How Occupy & the Tea Party use words on Twitter

This is a guest post by two Masters students in comp.social, Abraham Doris-Down and Joshua Teal. It’s an exploratory piece of work based on a class project.

It is election season again. As news about the Republican primaries and President Obama’s approval ratings dominate the airwaves, two relatively new political movements, the Tea Party and Occupy Wall Street, are inciting a kind of political activism that has scarcely been seen in recent memory. Many people claim to speak for each movement, claiming to know what proponents want; but in truth, these groups are both new and have evolving ideologies.

Rather than rely on the rhetoric of often self-proclaimed leaders of these movements, we can instead listen directly to the discussions of the movements’ followers. With this in mind, we rolled up our sleeves (metaphorically - Georgia Tech has yet to master consistent indoor climate control), pulled out some laptops, sudo-ed some libraries, and captured Twitter timelines to find out what members themselves actually talk about, and if the groups are as different as they and the media would have us believe.

Our Approach

Since we were not interested in what the Michelle Bachmans, Elizabeth Warrens, Keith Obermans, or Glenn Becks of the world were saying about these movements, we needed a way to find tweets of Occupy and Tea Party followers. To do this we researched news sites, blogs, and political sites to find 3 Twitter feeds for each movement that we felt were representative of that movement. Then we grabbed tweets from followers of each of those feeds.

The Twitter feeds we chose were not meant to be exhaustive, but representative. We felt it was important to focus on the followers of official and unofficial spokespeople of the respective movements, rather than the spokespeople themselves to ensure that the tweets we gathered came from genuine unconstrained conversations rather than from what each organization approved as its official message. A list of the accounts can be found below.

Once we had our sample, we collected only the latest tweet from followers who had completely public timelines, and we spaced out the tweet collection over several days between 12/10/11 and 12/15/11. The result was a corpus of 15,000 tweets from 15,000 different people (7,500 per group).

The following steps were taken to do an initial analysis of the data:

1) Word-counting/phrase-counting software was used to generate a list of words or phrases (excluding common and noise words and phrases like “http”, rt”,”via”,“com”) from each Tweet corpus which appeared 40 or more times. 45 word fit this criteria for the occupy movement and 77 for the tea party movements.

2) From that list of words and phrases, we determined a set of 17 relevant words which appeared 40 or more times in both corpora.

What we got

We performed a chi-square test on the 17 words to determine differences between each group of followers. Below is an example.

* We assumed that each term would be used only once per Tweet, so while “Christmas” would have a count of 55, not-“Christmas” would have a count of (7,500 – “Christmas”).

We also preformed a chi-square test on the number of unique words mentioned over 40 times that the occupy group tweeted versus the number of unique words mentioned over 40 times that the tea party group tweeted.

Pictured here is a diagram revealing a surprising finding: while there were approximately 24,000 total words in the combined corpus of 15,000 Tweets, only 4,000 of those words were used by both the Occupy Wall Street followers and the Tea Party followers; each group used 10,000 words unique to them.

__

So are they different?

In a nutshell: yes they are different.

70% of 17 the words we examined showed a significant difference in occurrence between the two groups. It is interesting to note that all but one of the words that showed no significant difference in occurrence between the two groups are typically used as verbs; “love”, “support”, “think”, and “read”. The only noun specific word with a non-statistical difference between groups was “world”. If we remove the verbs and focus on the nouns then 92% of the words show a significant difference between groups.

Since nouns generally relate to topics, our results show that the Occupy Movement and Tea Party have significantly different topics that dominate their conversations. This suggests that the Tea Party and Occupy Movement do have different motivations and interests. Since both are largely political movements it would suggest that they are both trying to bring out different changes in the United States’ political landscape. However, just because they appear to be different, we cannot assume, based on our data, that they are in opposition.

