The Digital Flames of Ferguson
How the nation’s attention turned to events at a St. Louis suburb
The recent tensions in Ferguson, Missouri has left the country deeply disturbed. Many of us have strong opinions about what went wrong. For others, the root grievance is more about misguided policies in impoverished communities for years. The Ferguson protests surface a deluge of sensitive issues, including race, class, disdain, militarization of police and even net neutrality.
What made the Ferguson protests become national news? In this post, I try to discern how the country’s attention gradually turned to the events transpiring in Ferguson, what were the driving factors and what it tells us about certain latent kinetics of the Social Web.
The data indicates there is specific evidence of persistent attention in certain cities and an ensuing social contagion effect, that possibly swung national attention to Ferguson.
For those of us without ground experience, viewpoint on Ferguson is largely influenced by how, when and what information we receive. Relevant information captures our attention. Therefore, to examine the information landscape with regard to Ferguson, it is key to first understand the attention dynamics of netizens.
But before delving into data, let’s recollect the fragmented media experience associated to Ferguson. Generally, the chain of digitization of a real-world news event takes the following sequence: social streams → social web news media →mainstream media → TV channels. This pattern wasn’t to be broken for Ferguson either.
Mainstream media had mostly ignored the initial phase of the protest. Due to the lack of coverage, activists turned to social media to raise awareness. Twitter, as usual, led the way as a tool for people to express themselves.
Starting August 11th evening, looting and riots were being reported on social media . ‘New Media’, including Buzzfeed, Vice etc. posted articles about the Ferguson riots as the looting began. Reddit Live had a great real-time chronological feed. Kmov TV did a commendable job of streaming events on the ground for online viewers. However, three nationwide mainstream media channels still hadn’t reported on Ferguson.
The story was finally picked up by mainstream media quite late into Monday night. People complained that news channels conveniently ignored the peaceful part of the protests, but jumped into covering the riots and looting once events led to destruction of American businesses.
Facts are born the moment a real world event is digitized. News may take a longer incubation period; probably because media organizations have different thresholds for fact checking/adjustments before they can broadcast. Nevertheless, as information reaches us via media channels, we start getting hooked to what’s happening. This prized resource of ours called Attention, is stimulated by receiving or being exposed to the information. The kinetics of this whole process can consolidate or decimate the attention of the masses.
What are some critical forces in this kinetics, that stimulate us into attention? First, we must have inbound channels that carry recent news. If I don’t watch TV or swipe apps on my phone, my only inbound information channel is perhaps word of mouth. Further, we all possess inherent affinities to certain categories of news. #MH370 vs. #Fracking — were the masses equally interested?
To expand a similar analysis on a national scale, we need to monitor the attention pulse at various locations across the country. There is hardly a better way to investigate such social phenomenon than by studying geo-temporal trend signals.
On Twitter, we receive all sorts of information streams. But even if people we follow are not tweeting about some news-worthy event, Twitter trending topics turns out to be a terrific way to expose yourself to issues that have captured Twittersphere’s attention. Twitter trending topics represents the attention landscape in the Twitter world. In fact, geo-located trends is a strong signal indicating the attention paid by users to some topic in that location.
The embedded Attention Signal
The ‘Trending Topic List’ (TTL) is the result of an algorithmic technique which detects sudden appearance of high frequency of words in tweets. For locations defined by Twitter, the TTL comprises of a group of 10 words that have highest frequency (tf-idf) of occurrence in tweets from that particular geolocation. We collected such TTLs across all cities, sampled in 5 min. intervals.
The TTL changes recurrently, depending upon what (and how rapidly) users in the location are tweeting. This rate of change is interesting — because it hints at the persistence of attention on some topic among users in that location.
In fact, more often the location TTL changes, greater is the volatility of attention in the location. Imagine a location’s volatility as an indicator for how fast it is pulsating. If you throw in a topic, it would be difficult for the topic to stick (persist/ keep trending in TTL) if the volatility is high.
On the other hand, a low volatility indicates one of two situations: (1) there is persistent attention on few topics from the location. For example, during the #KONY2012 campaign two years back, volatility across several major cities was significantly low, as the campaign captured people’s attention in all major cities across the United States. (2) The other possibility is that there isn’t enough consolidated topic hashtags within the tweets from the location. This pattern has been observed in previous protests, like the Arab Spring and Occupy Wall Street.
So, the volatility of different cities is a manifestation of their inherent variation of attention. I mapped the volatility of several US cities normalized over the time period of the Ferguson protests. The top 3 cities which are most volatile include New York (2.79), Los Angeles (1.82) and Philadelphia (1.59). Three cities with lowest volatility are Las Vegas (0.37), Memphis (0.33) and Baton Rouge (0.227).
