tutorial which gives a detailed explanation of how to build a visualisation to use data from a Google spreadsheet:

Shot scale distance distribution on The Grid and Flickr
Shot scale distance distribution on The Grid and Flickr - low values are long shots, high values are close ups
Distant shots predominate on Flickr
Distant shots predominate on Flickr

Close-up shots predominate on The Grid
Close-up shots predominate on The Grid

I found it far easier to do the analysis when I could compare the images side-by-side and so I created a new 3D view of them in five planes which correspond to my coding categories – intimate, personal, social, impersonal, and landscapes (for this study, I included other shots without any people in this category).

Visualising social distance in 3D planes - images from Flickr
Visualising social distance in 3D planes - images from Flickr

Flickr close-ups focus on children, food, drink, a dog at popular Guguletu butchery and outdoor restaurant, Mzoli’s. Photographers rarely feature in shots.Impersonal shots of landscape, buildings and distant township residents predominate.

Visualising social distance in 3D planes - images from The Grid
Visualising social distance in 3D planes - images from The Grid

Close-ups predominate on The Grid, often shot in self-portrait mode, with very few truly impersonal shots. Social distance is increased in some shots by the use of dark glasses and other distancing devices. Social distance is particularly difficult to code in some cases – if someone is photographed at what would otherwise be a ‘social’ distance, but in a provocative topless pose, it’s difficult not to code that shot as ‘intimate’.

If you’re interested in trying Processing, take a look at this introductory overview by my colleague Lyndon Daniels.

Local vs tourist views and mobile photography

Mobile snapshots
Online image-sharing sites such as Flickr currently reinforce the digital invisibility of the majority of the world’s population. This is a simple function of the fact that most people have not had access to consumer electronics, digital production and distribution, and even electricity. Recently cameraphones have become accessible to many more people, and digital publication is becoming more feasible, given that many platforms are now adapted or specifically developed for mobile use. For mobile industries eyeing emerging markets, multimedia communication practices can develop new markets for handsets and heavier use of mobile data networks. Academics and activists have spotted the possibilities of using mobile media to document grassroots stories, issues and new forms of journalism. But what possibilities do digital image-sharing platforms suggest for ordinary people? And to what extent will mobile publication platforms shift existing patterns of digital invisibility?

I’ve been working on a research project which investigates how cameraphone images are being published on South African mobile locative media-sharing platform, The Grid.Here is a short video which shows how The Grid’s designers imagined that their users might use the application’s locative features together with messaging and photographs taken with their cameraphones. The video shows trendy urban youth on the move, always in touch with one another, sharing their lives and emotions as they dart around the map, from home to train to mall to beachfront sunset to nightspot, co-ordinating busy social lives while capturing transient moments of beauty and fun.

VisualCultures

I’ve worked on the mobile snapshots posted to The Grid during a short sabbatical visit to the Multimodal Analysis Lab (MMA) at the National University of Singapore.  At NUS, linguist and multimodality researcher Kay O’Halloran and her team are working on a project called Mapping Asian Cultures, where they are developing visual tools to research and analyse visual culture in Asia. Their visualisation tool, VisualCultures, is currently in alpha stages of development, and during my visit I’ve been able to use the tool for my research into mobile photography from South Africa.

Cultural Analytics

The MMA VisualCultures tool is being produced in a joint effort with new media theorist, Lev Manovich, whose Cultural Analytics project at UCSD aims to use computational visualisation and supercomputing to study large sets of images. This will allow cultural scholars to gain insights into existing collections of artworks, but also provides new methods which are adapted to the explosion of visual images in contemporary culture, associated with user-generated content and Web2.0. The images above are taken from the Cultural Analytics Flickr page. They illustrate how VisualCultures uses the measurable  characteristics of individual works or pages (in these cases the mean brightness of Rothko paintings and of individual pages from manga title Zippy Ziggy) to create a large-scale overview of an artist’s oevre or of a manga title as a whole.

Brightness in Zippy Ziggy manga title
UCSD Cultural Visualisation project - Brightness in Zippy Ziggy manga title
Brightness in Rothko paintings (1830-1970)
UCSD Cultural Visualisation project - Brightness in Rothko paintings (1830-1970)

VisualCultures is  an open source project which runs in Adobe Air (free download). As the software runs on ordinary PCs rather than supercomputers it is useful for work with medium-sized image datasets (a maximum of 800 images). I’ve really enjoyed working with an early version to extend an analysis of South African mobile phone images that I started working on earlier this year. (I’ve written up an initial analysis in a short paper  accepted for design conference DIS2010).