Other cool stuff

There is a significant difference (p-value = .00255) between the numbers of words that Tea Party followers mentioned internally more than 40 times, and the number of words that Occupy Movement followers mention more than 40 times (77 for the Tea Party and 45 for Occupy). Since the Tea Party followers used the same words more often, we assume that they are all talking about the same topics or at least repeating the same phrases. This could suggest either that the Tea Party is either more ideologically organized, promoting a shared lexicon and phraseology, or that the tea party members have more homogeneous interests then members of the Occupy movement.

We noticed that the Occupy follower tweets contained a fair number of tweets in other languages compared to the Tea Party tweets. This could suggest that the Occupy movement has more interest internationally, or at least of people who speak other languages. In future research grabbing the geo tag information form the followers’ tweets would help shed light on these possibilities.

We noted that while OWS (Occupy Wall Street) and Occupy both received more than 40 mentions (57 times and 61 times respectively) in the Tea Party corpus, no words that directly related to the Tea Party were mentioned more than 40 times in the Occupy corpus. This could suggest that while the Tea Party is interested in the Occupy Movement and what they are doing, the Occupy Movement is not interested in the Tea Party.

The name of four prominent Republican presidential candidates and the word “debate” were all mentioned more than 40 times. This suggests that the Tea Party followers are interested in the Republican primary debates, and most interested in these 4 candidates. While the data itself does not suggest whether or not the Tea Party has positive feelings about the Republican candidates it does reinforce the cultural perception that the Tea Party is closely related to the Republican Party.

*Word cloud images from Tagcrowd.com

Selected Twitter Feeds

Tea Party Movement

Tea Party Patriots
Followers: 13,700
http://www.teapartypatriots.org/
http://twitter.com/tppatriots?_twitter_noscript=1

Tea Party.org
Followers: 3,700
http://www.teaparty.org/
http://twitter.com/teapartyorg?_twitter_noscript=1

Glenn Beck
Followers: 441,900
http://www.glennbeck.com/
http://twitter.com/glennbeck

Occupy Movement

Occupy Wall Street.org
Followers 136,400
http://occupywallst.org/
https://twitter.com/OccupyWallSt

Occupy Marines
Followers: 9,000
http://occupymarines.org/
http://twitter.com/occupymarines

Michael Moore
Followers: 928,000
http://www.michaelmoore.com/
http://twitter.com/mmflint

Why designing a carpool site is so hard

For the last 8 months, I’ve used MARTA to get to Georgia Tech. Let’s just say that it’s not the best-respected mass transit system and leave it at that. Still, I can get to work in 35 minutes and spend my time reading papers rather than fighting traffic. I don’t have to own a second car just for commuting.

But I would much rather carpool. It’s cheaper, faster, cleaner and safer. So when Georgia Tech announced their new arrangement with the ride-sharing site Zimride, I was the fourth employee to register. Carpooling is a very hard social site to get right.

Schedules. A commuting schedule is at the heart of any carpool site. If I don’t tell you when I want I a ride, how can you pick me up?

But I don’t want everybody I work with to know exactly when I arrive and leave everyday. I like to see my daughter before she goes to bed at night, and here Zimride makes me write it down. (Tenure committee—if you’re reading this—this is only an example. I never leave.)

I took my first Zimride a few days ago, and five minutes into it my driver explained how they often run late to take their kids to school in the morning. It’s not reflected on the Zimride schedule. It could be. The interface permits it. But people usually don’t want to be so literal with their schedules, especially when their colleagues can see them—something backed up by classic CSCW research.

Incentives. Let’s back up. Why should I do this at all? I don’t have a commuter car, so I stand to gain a lot from carpooling. But for somebody who could simply drive alone, why bother? My driver adds 7 minutes to their commute because I’m along for the ride. Zimride thinks it’s for the money.

When I tried to work out the money part, my driver shrugged it off. Honestly, they probably make quite a bit more money than me. Plus, it’s just plain awkward to hand over money in a situation like this. Sure, it’s great to offset your gas costs, but not at the social cost of ruining what would otherwise be seen as a good deed.

