December 08, 2015

What Would a Floating Sheep Map? The Manifesto

After six years of doing this FloatingSheep thing and something like 400 different posts, we finally decided to summarize a bit of our research in our newly published essay/manifesto, entitled "What Would a Floating Sheep Map?". A collaboration between the five members of the FloatingSheep collective and Rich Donohue and Matt Wilson of the University of Kentucky, our manifesto tries to provide some context to the broader debates about mapping that we've been participating in for the last several years.

Our manifesto is about maps, but it's not just about maps. We like maps and celebrate high-quality, aesthetically-pleasing, intellectually engaging and (yes) amusing maps. Our aim, however, is to do much more than express an appreciation for maps. We also want to challenge people to think deeply about space, construct maps that demonstrate an awareness of social contexts, and critique these very same maps. Thus, our manifesto focuses not on the specific techniques or technologies of map-making, but on the societies, spaces, and places that are being mapped, and from which maps emerge.

While we'll leave you to actually read the entire thing, as a bit of a teaser, here are the six key points that summarize our manifesto, and, ultimately, the last six years of FloatingSheep research that we've shared with you here on the blog.
#1a. Maps help us to understand the world.
#1b. But maps also produce the world as we know it. 
#2a. All maps lie.
#2b. But lies are the source of maps' power. 
#3a. Maps are now fundamentally different.
#3b. But maps are still fundamentally the same. 
#4a. Pretty maps are better than ugly maps.
#4b. But ugly maps will do in a pinch. 
#5a. Map or be mapped.
#5b. But not everything can (or should) be mapped. 
#6a. Maps of the digital world are a reflection of the material world.
#6b. But this reflection is imperfect and distorted.
Zook, Matthew, Taylor Shelton, Ate Poorthuis, Rich Donohue, Matthew Wilson, Mark Graham, and Monica Stephens. What Would a Floating Sheep Map? Lexington, KY: Oves Natantes Press, 2015.

October 09, 2015

Floating Sheep in Comic Form

Have you ever wondered what it is like to take a classes at one of home institutions of Floating Sheep? In this case, the Department of Geography at the University of Kentucky, a place where the land flows with milk, honey and cutting-edge insight with giant herds of critical GIS researchers roaming the tundra and bellowing out challenges?

OK, perhaps not. 

But it is a place where we have stock-piled Floating Sheep stickers and tattoos for the coming apocalypse. And sometimes these artifacts emerge and are reused and parodied by the denizens of the department.

Ergo [1], the following cartoon [2] has appeared in Kentucky, the work of the talented Emily Kaufman, Ph.D. student. And while the Floating Sheep collective would love to have come up with the "Computer is not a banana" as a metaphoric/allegoric/crazed bon mot -- if for no reason beyond its mysterious incomprehensibility -- we cannot take credit. This is the idea of Jeremy Crampton, who is actually the professorial figure featured in the cartoon.

And no, we don't under why a "computer is not a banana" [3] either.

[1] What a great word, love to use that word. 
[2] Technically a mixed media collage of paper, ink and tattoo.
[3] We are in cautiously in agreement with this statement although we think the idea of a banana-based computer is great idea.

September 21, 2015

The Prime Minister and the Pig: Mapping #piggate

“The creatures outside looked from pig to man, and from man to pig, and from pig to man again; but already it was impossible to say which was which.” - George Orwell
It has recently been revealed that David Cameron, the Prime Minister of the United Kingdom, has potentially engaged in sexual acts with the head of a pig. Yes, really [1]. We assume the technical term would be beastialnecrophiliality. Or Necrophilibeastialogy. Or Ick!

Because of our deep commitments to both data science and animal welfare here at Floatingsheep, we decided that we needed a better understanding of how the conversation around #piggate is spreading geographically. Also, how we can resist such a wonderful story about a "debauched and secretive society" at one of the home institutions of FloatingSheep. Hint: it is NOT the University of Kentucky.