VisualCultures helped me to compare geo-tagged images from Yahoo’s photo-sharing site Flickr and  from South African mobile locative media-sharing platform, The Grid. The images were taken or posted in low-income areas in Cape Town (they are mostly from Guguletu, but to increase the sample size slightly I included all images from neighbouring Nyanga as well).

While The Grid’s designers imagined prolific visual communciation between trendy groups of urban friends on the move, my study found quite different patterns of use. Users posted relatively few images, and they seldom document their environment, preferring to publish a couple of self-portraits. The Grid is still not widely used, and so, on this platform, visual communication seems to be taking place between geographically dispersed early adopters rather than between close friends whose daily lives are intertwined. The Grid  extends patterns of mobile interaction established by anonymous chat and IM, which are the most popular uses of dominant mobile platform in SA, MXit. The Grid seems to be used occasionally as a supplement or as an experiment, rather than as a primary means of communication. Nicky Allen from The Grid team reports that conversations on The Grid are primarily cross-gender (80%). The site seems to function to some extent as a dating app, with images often being used to introduce anonymous chat participants to one another. The limited number of images published may also relate to the fact that mobility costs money, and so does mobile bandwidth.

I compared the two sets of images along a number of dimensions. In this post I’ll focus on what I found about the geographical distribution of the images.

Multimodal Analysis Lab's VisualCultures - mockup
Multiple graph view in Multimodal Analysis Lab's VisualCultures (mockup)

Although VisualCultures doesn’t include a map function yet, I have mocked up a map view in Photoshop to show how the software allows one to display multiple graphs, and to investigate different views of the data at the same time.

The first graph is a map view, which maps the images according to longitude and latitude. The second graph (superimposed in the top left corner of the map) functions as a kind of user interface to the first graph, and allows the user to select categories of interest, in this case the two image platforms, Flickr and The Grid. By selecting the Flickr category on the graph I have tinted images on the map which belong to that particular category. These images of interest can then be viewed one by one in higher resolution if the user needs to take a closer look.

Locals and tourists - Guguletu and Nyanga images from Grid and Flickr
Locals and tourists - Guguletu images from Grid and Flickr

The map visualisation of the Guguletu data reminded me of the Locals and Tourists visualisations of Flickr images. Locals and Tourists contrasts the concentrated distribution of tourist shots (which tend to cluster around key tourist landmarks in cities) with the more dispersed distribution of photographs taken by people who live in these cities. The VisualCultures map was not entirely accurate, so I created a simpler visualisation with Processing (using code from Modest Maps) to get a more exact representation of their geographical distribution. Purple ellipses represent Flickr snapshots, and are concentrated around local township tour routes, with several shots taken around or near Klipfontein Road, the main thoroughfare through the area. A couple are taken from the N2 highway, which is the route to Cape Town International Airport. This is likely to be the only view tourists have of the area, if they do not take a township tour. In contrast, images from The Grid (blue squares) are distributed more evenly around the map, indicating more local patterns of use.

Flickr is used by both locals and tourists in the wealthier areas in South African cities. In contrast, the distribution of the Guguletu images suggests that there is very little local use of Flickr in these areas. Local views of South African townships may only be emerging on mobile platforms, where township residents are gaining access digital publishing opportunities for the first time. As these platforms are not always indexed, linked or aggregated on mainstream sites such as Google Images, these separate local mobile platforms may be perpetuating a certain kind of digital invisibility, and, for now, mobile creativity and expression remain on the margins.

The rest of my work on mobile snapshots involved looking at social distance in the shots, and I’ve developed my own visualisation tools for this – more on that in another post.

Ghost maps and urban networks

I am currently living in Singapore, where I am working on a research project in information visualisation. My first foray into the topic focused on getting very close to rivers of human waste, Victorian cesspools and the terrifying details of death by cholera.

Johnson, Steven. (2006) The ghost map: A street, a city, an epidemic and the hidden power of urban networks. London: Penguin.

Steven Johnson describes vividly how Victorian cholera victims fell ill, remaining mentally alert as, in excruciating pain, they witnessed their bodies emptying, fluids gushing out, leaving them shrunken, blue-skinned cadavers within 48 hours. His account is all the more terrifying because I am aware that similar deaths from cholera are still taking place pretty close to home, with cases in KwaZulu Natal recently and in Zimbabwe in 2009 and earlier this year as well. The difference is that Victorian authorities obsessed about the evil ‘miasmas’ or gases associated with the squalid living conditions of the lower classes that they believed caused the disease while we know both the causes and treatment for the disease. The vibrio cholerae pathogen is water-borne, and, ironically, cholera sufferers must be treated by rehydration. Today, cholera deaths are an absolute indictment of public health in a region.

