Today is the third and final day of the Lancaster Twitter & Microblogging conference. For more on the event see my Day One and Day Two liveblogs. Today there are only a few sessions over a half day so this will be a rather shorter post.
Firstly it’s off to:
Cracking the code: Towards a semiotic understanding of Twitter and its use by media fans by Rhiannon Bury
Use of Weibo by UK universities and Chinese students: A study of intercultural micro-blogging by Nick Pearce, Durham University (and Yimei Zhu, University of Manchester)
Nick teaches a lot of international students and works with recruitment areas of the university as part of his teaching work around social anthropology. It occured to me that we could be using Weibo to market what we do, but also interested in that idea of engaging with a social media in a language you don’t speak but where you can sort of see what’s going on. I have been working with Yimei Zhu and she does speak Mandarin so whilst I talk about not understanding what is being said, she does and her research is very much around analysing exchange on Twitter.
So today I’m going to talk about this in the context of UK Universities and some exploratory work.
UK Universities and social media – the focus here is on marketing/recruitment. It’s a cheap/low cost means of interacting with multiple audiences. So when I came to Durham I set up a Facebook page for our courses and we get maybe an enquiry a week through Facebook. Not sure if they might have contacted us another way but they are able to. And similar idea behind our Weibo page. Although this would be an institution-wide Weibo page. That multiple audiences aspect does matter though. You can see a top five chart of Universities on Twitter – ranked by followers and retweets (http://craig-russell.co.uk/demos/uk_uni_twitter/index.html). Not a surprise that OU is biggest – it’s the biggest university in the UK so that makes sense. They have an active Facebook page as well – although that was not all good news. You might get 1 in 10 bad comments, but 9 out of ten challenging that comment. You lose some of that control but your “product” (though I hate that word) gets out there. The chart of followers drops off fast… For Durham our Twitter profile is busier than our Facebook page.
There are issues in follower numbers. There is the issue of fake/zombie followers (zombie is a Weibo term but it’s a lovely term). There was a hoohah when Yale joined Weibo, and they got a huge number of followers instantly. But analysis showed that there were only a few hundred followers that were actually active. There’s a suggestion that Weibo may have had something to do with those fake followers, Yale certainly didn’t pay for them. But it’s easy to gather raw figures but it’s not really a good measure. For me Facebook comments and enquiries count far more than “likes”. Even more so for the OU perhaps. Interactions matter here. There are alternaive analytics – e.g. Klout, PeerIndex… but not much better than raw figures. It’s important to be careful and critical of numbers.
So what about China. Chinese students are a significant part of international cohort – 79k in 2012. Facebook/Twitter restricted in China (you can access them but only through dubious means, but they do get use). What makes that interesting is that those using Facebook in China tend to be more political, to go to that effort you need that. Weibo was set up in August 2009, has over 360 million users, it has restrictions but no one is being forced to use it. It is clearly popular and clearly became popular very quickly… regardless of whether we think many of those users are fake/zombies.
So looking at my Weibo profile it is quite Twitter-like. I follow people… but I’m not sure how I came to follow them. You can post in English here. And you have animated gifs – which I’d love to see in Twitter! But interface all in Mandarin. My colleague speaks Mandarin but uses no social media. I use social media all the time but don’t speak Mandarin… when she asked how to log out I found it faster – that’s a reflection of how like other sites Weibo looks. Other big similarities here. There are assymetrical follower relations (like Twitter, unlike Facebook) and there is 140 character limit. But in Mandarin that’s a lot! Ai Wei Wei says that that’s a short story. But some differences here. There is ID verification – it is supposed to be your passport number but there’s a suggestion that that’s not a rigorous checking process. Censorship is rather opaque, you only really encounter it by gaps and absences. But remember that Twitter is moderated – you get taken down or reported for some actions. Thinking back to Lee Salter’s plenary we saw people jailed for Tweets. There are differences there but also more similarities than first apparent maybe. And you have other differences: animated gifs but also gamification. So I am “level 1″. In Twitter it’s maybe about follower counts, in Weibo you get to new levels and you get a special patch/badge. So Weibo may have started as a clone but I think it’s gone beyond Twitter in some features.
So I went to my boss, the head of the University. I wanted to look at UK Universities on Weibo. There is no table of these. So Yimei did a manual search using the HESA list of universities and various search terms. I talked about ID verification but there is also verification of pages, a whole other levels. 58% had presences; 43% had verified presences. Posts are mainly in Mandarin, some in English. So, we did set up a Weibo page but verification is tricky from the UK. I got sent a very tricky Mandarin form, no indication of who should sign in. They wanted an official stamp, and that had to be in red. It takes weeks… So we have a page, tweeting going on – mostly retweeting comments about Durham.