In the map above, you can see that people in the UK - for obvious reasons - are most likely to be tweeting about the Prime Minister and his, erm, attachment to pigs [2].

The French, interestingly, seem to be largely ignoring the story. Whereas the Spanish, Irish, and Americans all seem to be highly tuned-in (perhaps because of the love of bacon and ham in all three countries).

The spatial dynamics of #piggate will, of course, evolve over the next few hours and days; but until then, this is perhaps the obvious next addition to any conversation that begins with 'where were you during #piggate 2015.' What a time to be alive!

Unless of course you happen to be the severed head of porcine.

[1] Yes, that is actually the accusation. We're still have trouble processing the news as well.
[2] Perhaps rapprochement is better term. After all it's a French word and makes it sound somehow better. And frankly this needs all the help it can get.

August 12, 2015

Punch Buggies vs. Perdiddles: The Geography of Road Trip Games

As the final days of summer come to a close, we start to wax nostalgic about long car trips from yesteryear. Sadly, the children of today -- hooked into their pads, phones and other devices -- will little understand the joys of being packed in the un-air-conditioned back seat with various siblings and/or cousins for 5+ hours with nothing to do but look outside and get on each other's nerves.

Of course, desperate parents came up with all manner of games -- license plate bingo, the alphabet game, twenty questions -- to try and keep the peace and a little bit of their own sanity. But truth be told, these were but stop gaps for the real games -- punch buggy and perdiddle -- which provide legally sanctioned channels for punching your sisters. While rules vary (the strip perdiddle game variant cited in Wikipedia is particular unnerving for family road trips), the basic rule is the first person who sees a "punch buggy" (a VW beetle) OR a "perdiddle" (a car with one burnt out headlight) "wins"; you now have tacit permission (according to the international charter of kid rules) to punch someone [1].

Over the years as multiple people stare blankly as I shout perdiddle, it has become apparent that these games (like most human culture) have lumpy spatial patterns, as is evidenced in our map of references to these games in geotagged tweets below [2]. 

Mapping Punch Buggy vs. Perdiddle Tweets

Using Twitter as a proxy for the preference for these road trip games, it is clear that "perdiddle" is mostly a midwestern to east coast phenomenon, with Ohio, Pennsylvania and Massachusetts being particularly prominent. While punch buggy had a greater overall number of references in our data (3612 total tweets, as compared to 1327 for perdiddle), it seems to be more spatially dispersed across the United States, with a somewhat greater prevalence in the west, and the west coast in particular. Perhaps because this is the natural habitat for VW Beetles?

We're curious if this distribution fits with your experience, so let us know.

[1] Of course, you might get in trouble with your parents but that's another issue.
[2] We ran searches for (1) padiddle OR pediddle OR perdiddle OR kadiddle, and (2) "punch buggy" OR "slug bug" for tweets sent from 2012 to 2015, aggregated them into hexbins and calculated a simple odds ratio to compare the two. This is basically the same approach we've done for awhile, so no need to rehash the details here. If you'd like to see more on this methodology, check out our forthcoming book chapter.

July 22, 2015

New job working with the Geonet team at the Oxford Internet Institute: 'Researcher in ICTs, Geography and Development'

Mark is now hiring a researcher to work at the Oxford Internet Institute to investigate low-wage digital work being carried out in Sub-Saharan Africa:

The Oxford Internet Institute is a leading centre for research into individual, collective and institutional behaviour on the Internet. We are looking for a full-time Researcher to work with Professor Mark Graham on the ERC-funded project Geonet: Investigating the Changing Connectivities and Potentials of Sub-Saharan Africa's Knowledge Economy. Combining archival research, surveys, and in-depth interviews, this ambitious project will critically assess the changing landscape of digital work in Sub-Saharan Africa, and ask who benefits (and who doesn’t) from those changes.

