Along with other #citylis students I attended the British Library Labs Symposium (Monday 3rd November 2014) during reading week. As well as being a fantastic event it provided an opportunity to think about not just participating in but also preserving a conference backchannel.
What is a backchannel?
An event backchannel is a secondary discussion via electronic media that takes place whilst an event is happening. It can be an event where the audience is co-located, like a conference or classroom, or it the audience could be highly distributed, like the audience of a television broadcast. By using social media an event can be amplified and extended beyond the live audience reception: it can reverberate in time and space. This allows people in the audience to immediately discuss or reflect on the event, even those that don’t like speaking up in large groups, and also allows those not in the audience to follow the event.
Generally backchannels are increasingly seen as augmenting the event for the better though there are arguments against: it diverts the audience’s attention from the speaker and distracts from the experience of presenting and listening, it is potentially alienating and can disinhibit hecklers. Whether welcomed or not they have changed the boundaries and communication dynamics of events and whether spontaneous or encouraged by the event organisers Twitter is an increasingly popular backchannel medium (not so much Google Wave).
The backchannel at the BL Labs Symposium
There was some confusion at the start of the Symposium as to the correct hashtag to use for the event. This is one possible pitfall of trying to follow a backchannel if there isn’t an agreed convention. I was initially following the backchannel using the twitter hashtag #bl_labs as that was the tag I’d seen used by event organisers in advance. Some people were also tweeting on the alternate hashtag #bllabs, shorter though slightly less human readable. Consequently I switched my search query to #bl_labs OR #bllabs to follow both backchannels as one conversation. Alternative tactics included including both hashtags in tweets whilst discussing what hashtag should be used (standardisation from below) until the event organisers started including #bl_labs on the big screen (standardisation from above). It should also be noted that neither of these hashtags is an event specific tag: they are used more broadly to discuss the work of the British Library Labs not just the Symposium.
Additionally those of using attending added our #citylis hashtag to posts directly relevant to classmates not able to attend (i.e most of them) and so the search #citylis AND #bl_labs could be used to see our activity.
With all our DITA led tweeting practice our efforts did not go unnoticed.
In terms of my attention I did try to balance listening during presentations with tweeting whilst tweeting my thoughts during transitions and breaks to help amplify the event to those who couldn’t attend, especially as we knew that classmates would be interested.
Whilst backchannels amplify events, like any echo they can only reverberate for so long. Tweets are ephemeral passing through Twitter streams rapidly unless propagated by retweeting and disappearing from search results via Twitter’s user interface or API after 7 days though they can still be accessed individually via their permalink if known. If you want to capture the signal and preserve it for future reference you have to find a way to record it so you can consult it or analyse it in future.
Curating Backchannel Narratives
One way might be to curate the story of the event via backchannel content. You could do this by using blogging to write up events and linking or embedding associated backchannel content. Alternatively you could use a dedicated tool for this kind of curation. Storify is a good example of tool that allows people to create their own retelling of events using a variety of material from across the web and social media. It is a remixing tool that can create event mashups: a kind of mediated, annotated meta-backchannel from a particular perspective. Here is Storify’s own story of their stories from 2013 and my colleague Caitlin Moore’s curated archive of #citylis tweets from the BL Labs event. There is still a question mark of the longevity of these curated ‘documents’ and this article highlights possible pitfalls and alternatives.
Capturing Backchannel Datasets
An alternative is to download the raw data from Twitter into a dedicated dataset that you store either privately or could share publicly. This affords more preservation and analytical options but may be less accessible to people without programming expertise and knowledge of the Twitter API.
There are plenty of online twitter archiving and analytics platform that hide the magic behind software as a service (SaaS), such as Tweet Archivist, those these can require payment and there remains the question mark over what happens when the service is withdrawn, like with the demise of Twapper Keeper.
Alternatively programming libraries and cookbooks can make writing code to access Twitter APIs easier. Twython is a popular library for Python and chapters from the O’Reilly book Mining the Social Web have been made available as iPython Notebooks for Chapter 1 on Mining Twitter and Chapter 9 A Twitter Cookbook and on GitHub (with an a accompanying webcast). Whilst Python is a simple language and coding can be as straightforward as following a provided programming recipe it is still a fairly inaccessible method for general use.
