Eric Fischer, a programmer and designer, created maps showing the locations of people when they send a Twitter message or upload a photo to Flickr. Orange dots represent the location of Flickr pictures, blue dots represent Twitter tweets and white dots represent the locations that have been attributed to both. Collectively, the dots yield powerfully beautiful images of the highly photographed and socially active areas of major cities like, New York, London and Tokyo.
Currently, Fischer has data visualizations for 30 different cities, each one with a different Flickr and Twitter geography. The images make for an interesting comparison. View them here.
We spoke with Eric Fischer to understand and explore the implications of these images in urban and transit planning.
Why did you choose to map the locations of Flickr and Twitter uploads?
Because they are the two best sources that I know of for information about the spatial patterns of large numbers of people’s lives. I had initially thought that Twitter and Flickr locations might show the same kinds of patterns, just for different individuals, but an initial look at the data for San Francisco showed a significant distinction between Flickr’s concentration in scenic and Twitter’s concentration in commercial areas. In addition, there turned out to be cities where Twitter is extremely heavily used that I had barely been aware of because Flickr is so rarely used there. So it is interesting to see what places they share and what places are specific to one or the other.
What do you think these images say about our interaction with our environment?
Well, at the most fundamental level, a tweet or a photo with a location signifies someone’s presence at a particular place at a particular time, and the presence of people in itself is indicative of the significance and interest of a place. And a picture goes further, saying that not only was someone there, they also saw something interesting enough to bother photographing and sharing. I’m not sure yet what the presence of a tweet says beyond someone’s presence. It might also mean having seen something interesting happen there, or it might just mean someone is filling time while standing in line.
When you look at the images, do you notice any trends and patterns among cities?
The standard pattern is that most cities have commercial areas that are well represented on both Flickr and Twitter, residential areas where Twitter is used but Flickr rarely is, and scenic places where Flickr is used but Twitter rarely is.
What are the implications of such images on city and transportation planning?
From a city planning perspective, what I think can be learned from this sort of data is both the positive of knowing what places are succeeding at attracting people and the negative of knowing where it is that people decide a place is no longer interesting enough to keep exploring and turn back. If you know where people are turning back, it may tell you where you could add a wayfinding sign to let people know about another interesting thing ahead that is just out of sight, or that you could change the characteristics of the unsuccessful place a little bit to make it more like the successful one and therefore make the hospitable area a little larger. And from the transportation perspective, it can tell you where the constituency of a place comes from, often how they are getting there, and perhaps what difficulties they are encountering on the way. My thinking about this owes a lot to Kevin Lynch’s mental mapping categories of paths, edges, districts, nodes, and landmarks in “The Image of the City,” all of which I think ought to be possible to derive from sufficient quantities of location data.
Are there any quirks about the visualization or the data that we should know?
The main disclaimer is that geotagged photos are only a tiny fraction of all photos and geotagged tweets are only a tiny fraction of all tweets. Many people never disclose locations for their tweets or photos and other people are selective about when they do.
If a location was referenced multiple times, it would have a bright location at the center and would then, with each additional reference, gain a dimmer halo around the edge. The idea is to give especially popular locations extra brightness, but to fuzz it a little so that it’s still possible to see the locations that are not the overwhelmingly most popular, instead of making the brightness of each pixel exactly proportional to the number of pictures or tweets for that exact spot.
What time frame do these images represent?
The Twitter tweets are from the past two months. I’m not sure what to tell you the exact range of the Flickr images is — they are a sample from the total Flickr set so that there are only as many images represented as there are tweets (about 61 million).
Is there a way to do this in real time? Would that even be useful?
It is certainly possible, and actually necessary, to monitor Twitter in real time, and Flickr has also recently announced a real-time “push” API. I think real-time search is more useful on Twitter than on Flickr, though, because there is often a lag between the time people take pictures and the time they upload them, whereas tweets are typically posted right as they are written.
Do you have any other projects you’re working on right now?
I am interested in doing further analysis on the Twitter data, paying attention not just to the locations but also to the text of what is tweeted in particular places, and in addition to the movement between places instead of treating each location in isolation.