In this exciting role, the Researcher will carry out 9-12 months of fieldwork among digital workers and organizations in Sub-Saharan Africa, as well as working at OII’s premises in Oxford. The Researcher will also contribute to the dissemination of the findings through peer-reviewed academic papers, project reports, events, blogs and social media.

Candidates should have experience of social science research in Development Studies, Geography, Sociology, Social Anthropology, Communications, Organization Studies, Management or related disciplines, training and practical experience in qualitative research methods.

Based primarily at the Oxford Internet Institute (with periods of fieldwork), this position is available immediately for 3 years in the first instance, with the possibility of renewal thereafter, funding permitting. For qualified candidates, there may also be opportunities to teach course modules on our ‘Social Science of the Internet’ MSc course.

The application form and further details, including a job description and selection criteria, are available on Oxford University's recruitment website.

The closing date for applications is 12:00 BST on Thursday 3 September 2015 and only applications received before then can be considered. Interviews for those short-listed are currently planned to take place in the week commencing Monday 21 September 2015.

July 16, 2015


We love mapping beer, there's no secret about that. We've been making maps about the digital landscapes of beer across the world practically since we started this blog six years ago, and this work is consistently some of our most popular. This includes some maps on the geographies of beer-related tweeting in the United States, building from a book chapter by Matt and Ate.

Now we want you to join in the fun of exploring this liquid landscape. To celebrate the rollout of a new online graduate program in digital mapping (New Maps Plus at the University of Kentucky) we offer up this interactive visualization of America's beer-related tweeting.

(click the image above to go to the interactive map)

Choose a type or brand of beer and see where people tweet about it or compare the attention to two different kinds of beer. Special thanks to Rich Donohue who built this slick interactive user interface with the Leaflet library. If you're curious about how this map was built and designed (or are interested in doing something like this yourself) check out the New Maps Plus program.

More saturated (darker) colors indicate a higher probability of tweets containing a textual reference to the selected beer type. You can visually explore a variety of beers by selecting a new beer from the drop-down menu at the top right. By default, a given beer is normalized by a random sample of the overall Twitter population, though you can also compare two different beers by selecting another beer from the second drop-down menu.  Hexagons without a significant number of observations/tweets do not show up. That's why some beers have more coverage than others.

Feel free to start playing right away, but in order to whet your appetite, here are some examples what you'll find. Starting with arguably the most locally-specific beer on our list, one can clearly see how Grain Belt beer is thoroughly grounded in the culture of the upper midwest, especially in Minnesota, and to a lesser degree Iowa and Wisconsin. It's interesting to note, however, that despite this very particular concentration, Grain Belt barely cracks the top 10 list for absolute beer references throughout the area. This obviously raises the important issue of recognizing that (nearly) all Grain Belt drinkers are Minnesotans, but not all Minnesotans are Grain Belt drinkers! We must admit we've not had the pleasure of trying Grain Belt ourselves, and we're not quite sure if that is a good thing or a bad thing.

Grain Belt

Although the Boston Brewing Company will be quick to tell you that it is still a craft brewery, Sam Adams is remarkably more diffuse throughout the US. However, one can also see that Sam Adams' home is very clearly in Massachusetts and extending into Maine, New Hampshire and Vermont, though the beer remains less talked about in these locations than a number of other non-local varieties.

Sam Adams

Tweeting about Yuengling, however, represents a few interesting deviations from the patterns seen with Grain Belt and Sam Adams. For one, Yuengling has a much more prominent role within the Pennsylvania area, asserting itself as a top-5 beer-of-choice throughout the state, and even coming in as the #1 beer referenced in the area around Bethlehem, PA, not far from the Yuengling brewery in Pottsville. Second, while Yuengling is similar to Sam Adams in its wider distribution throughout the US, the number of references to the beer drop off significantly to the west of the Alleghenies, and are practically non-existent to the west of the Mississippi River. Finally it's interesting to note that Yuengling also represents the unique case of a regionally-specific beer that is actually multi-polar, as the beer is also prominent in Florida due to its secondary brewery being located in Tampa.


