Floatingsheep and the Open 2.0 Event

Those of you in (or near) Lexington, KY should check out the Open 2.0 Conference happening at the University of Kentucky. Lots of interesting speakers but in the spirit of shameless promotion, Matt will be giving a talk on the 19th at 9:00 am on "Open Geographies".

More Flickr Mapping

Building on our visualisation of 34 million geotagged Flickr images, we have decided to map the data normalised by population and area. In doing so, some quite interesting patterns are evident.

Flickr Images per 100,000 people

Flickr Images per 100 square km

Predictably, we see some of the same core-periphery patterns that are observable in other types of user-generated content (e.g. Wikipedia). More surprising is the fact that unlike the geography of Wikipedia content, there are a significant number of low-income countries with relative large amounts of content (i.e. images) per every 100,000 people and 100km. Cambodia, Oman, Namibia, South Africa, Nepal and a host of other countries all score highly using these normalised measures.

I would hypothesise that two factors are at play here. First, there are lower barriers to entry on Flickr versus Wikipedia. In other words, despite the openness of Wikipedia, it is still easier to upload geotagged photos to Flickr than to create a new article and defend it's existence against nominations for deletion and overzealous editors. Moreover, the binary developed vs. developing country division has always masked the range of differences between and within countries, e.g., an interesting comparison between Oman and Yemen.

Second, it is also probable that much of the content in low-income countries is created by visitors and tourists. For instance, a significant number of photos geotagged to Cambodia are likely tourist shots of the Angkor Wat temple complex rather than locally created scenes of more everyday events.

Whatever the reasons are, more research is clearly needed on the topic to uncover what the specific biases in authorship are. Furthermore, irrespective of the specific reasons, it remains that these maps continue to show significant unevenness in user-generated content around the world.

For further reading see:

Graham, M. 2010. Neogeography and the Palimpsests of Place. Tijdschrift voor Economische en Sociale Geografie 101(4): 422-436.

Zook, M. and M. Graham. 2007. The Creative Reconstruction of the Internet: Google and the privatization of cyberspace and DigiPlace. GeoForum 38(6): 1322-1343

More Fun with Spurious Correlations: Feminist and Falafel, Gay and Shepherd's Pie

Our series of posts on statistical correlations have shown that references to "Christianity" within cyberscapes are highly correlated with obesity, while places with many references to "feminist" and "gay" have the opposite relation. Despite the appeal of some of these correlations, per capita income likely remains the most important, albeit not the only, underlying factor in shaping these relationships.

Given the important causal role of income to obesity we were curious whether this might extend into the virtual world as well. In other words, can we correlate references to income within cyberscape with references to other cultural markers such as "gay" or "feminist". Earlier maps we made of references to "rich" and "poor" within the Google Maps database revealed concentrations of rich in expected places, e.g., the coastal regions of the U.S. Thus, we wondered how the number of references to "rich" would correlate to references to "feminist" or "gay".

Gratifyingly, there is a strong positive correlation between references to the keywords "feminist" and "rich" (0.795) and "gay" and "rich" (0.66). This is consistent with our idea that the virtual representations reflect(albeit not perfectly) many material practices, events and places. This is an idea that we explore in more detail in many of our more academicly geared papers and publications.

Feminist and Rich

Gay and Rich

But we'd be doing a disservice to the loyal followers of Floatingsheep if we didn't extend our line of questioning to the next improbable, and highly spurious, level. Given that we started our series of correlation experiments by discovering the negative relationship between obesity and virtual references to "feminist" and "gay", we must now ask what foods are associated with these terms within the cyberscape? A burning question to be sure.

As the graphs below illustrate, references to "feminist" and "falafel" are highly correlated (0.658), as are references to "gay" and "shepherd's pie" (0.629). Not exactly, the correlation that we expected, but data does not lie [1]. Perhaps the route to weight loss is falafel and shepherd's pie?

We're not entirely sure how to interpret this, but it is (ahem) food for thought [2].

Feminist and Falafel

Gay and Shepherd's Pie

[1] Data does however roll over and allow itself to be manipulated shamelessly. It really is a bit of a milquetoast. Sometimes we wish data would have a bit more of a backbone.

[2] Actually, we're fairly sure that the underlying cause of these relationships is once again income. At least in the case of falafel, i.e., falafel in the U.S. =~ more cosmopolitan =~ higher income. But we're intrigued by the significantly higher levels of shepherd's pie in New England (New York, Massachusetts, Vermont and Rhode Island) and the outlier of Maryland. Yum?