What does this have to do with information visualization? Johnson details how Henry Snow, a physician and amateur scientist collaborated with devoted clergyman Henry Whitehead to solve the riddle of the source of a cholera infection around Broad Street in London (now Broadwick street in Soho). The two used the local knowledge of the infected community to make the maps that helped to counter prevailing orthodoxy, save many lives and introduce the kinds of public health strategems that make today’s mass human settlements viable.

John Snow's 'ghost map' of deaths in the Broad St cholera epidemic

John Snow’s ‘ghost map’ of deaths in the Broad St cholera epidemic

Snow mapped the locations of  the sick in relation to water sources, and his map showed how the site of the outbreak corresponded with the actual distances that various residents had to walk to get water in relation to the infected Broad Street pump.

Voronoi diagram showing cholera deaths and the community that used the Broad Street pump
Voronoi diagram showing cholera deaths and the community that used the Broad Street pump.

This particular visualization technique is known as a voronoi diagram, and as historians have pointed out, in this case Snow was mapping time (the time it took to walk to a source of water) as well as the spatial layout of the epidemic. With the help of Whitehead’s intimate knowledge of the affected community, Snow was able to prove that the index case of the cholera epidemic, or patient zero, was a young baby whose mother had washed nappies and disposed of water into a cesspool that drained into the pump. This action sealed the fate of the many nearby residents who walked to the Broad Street pump and drank its water over the next two weeks.

Johnson’s account left me with a renewed optimism for the rapidly growing urban networks at home in South Africa, and the probable future megacities which are likely to form if our current rapidly urbanising trend continues. As he points out, the Web links institutional knowledge with local knowledge of amateurs, and (in certain contexts at any rate) it has never been easier for local knowledge to find its way onto a map.

“Where Snow inscribed the location of pumps and cholera fatalities over the street grid, today’s mapmakers record a different kind of data: good public schools, Chinese takeout places, playgrounds, gay-friendly bars, open houses. All the local knowledge that so often remains trapped in the minds of neighbourhood residents can now be translated into map form and shared with the rest of the world.” (2006:219-220)

This is one of the reasons that I am particularly interested in exploring applications of the South African mobile locative social network The Grid. I’ve done some research into local geo-tagged images posted to The Grid from Guguletu which has been fascinating. I should say that my optimism is tempered by awareness of the many barriers and mediators between local knowledge, online publication, and institutional knowledge in the South African context. Johnson discusses how such online networks assisted in the measures taken against avian flu or H5N1. The more recent story of the tweet-fuelled H1N1 panic shows how such systems change the landscapes that they map, and is also a reminder that publics, media and public authorities can be all-too-fallible when they use such systems.

Many thanks to my host, Denisa, for letting me stay in her apartment and giving me access to her wonderful library while I am here.

Lyndon Daniels and I will be collecting examples of open source visualisation tools and track the progress of our project by posting them here over the next two months. I will kick off with Steve Fortune’s  voronoi polygon generator, written in python.

Guguletu and social media

I had lunch a couple of weeks ago with Richard Chalfen, an anthropologist from the Centre on Media and Child Health at Harvard. He has written some wonderful books about families’ collections of snapshots, or the ways in which ordinary people use photography in their lives. Chatting to him  inspired me to take another look at some data I collected a couple of months ago. It’s two sets of photographs from social media sites which shows up the contrast between the collections of images on the mobile social networks which are becoming popular in developing countries, and the photos posted to image sharing sites in the north, which tend to record the global excursions of wealthier northern tourists.

For the sake of the comparison, I chose Flickr, a well-established image-sharing site for digital photography, much loved by the digerati, and The Grid, a South African mobile social network with locative features which allows users to upload text, photographs, and video. Both Flickr and The Grid have a locative dimension, since Flickr users can geotag their images, and all content in The Grid is displayed on a map. I thought it would be interesting to look at what users are making of this spatial co-ordinate metadata. In South Africa’s recent history, apartheid’s racial policies of ‘separate development’ meant that space was destiny. Sadly this has not really changed much since then, so much so that even a technical term such as ‘location based service’ is unavoidably tainted by our past. The word echoes ‘location’, the term used for the under-serviced suburbs where black people had to live, and which still lives on in ekasi or township. I decided to look at images of Guguletu (one of the older townships in Cape Town) on both sites. Gugs is definitely off the beaten track for many tourists, but is on the route taken by some ‘township tours’ which take tourists away from their plush hotels and game farms and give them a smidgeon of social history, and a glimpse of the lives of ordinary South Africans.

