Written by: Spencer Greenhalgh
Primary Source: Spencer Greenhalgh
My first introduction to Twitter was in a class on the intersection of technology and Mormonism that I took from David Wiley at Brigham Young University. During the class, David encouraged us to try experiencing the sessions of the upcoming semi-annual LDS General Conference in a new way: by following the #ldsconf hashtag. My very first tweet—as seen below—is evidence that I gave this a try, and even though I now spend more time tweeting about inter-rater reliability than religion, there’s always been a connection between my faith and my favorite social networking site.
Signed up for Twitter especially for #ldsconf. :-)
— Spencer Greenhalgh (@spgreenhalgh) April 3, 2010
About a year ago, when I first started learning about tracking and collecting tweets, I started to wonder if it would be feasible to collect all of the #ldsconf tweets associated with a General Conference for scholarly purposes (also, for funsies). Trying to collect these tweets taught me a number of valuable lessons about Twitter collection: For example, I became a TAGS fanboy when I realized the tracker I’d previously been using had certain collection limits that made it less-than-ideal for collecting tweets during high periods of traffic. Eight months later, shortly after the next Conference, I learned that if you let a TAGS collect for too long, it will overwhelm the Google Sheet and lose all of your data.
Well, the third time is the charm, and I finally managed to collect a reasonably-full set of #ldsconf data this past April. There’s a lot of work yet to do, ranging from big picture questions of how to frame this activity (informal learning in an affinity space?) to more granular things like coding tweets and Twitter profiles (one of the most interesting thing about the #ldsconf hashtag is that it’s used by everyone ranging from official LDS institutions to critics of the church) to even basic things like properly cleaning and standardizing my data (I’m not 100% sure that I’m using the same data for all of the plots below…). Those take time that I don’t currently have, but I did want to do a little bit of tinkering just to play with the data and see what’s going on here.
First, some basic descriptive statistics. The #ldsconf hashtag is used year-round, but I’m looking at data from March 19th, 2016 to April 10th, 2016 (I’m defining midnight in terms of MDT because Utah). This time period starts a week before the first session of the April General Conference and ends a week after the last session of the Conference. During this time,
- 21,536 different Twitter users posted a total of
- 24,958 original tweets and
- 63,134 retweets.
That’s quite a bit! If that’s evenly distributed across days, it represents over 1,000 original tweets per day and over 2,700 retweets per day. It’s not likely to be evenly distributed, though, so let’s take a look at activity plotted on a timeline. Below is a plot of four measures of activity (users, new users, tweets, and retweets) for every hour during the entire three-ish weeks I’m looking at:
By the way, please ignore the “by day” in the legend and mentally replace with “by hour.” I’m reusing code here, I forgot to make a change, and I’m too lazy to fix it for a blog post.
The three big spikes correspond with the three days that Conference sessions took place, and if you look close enough, you can even divide the last two spikes into subspikes based on individual sessions (three on Saturday, two on Sunday). The early spike—corresponding to the General Women’s Meeting—also seems to divide into subspikes, and we’ll see why in a bit.
To get a clearer idea of what this activity looks like, I zoomed in on Saturday, April 2nd, and Sunday, April 3rd, when five of the six sessions take place. Now the session spikes are really visible, but we can also see some distinctions between different types of activity. For example, we see more retweeting than tweeting original posts; we also see spikes in new users with every session, which wasn’t what I expected, so it’s either something exciting to look into or a problem with my code.
As seen below, the General Women’s Session is pretty similar and even has two clear spikes with a total absence of activity in the middle. Unfortunately, the reasons for this aren’t that exciting. After checking my data, it looks like I’m missing an entire hour during that time, which is a bummer. I know there was a lot of excitement about the announcement of the I Was a Stranger refugee initiative announced during this General Women’s Session, so maybe that broke Twitter (or at least TAGS)? If that’s the case, it’s a real shame, since that would be one of the most interesting things to study from Twitter activity during this last Conference.
Two last charts before I get back to work. First, a map of where these tweets are coming from. That’s not entirely the right way to put it: Each point on this map corresponds with the listed location of a Twitter profile who tweeted or retweeted at least one post with the #ldsconf hashtag. I have no way of knowing whether they were actually in that location during the Conference, especially since it’s not uncommon for people to travel to Salt Lake City for the occasion.
So, I was expecting heavy concentration in the Western US, which I got, but there’s actually not that many more dots in the West than in the East, which I (as a Mormon who grew up “out East” in Kentucky) appreciate. Then again, not sure if dots are overlapping each other in this plot or not (need to check documentation for the functions I used), so more questions to be asked before making any bold interpretations.
A look at the whole world continues to challenge the “Utah and Utah only” expectations that even I had going into this project. My method for identifying the location associated with a Twitter profile is far from foolproof, so before looking too much into Mormon Twitter use in Algeria, I’d like to see if the Google Maps API I’m using is misinterpreting anything, but it’s still neat to see people from all over the world using this hashtag.
So, that’s about it for now. I’m sure I’ll be back to this data soon!
Latest posts by Spencer Greenhalgh (see all)
- Public data and digital research ethics - September 11, 2017
- Using notebooks for beginning-of-semester planning - September 5, 2017
- A couple of podcasts on screencasting - August 23, 2017