Youtube Censorship

An editorial in the NY Times this morning reminded me that Google is making some interesting data available about user created content, censorship and Geography on the Internet, namely Google's Transparency Project. It includes both data on government requests to censor and traffic flows by country. The direct requests to censor is a difficult variable to understand (i.e., not well specified, see the FAQ) but the traffic data reveals some really interesting patterns.

One of the most striking examples of censorship relates to Youtube which is also clearly tied to user generated content (of great interest to Floatingsheep). Using the Google interface one can see how access to this stream of user content has been restricted over time and space.

This is essentially one look at the flip side (location of consumption of user created content) of what Floatingsheep normally looks at (location of the production of user created content)

Some of the more striking examples are below (or play with the data yourself).

Given Canada's strong protection for personal liberty it is a good comparison to other countries. Note: all countries have some variation likely tied to normal congestion/outages on the Internet. The key is to look for big changes (i.e., drops) in the level of access. You can also compare the general level of access to Youtube between countries, e.g., compare Canada to Cuba.

CANADA, an example of a country with strong
protections for personal liberty

IRAN: Note the drop right after the disputed
2009 elections and protests


CHINA: Youtube is just not an option


TUNISIA: Also blocking Youtube

CUBA: Some access but lower than Canada

PAKISTAN: Access but an effort to censor during
a specific time period

TURKEY: Perhaps the most interesting, declining availability
of Youtube over time.

Visualizing Digiplace

Two of the basic concepts we use here at Floatingsheep are:

Cyberscape, or the cloud of geo-coded information through which we move everyday (handy visualization of cyberscape here).

AND

Digiplace, or the sorting of cyberscapes, often by software algorithms such as Google’s pagerank, to filter content and avoid information overload. A key part of digiplace is how it makes some parts of the cyberscape more visible while peripheralizing other areas.

We’ve never had a good visualization of digiplace…at least until now. But Taylor’s recent search (via his iPhone) on Google Maps for the keyword "Kroger grocery" resulted in a nice example. In the image below, Kroger’s is in the upper left and is marked by a red push pin. Interestingly, Meijer's, another grocery store chain, also shows up as a sponsored link. Digiplace in action.

Digiplace

While Google Maps does clearly label Meijer’s as a “sponsored link”, i.e., "paid for placement”, the digiplace produced by this search does change visibility within the retail landscape.

For further reading, see:
Zook, M. and M. Graham. 2007. The Creative Reconstruction of the Internet: Google and the privatization of cyberspace and DigiPlace. GeoForum 38(6): 1322-1343

Zook, M. and M. Graham. 2007. Mapping DigiPlace: Geo-coded Internet Data and the Perception of Place. Environment and Planning B 34(3): 466-482.

Floatingsheep in Mapping America

As any of you who have kept up with Floatingsheep for very long may know, we get quite excited when our work appears elsewhere - on the web, in books, in journals, on t-shirts and even beer steins. Most of the time such appearances are unexpected, and we just happen to stumble across them. Other times, the left hand doesn't know what the right hand is doing and 2/3 of the Floatingsheep crew is wondering how in the world one of our maps ended up in a new book, when the other 1/3 had spent his precious summer hours making a map quasi-acceptable for said publication...

Bitterness aside, a very fresh and revised version of our map depicting the prevalence of abortion providers and abortion alternatives has been published in a new book, Mapping America: Exploring the Continent by Fritz Kessler. Thanks to Martin Dodge for taking a picture of our appearance in the book and alerting some of us to its existence for the first time.

More Fun with Correlations

As we explored in an earlier post there are some interesting correlations (at the state level) between the number of references to beer and Christianity in the geoweb relative to obesity. (Recap: while it looks like Christianity is causing obesity it really a matter of income).

We're starting to look deeper at the interrelation of content on the geoweb but before we do, we thought we finally post a couple more obesity correlations. And to reiterate, the obesity data is coming from the CDC, not us.

This time around we compared the relative frequency of geoweb references to "feminist" and "gay" to the offline level of obesity at the state level. As the two graphs below demonstrate, states with more references to these two topics are less obese. Again, per capita income is likely an important (albeit not the only) factor in shaping these variables.

It is, however, interesting to look at the outliers from the general trend, e.g., Colorado in particular. Although it has the lowest level of obesity in any of the states, the relative number of references to "gay" and "feminist" in its cyberscape is much lower than states with similar obesity. In other words, to fit into the overall norm of U.S. states, Coloradans need to either (1) gain a lot more weight or (2) start using the terms "gay" and "feminist" more.

Food (pun intended) for thought. Stay tuned as we belly up for some falafel, wine and shepherd's pie.

Feminist and Obesity

Gay and Obesity