A reverse of this spatial distribution can be seen in the case of Shiner Bock, whose references are dominant in much of Texas, especially around the Spoetzl Brewery in Shiner. Though concentrations extend beyond the Lone Star state into Louisiana, Arkansas, Oklahoma and Kansas, don't even try to get midwestern or New England states on board with this Texas brew.

Shiner Bock

There's nothing like the simulated authenticity of drinking a cerveza when trying to cool down on a hot day. But, as a comparison of Corona and Dos Equis shows, which Mexican beer you choose is likely (at least a bit of) a function of where you are. While Corona tends to be more concentrated in California, Florida and parts of the northeast, Dos Equis tends to be concentrated in the middle part of the country, especially centered on Texas.

Corona vs. Dos Equis

Last but not least, we thought it important to take a closer look at the geography of the country's two most popular beers, Bud Light and Coors Light. And while Bud Light sales were well over double those of Coors Light in 2014, tweeting activity around these two popular watery substances (sorry, we're solidly in the craft beer camp) reveals some interesting caveats to this seemingly one-sided competition. Indeed, just to the west of the Mississippi River appears a fairly clear dividing line at which the bevy of Bud Light in the eastern United States gives way to a western preference for Coors Light. 

Bud Light vs. Coors Light

And while the eastern seaboard between New Jersey and Rhode Island seems to be the one eastern outpost of Coors Light, Bud Light actually remains the most popular beer being tweeted about in these areas. But because the statistical comparison looks not at absolute numbers, but the prevalence compared to the expectation at the national-level, the seeming competition here is a bit deceiving. Indeed, references to "Coors Light" itself are incredibly sparse throughout the US, and the term rarely cracks the top 10 for any given locale, although the more generalized "Coors" in these areas makes clear the regional preference.

Whew! That was a lot of work. We're off to kick-back and enjoy a cold one. Have fun with the map and be sure to tell us which beers we didn't include...We suspect there will be many future iterations!

If you want to learn how to make maps like this, check out the new online graduate certificate and master's degree in digital mapping from New Maps Plus! The first batch of classes start October 4!

April 04, 2015

Cats and Badgers: Mapping the Final Four

Later tonight, the beloved Wildcats of the University of Kentucky men’s basketball team face the University of Wisconsin Badgers in the Final Four. And while the score will ultimately be settled on the court, this (sort of) rivalry can’t help but spread to other arenas. You’ve got the greatest tradition of cheese-making and beer brewing in the United States up against the greatest tradition of burning couches, making (and drinking!) bourbon and betting on horses. You’ve got those former Wisconsin-ites of Maptime Boston starting stuff with the good hearted Kentucky folks at MaptimeLEX. You’ve got new maps versus old maps.

Indeed, the map itself offers an important terrain on which to fight the battle between Kentucky and Wisconsin. The unsourced map below represents just the first cartographic shot fired in this war. While Wisconsinites might like to think that only their Minnesotan rivals to the west are the only ones that would side with Kentucky in this match-up, how are we to believe such cartographic assertions without interrogating the context of the map?

From what dataset is this map constructed? Why is this data aggregated to the state level, anyways? Why isn’t Canada similarly disaggregated into province-level data? Are we really to assume that such administrative regions are internally coherent in terms of their sports-rooting interests? I mean, UK even has a half-Canadian player! Even if we are assuming that, why are we assuming that Mexico doesn’t like college basketball, too? Similarly, why are there no gradations to reflect those places that might be relatively conflicted about who they’re rooting for? Why is there no context regarding the motivations behind such rooting interests (e.g., Michigan State and Duke fans who would much rather play Wisconsin than Kentucky in the championship game)?