Fortunately Martin Hawksey created an alternative method to use that swiss army knife of data analysis, the spreadsheet (specifically Google Spreadsheets), to archive and and perform basic analysis on twitter search results. The solution is known as TAGS standing for Twitter Archiving Google Sheet. There is an additional web based visualisation tool, TAGSExplorer, for further graphing of Twitter archives stored in TAGS. A full example of functionality is provided using #jiscel12 tweets. Whilst you still need to have created a Twitter Application ro authenticate with Twitter and have a Google Account to use Google Spreadsheets the TAGS approach is one of the most straightforward available for archiving and exploring a Twitter dataset. Additionally the spreadsheet can be configured to check periodically for updates meaning the archive can extend beyond the Twitter 7 day search limit if created in advance.
BL Labs TAGGED
I created a TAGS for #bl_labs during the DITA Labs session that introduced us to TAGS on 20.10.2014 a week before the event and it’s now just about a week after the event. Whilst tweets are already starting to disappear from the web user interface of Twitter for this hashtag I have an archive of 1159 Tweets from this period with a noticeable spike in activity during the Symposium.
Simple Archive Analysis
The TAGS spreadsheet provides each tweet listed in an Archive tab with a Summary tab and a Dashboard tab that provide some simple analytics of the kind that are easy to do with spreadsheet functions. No surprise that James Baker in the BL Digital Research team is the top tweeter with 58 but our very own Caitlin Moore was 3rd with 52 tweets.
As it’s a spreadsheet you can use formula to add your own calculations on the data. There were 270 tweeters in this period but the top 30 were responsible for 60% of the tweets indicating quite a long tail for the backchannel (that echo reverberated). Also despite our activity the #citylis hashtag only appeared in 12% of the total tweets with four of us in the top four tweeters indicating there was lots of other activity going on.
These is fairly simple, standard, and long standing, but useful number crunching that can be done once you have a TAGS. However, if you publish the source spreadsheet so that it is publicly available, Martin has also been developing some web visualisation tools to provide alternative user interfaces to the archived dataset.
The more straightforward is a simple web view of the Twitter Archive that replicates some of the look and feel of the Twitter stream along with some filtering options.
Mapping the Archive Network
A more complex visualisation is the TAGSExplorer. This represents the archive as a network graph allowing fuller interaction with the data. There are graphs of the top tweeters and the top hashtags and the searchable archive is also available from this interface. Of most interest though is the way this visualisation maps conversation through the archive. Conversations, I think, are treated as when another @tweeter is mentioned in a tweet whether it be a mention or included in a retweet.
It provides four mechanisms to do this:
- A graph of the top conversationalists measured by degree (connections with other tweeters)
- An interactive map of the connections between tweeters in the archive. A lot of fun can be head pulling this around.
- If you click on a tweeter in the map a summary archive of their tweets, replies and mentions is displayed.
- A further visualisation allows you to then replay that tweeter’s conversations via an interactive timeline so you can watch a conversation grow
I didn’t really appreciate this visualisation initially but this week I’ve also again been studying the University of Southampton’s Web Science Mooc on FutureLearn which provided some useful insight into network properties and analytics.
In network analysis a graph is a mathematical representation of a network. It contains nodes, in this case tweeters, and connections or edges between these nodes, in this case mentions or retweets. This graph is directional: just because a tweeter mentions or retweets someone it doesn’t mean it is reciprocated.
A degree is the number of edges that connect to a node. So in TAGSExplorer each mention or retweet is a directional edge. A higher number of edges in the network would indicate more connections and conversation between tweeters.
Analysing the Archive Network
The network also exhibits properties of a scale free network. In these types of network there are a large number of nodes with few connections and a few hub nodes with many connections. Tweeters like James Baker, Caitlin Moore and Ben O’Steen are hubs in the BL Labs archive network. As more people join the back channel and add to the network preferential attachment means they are more likely to notice and therefore mention the hubs making them even more influential.
Social Network analysis allows us to think, albeit simplistically, about influence in the network. Influence could derive from three potential sources:
- Position (their role in the network)
- Propagation (retweeting others to amplify their tweet)
- Participation (replying to others and joining in conversations)
Based on this tweeters like James and Ben, both of whom work for the BL, have influence because of both their position in organising the event and their activity in retweeting and replying to others. Some like Caitlin and DigiVictorian were influential because of their activity.
Overall the large number of tweeters in the graph, but the lack of connections between all but a few of them suggest that for this event the backchannel was maybe used more for propagation than for discussion to draw a simple conclusion from the dataset.
- Whilst going to post this I noticed my classmate David has written a similar post on his experiences at the Symposium and analysing #bl_labs with TAGSExplorer
My rough and ready notes in Evernote. These are unedited: all errors are mine.