I started the visualisations with social network analysis tool UCINET, but then switched to NodeXL when I realised how nicely this MS-Excel-based tool could handle images as data in visualisations of social networks.

Guguletu on Flickr - visualisation of geotagged images
Guguletu on Flickr - geotagged images

This is a collection of Flickr images that were geo-tagged and pinned on Flickr’s world map in the region of Guguletu. The process of geotagging images shot with digital cameras is a manual process for most photographers. A couple of the shots were definitely not taken in Gugs, while some of those which do depict Gugs are tagged and titled rather vaguely or inaccurately (e.g. ‘Khayelitsha’ or ‘South Africa’), depending on how serious the tourist was about tagging the photographs individually, (or perhaps whether they even  noticed details such as the name of the place they were visiting!)

As I’m not focusing on tourists’ use of social network sites, I coded the pictures and then grouped them roughly according to their theme. There’s a set of Driemanskap’s promotional pictures (they are a local Hip Hop crew), one lonely baby pic, a  documentary about Elections 2009 and an orphanage for HIV/AIDS orphans, several other pictures of children playing in the street, and the rest are pretty standard township tour shots by tourists and some visiting photographers. The visitors seem fascinated by South Africa’s housing problems and have definitely had a great time eating meat and drinking at Mzoli’s (some are travellers from abroad, while other travellers are from suburban South Africa).

I am interested in how South Africans are using social networks to share media, and so I coded the photos from The Grid differently to the ones from Flickr. For each picture I collected the comments that were posted in response to it, how many views each picture received, and what ratings they received from other users.

Guguletu on The Grid - mobile locative social network
Guguletu on Flickr - geotagged images

The visualisation above depicts a set of images posted to The Grid (where they are called ‘blips’) . (These images are all publically available via the mobile app or via The Grid’s website). The Grid uses mobile phone network data to work out where users are located  (more or less), and all user generated content is automatically geotagged on The Grid.I chose a set of images which were posted from Guguletu (and environs – the locative features do not necessarily respect suburban boundaries), and also captured the comment networks that sprang up around them. The size of the images  shows which ones were viewed most often, while the network indicates how many comments each picture received.

The contrast between the two collections is a result of many factors, which I’m exploring in a paper that I’m writing. Here are just some of them:

Most Flickr users were posting photographs taken with digital cameras, while The Grid users tend to post pictures taken on their phones.

Flickr has been around for ages, while The Grid is still only home to a small group of early adopters. Flickr users have been uploading large numbers of pictures over a long period of time, while on The Grid, many users only seem to try uploading a pic once or twice and then they often decide to return to their usual pastures on MXit.

Wealthier northern photographers with digital cameras have oodles of storage and cheap bandwidth to store and upload many shots, while in the south, mobile phone users tend to delete pictures as they run out of space, and can’t always afford the bandwidth for uploading and downloading lots of images from the web.

Social context is a key factor – what are people doing here? How does this shape the place they are representing? Is their physical location even important to what they are doing? In other words, what audience is addressed by the pictures, what kinds of conversations are taking place around and through the photographs on the two platforms. How does the architecture of each system influence the range and nature of social interactions that take place?

It struck me that Flickr geotaggers are operating in the third person. They are using these pictures to tell a (somewhat predictable) story about an exotic place.  The only exception to this are the promotional pictures for Driemanskap (where a professional photographer took the pictures in the Guguletu setting). In many cases, photographers are not depicted in the shot, but their friends and family feature, particularly in the party shots taken at Mzoli’s.

The Grid users are almost all posting self-portraits, with a few family portraits or peer portraits (‘me and my frendz’).  Although every image is geotagged there are very few references to the spatial environment in what the users choose to represent – only about two or three images represent a place rather than a person.

The Grid encourages users to rate one another’s pictures, which means that most comments centre on the them of ‘hot or not’, and they can almost all be characterised as insults or compliments.  The architecture encourages exhibitionism (at the moment I don’t think it is possible to share images with just one or two friends). As a result there’s a marked  intimacy and individuality to the pictures, collecting the photographs felt like overhearing snatches of conversations between friends.  This visual eavesdropping is another popular activity on The Grid. The social network looks cohesive, with many commenters commenting on more than one image from Guguletu. On closer inspection, these links between the separate cliques are in almost all cases ‘haters’ (or ‘trolls’)  – users who specialise in going from image to image ‘rating’ and flaming others.