Quite obviously, we can’t trust this map. Maps are fallible and power-laden and subject to the whims of the always-partial cartographer, in this case some Wisconsin fan who needed to make themselves feel good before a second-straight Final Four defeat at the hands of the greatest program in the history of college basketball. What we can trust, however, is that always transparent window into the collective social psyche of 21st century America: our database of geotagged tweets. This kind of data is objective and apolitical, a true representation of the world as it actually is. So, collecting nearly a month’s worth of tweets in the continental US from March 1 thru March 30 for the more-or-less official school hashtags of #BBN and #OnWisconsin, we sought to create a more objective, data-driven representation of the geography of Saturday’s matchup, as seen below.

Comparing #BBN and #OnWisconsin Tweets [1]

At the broadest level, #BBN tweets outnumber #OnWisconsin tweets by over 5x: 5,707 to 1,039. For a state that’s only about three-quarters as populated and a university that has only about two-thirds as many students, I’d say that’s a resounding victory for the Big Blue Nation. While each team dominates its own state (though we should note, there are some parts of Wisconsin without any pro-Wisconsin tweets at all!), those areas with more #BBN tweets than #OnWisconsin tweets are a bit more numerous throughout the country. Importantly, Kentucky fans have taken over key outposts such as Chicago and New York City, along with Nashville (the site of this year’s Southeastern Conference tournament), Cleveland (the site of the Midwest Regional) and Indianapolis, the site of the Final Four itself, not to mention much of the south, more generally. These are important strategic victories in the geography of basketball fandom this week, as the only other state Wisconsin fans have dominated apart from their own is Minnesota, in clear contrast to the stylized representation seen up above. Indeed, it’s quite telling that based on our hexagonal aggregation, no place in the United States has more than 100 more #OnWisconsin tweets than #BBN tweets: not even Madison, Wisconsin!

While there are still a number of places throughout the country where Wildcats and Badgers seem to be coexisting peacefully (thus far, at least!) -- and a good number more that seem not to particularly care -- we can say with some level of certainty that the members of the BBN are crazy and the Cats won’t be stopped on their way to another national championship. At least those folks in Madison are good at making maps! [2]

[1] Many thanks go out to Eric Huntley, PhD student in the Department of Geography at the University of Kentucky, for his help with this map!
[2] No hard feelings, Wisconsin fans. We love y'all. But seriously, GO CAYTS! And may we have another occasion to publish an academic paper about UK fans celebrating a national championship!

April 01, 2015

New paper accepted – Mapping Information Wealth and Poverty: The Geography of Gazetteers

  Spatial distribution of placenames in the GeoNames gazetteer

  Spatial distibution of placenames in GeoNames included in the dataset of populated places with more than a thousand inhabitants, compared to the spatial distribution of population

Mark and his colleague Stefano have put together a short paper that will be forthcoming in Environment and Planning A. The paper focuses on the geography of geographic information, and builds on Mark's work into the uneven geographies of information. It highlights how the very information systems that we use as 'ground-truth' are themselves characterised by significant biases.

Gazetteers are dictionaries of geographic placenames that have important implications far beyond the worlds of geographers and cartographers. By containing ‘definitive’ lists of places, gazetteers have the ontological power to define what will and won’t be geocoded and represented in databases, maps, search engines, and ultimately our spatial understandings of place.

The paper focuses attention on GeoNames, which is the world’s largest freely available and widely used gazetteer. It illustrates how content in GeoNames is characterised by highly uneven spatial distributions. There are dense clusters of placenames in some parts of the world and a relative absence of geographic content in others. These patterns are related to not just the wealth and population-size of a country, but also its policies on internet access and open data. The paper then traces some of the specific implications of this information inequality: showing how biases in gazetteers are propagated in a variety of geographic meaning-making.

Graham, M. and De Sabbata, S. 2016. Mapping Information Wealth and Poverty: The Geography of Gazetteers. Environment and Planning A. (in press).