Now, why is it important to measure the volatility of cities? It’s because volatility represents the inherent attention fluctuation in the location. And it is pre-requisite to understand this inherent fluctuation before studying Ferguson-related attention specifically.
Imagine a city’s attention like a moving carousel and an impending trend is someone trying to get aboard. The more the volatility, the faster the carousel is spinning. If you are an impending trend, it is much easier to jump into a slowly spinning (less volatile) merry-go-round of attention and stick to it, than it is to get on and tag along on a carousel which is spinning very fast (more volatile).
Now, back to Ferguson…
Ferguson-related Twitter trends
I sampled 37 Ferguson-related trending topics that appeared in the TTL of 32 major US cities. Examples of such trending topics include ‘#Ferguson’, ‘National Guard’, ‘#justiceformikebrown’ etc. Then, I analyzed when and how long these topics trended in each city. I inspected their origin, where it spread, how rapidly it spread etc.; monitoring all this in coordination with events that transpired on the ground in Ferguson.
* Signal Strength of Ferguson-related trends
Since the data comprises of the TTL instances collected every 5 mins between Aug. 9th and Aug. 19th, 2014, I inspected the number of Ferguson-related trends in every TTL and aggregated them hourly. Thus, for every hour, I could calculate a Ferguson-related signal value for the location. The more Ferguson-related trends occur in the TTL of the location, the higher will be the amplitude of the signal. Over time, this represents the attention signal of the city towards Ferguson-related trends.
This attention signal is distinctively different for different cities. The figure on the left shows the attention in New York over time. We see the Ferguson-related trends first appeared in New York on Aug. 10 at noon, around the time when the first peaceful protests were being organized in Ferguson. Notice the tall spikes on certain other days. These points represent attention paid in New York to major developments in events unfolding in Ferguson.
Now, compare this with the attention signal in St. Louis itself (shown below).
First observation in St. Louis is presence of many more (and frequent) spikes. The signal amplitude never reaches zero. Attention not only keeps spiking intermittently, it is maintained at a higher average level. This is expected, since St. Louis is the epicenter of the news.
Also, notice the signal originates on Aug. 9th in St. Louis, one day prior to the major protests and riots and the day of the actual shooting. However, the amplitude of the signal is comparatively lower on Aug. 9th compared to the next few days. It seems less tweeters in the area related with ‘#justiceformikebrown’ (which trended earlier) than ‘#Ferguson’. The latter trended when the riots started in the final hours of Aug. 11, and led to the largest spike. Alternately, it was possible that people were using one consolidated hashtag before and many afterwards.
Here are the attention signals for some other cities…
Since all these signals are essentially time series data, let us look at the average level of signal strength maintained at a particular location. Not surprisingly, St. Louis possesses the strongest attention signal strength (31.8). In fact, the next five cities with highest attention signal strength are comparatively much lower, and include Atlanta (9.15), Las Vegas (8.57), Washington DC (6.38), Indianapolis (6.17) and San Francisco (5.86). You can check out the interactive visualization for exact values for other location.
* Which cities get the most credit for maintaining focus on Ferguson?
So, cities have inherent volatility. They also have attention signal strength about a particular topic (here Ferguson). I wanted to identify which cities maintained their focus on Ferguson-related trends, in spite of their inherent tendency to shift attention quickly. In other words, which cities can sustain their attention on Ferguson in the face of several competing trends. This is a typical case of measuring the signal-to-noise ratio (SNR), where volatility is the background noise trying to disrupt the signal we want to detect (in this case, attention on Ferguson).
One would expect St. Louis to have the highest SNR, being the epicenter. This is what we find. Outside the epicenter, New York (12.45) is distinctive in sustaining its attention on Ferguson-related trends in the face of contesting non-Ferguson trends, followed by Atlanta (8.55) , Washington D.C. (7.69), Chicago (7.66) and Miami (5.44).
Maintaining attention in the face of competing topics is a durable quality. But that is just one part of the puzzle….
The other part of the puzzle, is how quickly did the cities become aware of what was transpiring in Ferguson. In other words, how sensitive were they do Ferguson-related trends?
* Contagion of Ferguson-related Trends
Geographical information diffusion on Twitter is driven by the social interconnection structure among users in various cities and the sensitivity of a city’s tweeting population to new topics. Complex contagion is the study of social networks as conduits for ‘infectious’ idea/topic transmission. The simplest way to study contagious information spread is to examine when the various cities got infected with a Ferguson-related trend.
The timeline below shows when a city was first exposed to a Ferguson-related trend, and how long the trend had to persist there before a new city in the system got exposed/infected.
