One of the great promises of social media was not just the ability to hear from ordinary people all over the world. It was to be able to put all of those voices on a map, to see what people in each corner of the earth were thinking and feeling. To flag a flurry of posts about an earthquake or a building fire and know where they were happening, down to the precision of a street corner. With its public posts and machine friendly data feeds, Twitter embodied this mappable promise, but it turns out that this grand vision of mapping social media was based on an assumption that proved false and a privacy setting that Twitter changed in April 2015, meaning our mappable future is becoming fuzzier and fuzzier and fading rapidly.
Social media isn’t just about giving ordinary citizens the ability to publish their thoughts to the world. It is about the networked infrastructure to help those voices be heard. Unlike the “social” platforms that preceded them and offered publication, but not promotion, the modern wave of social media platforms promised the ability to spread their users’ content far and wide to be heard by the world.
As social media became increasingly produced and consumed from GPS-enabled mobile devices, platforms rolled out geotagging to make it trivial to assign one’s precise geographic coordinates to each post.
This led to the grand vision of being able to map society’s dreams and fears, its events, narratives and emotions, in realtime. An earthquake could be detected in realtime and its damage assessed by those actually living in the affected areas. A protest could be experienced from ground level through the images, video and commentary of those actually participating and those witnessing it firsthand.
When I published one of the very first in-depth explorations of the geography of social media back in 2012, geotagged tweets accounted for just 3% of all tweets, but with Twitter spreading rapidly and virally across the world, the prevailing assumption of the social world was that it would be only a matter of time before everyone everywhere was sharing their realtime location with the world in the same way they were now sharing their every thought.
This assumption proved remarkably wrong. It seems that for all the technological capabilities of the major social platforms and their enormous success in breaking down societal norms around privacy and sharing, the one thing they’ve been unable to change is our intense desire for locative privacy.
We’re perfectly happy today sharing our most intimate thoughts with the world but allowing others to know where we were standing when we shared that thought is a line we aren’t yet ready to cross as a society.
In turn, the percentage of tweets that are geotagged has actually plunged from just over 3.5% seven years ago to 1.5% today. It seems we are more and more willing to share our daily lives with the world, but less and less willing to share our location.
It is unclear why location remains the final frontier for privacy and sharing online. A Twitter spokesperson confirmed the low density of geotagged tweets but declined to comment on why the platform believed so few were willing to share their location.
It turns out the picture gets even worse.
Shortly after introducing precision GPS-tagged tweets, Twitter rolled out “Places” in which privacy-consciousness users could share only a coarse bounding box representing their location rather than their precise GPS coordinates. Instead of broadcasting the location of their living room, users could tell the world only that they were tweeting from somewhere in San Francisco.
By mid-2012 Places had become the dominate way people expressed their location on Twitter, favoring its city-level coarseness over letting strangers know their home address.
Despite the popularity of Places, for some reason users still kept sharing their GPS coordinates. All of those tweets tagged as just “San Francisco” still for some reason included the user’s precise GPS coordinates in the tweet’s JSON record.
The reason for this was that Twitter was quietly recording users’ GPS coordinates into their tweets even when they requested that it share only their city-level location. As this practice and its privacy implications became more widely known, a company spokesperson confirmed that it reversed its policy in April 2015 and recorded GPS coordinates only for those users who explicitly authorized the publication of their precise location.
This led to a 70% decline in GPS-tagged tweets.
Of those remaining Place tweets, 98.7% specify only the user’s city, state or country. The total number of distinct locations on earth captured in Twitter’s geotagging data has collapsed.
Putting this all together, mapping tweets today faces a trifecta of challenges. The total number of tweets per day are decreasing, from a high of 500 million tweets a day in 2013 down to just 320 million tweets a day at the end of last October and falling steadily. As Twitter shrinks, the percentage of the remaining tweets that include verified geographic information has plunged as well, meaning a greater and greater loss of geography. Finally, of those ever-shrinking number of geotagged tweets, their geographic specificity is increasingly limited to the city level, no better than traditional news coverage, rather than the GPS coordinates that once ruled supreme on social.
It seems that the era of high precision mapping of social media using GPS and capturing the views of the entire planet’s citizenry on a live map was nothing more than an unrealizable fantasy, based on false assumptions and a privacy policy that was eventually changed. In its place, the geography of social today is no better (and in fact quite a bit worse) than the view we’ve long obtained from traditional sources like news media.
It turns out the tweeting public never meant to allow themselves to be mapped. The great GPS-tagged social revolution that ushered in the precision maps of global society that have captivated us for the past decade were the inadvertent result of a public that didn't fully understand how their location was being shared. As the graphs above tell in stark detail, we never wanted to be placed on a map.
In the end, perhaps it is time for us as data scientists to recognize that the era of precision mapping of social media is coming to a close and look to the future of how we might use the “big data” era to map a society in motion in a privacy-conscious and user-consented way.