March 11, 2015

New paper accepted in Landscape and Urban Planning!

We're very happy to announce that three-fifths of the collective have recently had a new paper accepted for publication in the journal Landscape and Urban Planning, as part of a special issue on critical visualization edited by Annette Kim, Katherine Foo, Emily Gallagher and Ian Bishop. Taylor, Ate and Matt's paper, "Social media and the city: rethinking urban socio-spatial inequality using user-generated geographic information", builds on our earlier calls to go 'beyond the geotag' in order to develop alternative conceptual and methodological approaches for the use of geotagged social media data, drawing attention to the variety and complexity of socio-spatial processes embedded in such data.

Using Louisville, Kentucky as a case study, our paper examines the socio-spatial imaginary of the '9th Street Divide', separating the city's largely poor and African-American West End from its more affluent and predominantly white areas to the east.

While a more conventional analysis of these inequalities as reflected in geotagged tweets might look a bit like the map above, we argue that such maps of isolated, atomistic dots do little to reveal the nature of inequality between places, and do a disservice to the data itself by stripping it of much of its context. So, rather than just arguing that the West End seems to have a relative lack of tweeting activity compared to other parts of the city -- and thus deducing that the digital divide is persistently reflected in this data -- we put these different areas of the city in comparison to one another in order to understand how both individuals and groups move through the city and (re)produce landscapes of segregation and inequality through their everyday practices and mobilities.

Using a novel method for analyzing this data, we attempt to demonstrate how the idea of the West End as a separate and apart from the rest of the city is challenged by the realities of people's everyday movements. Rather than being isolated, West End residents are actually much more spatially mobile within the city, while East End residents tend to be much more confined to their own neighborhoods.

So while the 9th Street Divide remains a key way of understanding and highlighting the spatial dimension of urban inequality in Louisville, we tend to think that this framing actually reinforces the understanding of the West End as a kind of 'problem area'. And while only a partial contribution to this argument, we hope that understanding the West End through its relations with, and connections to, other spaces and places ameliorates the vilification and pathologizing that is so common in discussion of racial and socio-economic inequality in highly segregated cities.

Ultimately, we hope this paper can allow for an alternative conceptualization of urban inequality in Louisville and the West End, while also demonstrating the utility of a situated and contextualized, mixed methods approach to the study of geotagged social media data, emphasizing the full range of socio-spatial processes embedded in this data that can't be captured in just a single point on a map.

The full citation for our paper is below:
Shelton, Taylor, Ate Poorthuis, and Matthew Zook. (Forthcoming) Social media and the city: rethinking urban socio-spatial inequality using user-generated geographic information. Landscape and Urban Planning.

February 19, 2015

Happy Year of the Sheep: Mapping "Literally" Every Sheep in the United States

The recent blog post by the Washington Post's Christoper Ingraham mapping “Literally every goat in the United States" was interesting in a number of ways. First, it continues the rhetorical trend of making maps (or at least claiming to make maps) that include everything.

If nothing else, it provides a compelling example of the current cultural capital value of "big data" in society. This is an interesting cultural moment in popular cartography, since the fundamental task of maps is abstracting and representing. And even the map that supposed showed "every goat" was actually a representation with each dot on the map standing in for 500 goats.

Secondly, we very much doubt that any map produced by the USDA Agricultural Census has ever received this much attention in the history of the agency.

Thirdly, why are the goats getting all the press? After all, isn't this the year of the sheep according to the Chinese zodiac [1]?  This is not the way Pan, god of shepherds, meant it to be. 

So being the ovis-chauvinists we are, we wanted to point out that there are actually twice as many sheep as goats in the US, and so the sheep population could probably take the goat population if it ever came down to hand to hand (or hoof to hoof) combat.

Also the USDA has made some fascinating interesting maps of sheep. Not baaaaaad.