The data shows that on Aug. 10th, Ferguson-related trends were observed in four cities, namely the epicenter St. Louis itself, Miami, Boston and New York. The time between subsequent infections in every case was more than 3 hours .
Past midnight on Aug. 11th, Ferguson-related trends had persisted in New York for about 15 hours. Then something interesting happens. The trend first appears in Washington DC. It trends in Washington D.C for about 70 minutes, before erupting into national trend — infecting 12 cities in the next 3 hours! This is a remarkable escalation in dispersion, given Ferguson-related topics trended in only 4 cities in the previous 53.4 hours. Following the topic’s persistence in New York and inception at Washington DC , 62% of the remaining cities get infected in the next 5 hours.
This behavior in ancillary with what had been happening on the ground in Ferguson, Missouri around midnight on Aug 11th. A QuickTrip was burned, there was looting and burglaries being reported.
Such points in a social phenomenon’s space-time, like the one observed in Washington DC just after midnight, are called ‘tipping point’ or ‘critical mass point’. At the tipping point, a huge fraction of the cities adopt a previously rare trending topic over a drastically short period of time.
Contagion tipping points can be possibly attributed to three causes:
(1) The active users (who were tweeting about Ferguson) in New York have a stronger network-community/connectivity with users in several of the subsequent infected cities; much more than Boston, Miami or St. Louis do. Thus, the two cities combined were sufficient to spread the news to a critical mass of interested Twitter users in other cities.
The next two possibilities eradicate a specific location’s impact in generating critical mass. (2) There could be a minimum number of cities (and in turn, no. of people) that need to be infected before a trend goes becomes national, i.e. there is an inherent tipping point (based on number of cities infected, irrespective of actual locations). (3) There is an exogenous agent that caused the contagion, such as a news story run by TV media etc.
The elementary reason why such social contagions and tipping points are decisive in information dissemination is that it draws national attention to a trending topic quickly, and in many cases, can be the difference between the news going national or not. A possible hypothesis of why this phenomenon emerges in the Ferguson scenario is that several journalists/media are positioned in New York and the city might be a critical node en-route to a national trend. Sustained attention in New York (~ 15 hours) could be a key factor in making this “trend” tip.
* Not all trends initiate contagions
As a contrasting example, I decided to analyze the geo-attention signal around the Zimmerman trial (trending topics: #Zimmerman, #TravyonMartin etc.). Recall the results of the verdict was disclosed by the mainstream media around 10 PM on July 13, 2013.
Our data shows Zimmerman-related trends were observed in only four cities after the verdict (on July 14th). It was not observed nationally across the major cities afterwards either.
In New York, Zimmerman-related trends were observed after the verdict for about 3 hours, during a protest held near Union Square (this is probably much lesser than the critical trending duration needed for contagion). There were no tipping points. As a consequence, we would predict that the volume of Twitter conversations around the #TravonMartin case would be significantly lower than Ferguson. And that is precisely what a Pew Research study finds.
This also hints at interesting possibility, that the latent kinetics of contagions might be fundamentally different when initiated or originally driven via social media vs. mainstream media channels.
With respect to Ferguson, we see two key developments that lead to a tipping point. First, New York’s incredible ability to sustain attention on the topic, in spite of several competing trends trying to gain attention (recall it is the most volatile city). Second, the tweeting population of New York reaches a critical mass of users in other cities, enabling the kinetics to infect them with news about Ferguson.
What does this all mean…
As David Carr explains in the Media Equation column of the NY Times, while much of mainstream media leaves communities of color unmoved —Twitter becomes the place many users rely on for information. Among other things, it is the lack of information at the right time and the frame of choice that bemuses our opinion. Whichever information carrier we embrace, it is important to remember that the chosen media channel has the power to shape opinion by governing how news reaches us.
Trending topics can have varied intensity of reception in the Twitter world, coupled with a myriad of reactions. In the Ferguson case, there is evidence that cities possess unequal capacity in sustaining their attention on the issue. We also notice the emergence of contagion effects for Ferguson-related topics within social media channels.
Past the tipping point, every node gets affected and it is almost impossible to remain oblivious to news about Ferguson for most parts of the network. With the optimal force, contagion effects and tipping points emerge in social media channels. And just like in the much-needed case of Ferguson, it can quickly disseminate valuable information to connected users and raise awareness by capturing national attention.
I would love to hear your thoughts/comments. Find me at : @_RoySD or suman@betaworks.com. All the above analysis is in an iPython notebook, which I am happy to share with dataphiles.
Thanks to Gilad and Brian for helpful comments in drafting this post.