Identity. By partnering with Georgia Tech, Zimride forces you to authenticate as an employee or student at Tech. That’s a start. They also encourage you to use your Facebook identity on Zimride. I think this is a great move, and I would rather Zimride make it mandatory. If I’m going to get in your car, I should be able to see your Facebook profile. In fact, I should get to see much more of it than some random internet user. I want to learn what you care about, how snarky you seem in your posts, and whether you seem angry, right-wing, or hopelessly hippy. I’m about to put myself in a small, confined space with you: these things matter. But this isn’t Zimride’s fault. There’s no “getting to know each other” setting on Facebook.

p.s. Hat tip to Amy Bruckman for putting this topic in my head last week.

Social Translucence

This semester I taught one of my favorite papers, Social Translucence by Tom Erickson & Wendy Kellogg. It’s more than 10 years old—pretty old in this field. I’m writing about it now because I think it can still do lots of good.

“Socially translucent systems [make it] easier for users to carry on coherent discussions; to observe and imitate others’ actions; to engage in peer pressure; to create, notice, and conform to social conventions.”

Erickson & Kellogg argue that we should look to what makes communication work in the real world when designing social media. For example, what makes a crowded city street work? New to the neighborhood and want to find a good restaurant? Follow the people who seem to know what they’re doing. Why don’t people throw one another to the ground when the train door is about to close? Because I know the rules, you know I know the rules and everyone around us is watching. The authors call it visibility, awareness and accountability.

A lot of this has been implemented in popular social media. But not everywhere, and this is where I think practitioners and founders will find it useful. Shine social translucence in some places and you’ll often be surprised. Take email. Did you get my message? Do you have time to read it? Am I bothering you? I know it’s my vacation pictures … oh you immediately hit the delete button? … oh. It’s ten years later, but the theory still provides ideas for building new stuff—things to imitate what we accomplish so effortlessly IRL.

The internet has always been social

I do research on social media (aka social computing). I look at ways we might build social media differently and what we can learn from massive datasets online. When I tell people what I do, many well-meaning people say “that’s a hot topic!” Inwardly, I always grimace a little. “That’s hot” has a hint of faddishness to it. Fads go away.

For most people I meet, the phrase “social media” evokes post-2004 web technologies: Facebook, Flickr, Twitter, Foursquare, … (The phrase came into widespread use around this time.) But since its earliest days, the internet has been a very social place. I love the following quote from a SIGCOMM review of ARPAnet, a precursor of the internet: 

“The initial goals in creating the ARPAnet were to promote more effective use of geographically dispersed computing facilities, and to develop the underlying packet switching technology that might later support advanced military information networks … A new use emerged, however, with the realization that the network could also be exploited as a medium for intercommunication among its human users … Network message service was an immediate success. Message flow grew in volume to become the most visible (if not the heaviest) traffic on the network.” (Henderson & Myer 1977) 

It’s easy to forget about little old email, perhaps the internet’s first social medium. Today’s status updates and location-based services recall the UNIX command finger. The many-to-many, instant-chat idea behind Google Wave made me think of the talk commands, dating back to the 1970s. If you give us something we can use to socialize, that’s what we do. We tell stories. We check each other out. That’s not going anywhere.

comp.social lab is founded

I’m Eric Gilbert, a new Assistant Professor in Georgia Tech’s School of Interactive Computing. I work on building and studying social media, and my new research lab is called comp.social.

I’m looking for bright students who want to build fun, daring social media systems or analyze them in surprising ways. (Or both.) In the past, I linked emotion expressed online to the stock market. I also built an experimental application called We Meddle. These are examples of what I want to do in comp.social, although we’ll swing bigger now. You can learn more by visiting the comp.social site. (My email is also posted there).

p.s. The name “comp.social” refers to the structure of Usenet. Social media may seem new, but it’s as old as the internet itself.