Every Sheep in America

The USDA's Agricultural Census found that there were 5,364,844 sheep in the US in 2012. California was the sheepiest state in the USA with 668,517 sheep. However, Weld County, Colorado is the sheepiest county in the USA with 204,694 sheep reported in 2012. That's more sheep than Alaska, Delaware, Rhode Island, Connecticut, New Hampshire, Louisiana, Maine, Massachusetts, South Carolina, Mississippi, New Jersey, Florida, Vermont, Arkansas, Maryland and Alabama put together. Go Weld County [2]! We don't know about you, but we're booking our tickets and hotels to go visit ASAP.

However, Van Zandt County, Texas had the most valuable sheep compared to other crops with 68.38% of total market value of agricultural products sold originated from sheep, goats and their products (milk, wool, etc.). We've also found that the map of every sheep in the US opens up many perceptual rivalries with optical illusions hidden within. Yes, we are comparing our map to the illustrations of Sandro Del Prete. Please comment on this post--what do you see in the illustrations? The profile image of a lady in a bonnet? A man's naughty bits? A sheep?

 [1] We are aware of the sheep/goat confusion, but come down firmly on the side of the sheep.
 [2] Read more about Weld's extraordinary sheep processing.

February 11, 2015

IronSheep 2015. Be there! Or be an Iguana!

We are delighted to announce that we will be hosting IronSheep 2015 at TechNexus[1] in Chicago, Illinois on April 23 from 6-9:30pm. This is right in the middle of the AAG meetings so a great chance to get your sheep on during an academic conference.

If you are wondering "What is IronSheep?", the short answer is that it is our annual hackathon/mapathon modeled after the TV show IronChef (or project runway). Everyone gets the same set of data and competes to make the best map in a limited amount of time. And William Shatner will judge your map!!  Ok, the last bit is not true but we're working on it.

You'll be assigned a team with members who have a range of skills and all teams will have a mission to complete with common data.  

The best product at the end will win an award. The worst product at the end will win an award. Ok, almost everyone will win an award. We're just nice that way.

Anyone can join in the fun but we do ask that folks register for the event here:

There are also many AAG sessions relevant to our IronSheep event including a couple of "Future of Mapping" panels right before. We'll let you know in an upcoming post about those.

Check out our previous IronSheep events:

2012 at Pivotal Labs in New York
2013 at LARTA in Los Angeles
2014 at the Wave in Tampa Bay

[1] A big thank you to TechNexus who is sponsoring IronSheep by providing a space with wifi and breakout rooms.  We'll provide dinner and stickers, and you bring your laptop and your sheep (aka labor).

January 09, 2015

Mapping the Twitter Reaction to the Charlie Hedbo Attack

Following the attack on the offices of the French satirical magazine Charlie Hebdo, Twitter -- and those who make maps of it -- were all aflame with discussions, speculations and conclusions. In order to process the geographic extent of the reaction to the Charlie Hebdo attacks, we collected approximately 73,000 geotagged tweets created in a roughly 36-hour period from January 7th to noon (EST) on January 8th, that contained either of the hashtags: #charliehebdo OR #jesuischarlie (English translation: 'I am Charlie').

We then aggregated these tweets to the country level and normalized these tweets by a random sample of tweets in each country during the same time period [1]. We excluded countries that did not meet a minimum threshold of activity (15 tweets) to exclude places with extremely low levels of engagement. The map below was created by Rich Donohue, a post-doc at the University of Kentucky Department of Geography, whose cartography will be showing up on the blog more in the near future. The interactive version of the map allows you to pan, zoom and select specific hashtags to reveal different patterns between the tweeting.

Normalized Distribution of Geotagged Tweets 
containing either #CharlieHebdo or #JeSuisCharlie
Click here for an interactive version of this map.

Those countries shaded in orange demonstrate a greater level of Charlie Hebdo-related tweeting than one would expect based on typical levels of tweeting [2], while those countries shaded in blue demonstrate a lower amount of tweeting than one might expect [3]. Countries shaded in grey failed to meet the minimum threshold of tweeting activity to be included, while the handful of countries in red -- France, Belgium and French Guyana -- have the highest relative number of Charlie Hebdo-related tweets.

As expected, the reaction to the Charlie Hebdo attack has mostly captured the public's attention in Europe, especially (and unsurprisingly) in France and Belgium, with a seeming distance decay effect as one moves away from Paris. But outside of Europe, one can see greater levels of tweeting about the attack in countries with historical -- often colonial -- ties to France, such as Algeria, Tunisia, Senegal and Canada, as well as French Guyana which has significantly more tweeting about the attacks than one would expect based on usual levels of tweeting [4]. Other countries, such as Australia, India and Pakistan, also demonstrate significant levels of tweeting about the attacks, but don't have the same kinds of historical connections to France that might explain such heightened awareness.

Countries with the Greatest Relative Number of Tweets 
containing either #CharlieHebdo or #JeSuisCharlie 
Note: A location quotient greater than 1 indicates a relatively higher higher level of tweets with hashtags relative to the normal amount of tweeting taking place. A location quotient less than 1 indicates a relatively lower higher level of tweets with hashtags relative to the normal amount of tweeting taking place. 

In addition, there are a number of noteworthy patterns that we wish to highlight although are not prepared to explain at this time.

While such patterns are fairly obvious and could easily be predicted, the data leave us with a number of lingering questions that we don't have ready answers for. For instance, why is there a greater level of attention to the attacks in India and Pakistan than in Turkey or Egypt, which are both nearer in absolute distance and, in some ways, social distance to the attacks in Paris? Why are Canadians more focused on the issue than people in the United States? Why are people in the United States roughly 15x more interested in the Charlie Hebdo attacks than in the attempted bombing of an NAACP branch in Colorado?

It's also interesting to explore the differences in how each hashtag is used, and how this effects the spatial distribution of the tweets. Is the use of #charliehebdo a simple indicator of attention to the event, while #jesuischarlie indicates solidarity with the magazine? For example, the UK has a relatively low amount of #charliehebdo tweeting (LQ = 0.84) but a much higher level of #jesuischarlie activity (LQ = 1.35). In contrast, other nearby countries such as Spain, Portugal, Algeria, Morocco, Tunisia have relatively more #charliehebdo than #jesuischarlie activity perhaps connected to a more fraught relationship with local populations and the satire contained within Charlie Hebdo cartoons. To be clear, the causes behind the observed patterns require much more in depth work than we can provide here and now.

Moreover, as always it's important to think about what kinds of discussions aren't captured in this particular dataset, such as discussions of the attacks in Arabic-speaking countries such as Saudi Arabia or Egypt, which use entirely different alphabets than we used in our search. While we don't want to read too much into these differences without further research, these issues do represent potentially interesting differences in the use of social media, both across space and different social groups.

It is also useful to track the distribution of tweets over time, which began shortly before noon Paris time and peaked approximately ten hours later.

Number of Geotagged Tweets Overtime (in ten minute blocks)

While we have surely raised more questions than we have answered in this post, hopefully this early attempt at mapping the response to the attacks provides some further food for thought for those wishing to delve deep into understanding the nature of the attacks and the response to them via social media.

[1] We used the following formula (location quotient) to normalize the data:

(# of tweets with hashtags in country / # of total tweets in country)
(# of tweets with hashtags globally / # of total tweets globally)

[2] With a location quotient greater than 1.
[3] With a location quotient less than 1.
[4] There were a number of Francophone African countries that had high location quotients but were excluded from this map because they did not meet the threshold of 15 tweets. This includes Côte d'Ivoire, Gabon, Burundi, Benin, Togo and Congo.  Other countries with strong ties to France -- New Caledonia, Fiji, and Saint Martin -- exhibited similar patterns.