Looking at and understanding Weibo when you don’t speak the language… it’s odd. Twitter isn’t global. Weibo isn’t global either, although expanding and just launched a Thai version. No reason that Weibo couldn’t launch an English language interface and have that take off. People are happy to sign up to Apple and hand over power and choice to some extent. I’m not saying that will happen with Weibo but seeing that other cultural context lets you look at these things in a new light.
Q) The counts on your profiles are different on Twitter and Weibo. Any of those numbers can be normalised in some way. APIs for both will show you more detail of that data.
A) Yimei has been looking at content and interactions and she’s been noting changes, interactions and the role of time.
Q) Your comparison of censorship – I don’t think Twitter is that harsh.
Comment) There was an article this morning, in Hong Kong there are removed Tweets and censored Tweets. Also papers on censorship of Weibo, based on large data set.
A) Chinese Communist party control the broadcast media. Their response to social media, there are some who fear social media… You can censor afterwards but you can’t stop people tweeting. One of the founders of Weibo, a private company, was making a democracy point.
Me) Different types of censorship: political in China, commercial on Twitter – much more about brands etc. Now that may have different impact and ethical implications but those are both forms of censorship.
Comment) Yes, I think so. In China censorship really isn’t a line here though, it’s never clear what is/is not censored. Sometimes things appear to have been censored relatively at random.
Comment) Weibo functions because the government lets it and works within it’s mandate. Twitter chooses to censor
Me) Yes, but Twitter is not just making choices, it also comes under government pressure to censor – they have censored tweets in the Middle East after pressure from some governments, they were also pressured to censor and restrict during London Riots
A) Find that feature of social media very odd: people like Pinterest ignore copyright law and sort of reset rules in how they run having ignored those rules. Would YouTube have taken off if all non-cleared video content been removed/censored?
And with that we have to finish a really lively wee Q&A.
Johnny is introducing Dr Ruth Page and mentioned her book Stories in Social Media, and her article on Self Branding and Celebrity in Social Media. Last year Ruth organised an event at Leicester on social media which was a particular inspiration for this week’s event. Ruth is also Chair of new special interest group in Linguistics. David Bartam, Johnny Unger and Ruth are currently coauthoring a book on researching social media.
I am a little ambivalent about going last but at least I should get the last word (ish). So my talk is on “saying sorry”. I’m going to start by making some opening remarks on Twitter, why it is significant to corporations, a bit about the data set. And I need to make an apology myself – my section on corporate “talk” is not about apologies per se but contextualising apologies. Then I’ll look at approaches to apologies, characteristics of corporate apologies, and the application of linguistics. There are some interesting potential approaches coming from our highly varied backgrounds and disciplines here.
Twitter is public, participatory environment (Jenkins 2006); virtual marketplace (Bourdieu 1977); driven by value of attention and visibility (Marwick 2010). There is a direct access there – complaining to Starbucks or an airline say – but that isn’t evenly distributed power. In 2012 I argued that it works like a virtual marketplace and that that is around attention and visibility – reach of tweets, scale of followers, influence, etc. But that’s not the only way that attention and visibility shows up. It also shows up in linguistic choices in Twitter, how people shape interactions to those affordances, and an opportunity to see how those inequalities and hierachies works.
How is Twitter used? It’s electronic word of mouth (Jansen 2009) – and that matters to company, they mine that all the time. 51% of users follow users/companies (Edison research 2010). But there are different types of twitter accounts, there are corporate and/or personal accounts. But I’m interested in corporate and branded accounts. You also see distinct accounts for specific purposes, e.g. customer care accounts.
So, where did I start to get intereted? My data isn’t designed in response to a specific research question, the research has evolved organically from my work for last few years. My data is based on around 180k tweets harvested from 100 publicly available accounts, using custom Python codes pulling data from named accounts. Firstly I was comparing company use with “ordinary” use – although “ordinary” isn’t really the right word here. people said “you haven’t looked at hashtags”, and I did. So I started to look at the very different corporate use of hashtags. I had 40 companies, 30 celebrities, 30 “ordinary accounts”. Gathered data in 2010 and 2012. And today I’m talking about around 1200 tweets with apologies in them.
When I harvested the data I wanted to distinguish between updates, things that were public but with an @username, and the RTs. It doesn’t take into account quoting or MTs etc. as those are newer practices. So I was interested in the distribution of those types of tweets. In 2010 all types of users favoured the one-to-many broadcast pattern (the update), what does that say about identies and how individuals manage their interactions with others?
So how do companies use Twitter updated? There are interactions initiated by the company – pushing things out; broadcast the brand – through hashtags; broadcast across platforms – link analysis; broadcast conversational snippets – modified RTs (less occurance in new style RTs which isn’t covered here). So looking at occurance in hashtags we see that hashtags occur in the updates, they are in the one-to-many not the @reply one-to-one posts. And the use of hashtags is increasing. But this is odd, hashtags started off as folksonomic phenomenon to allow your topic to be promoted and found. Twitter changed their search algorithm so that you can find topics easily BUT the use of hashtags is increasing over time. The most frequest hashtags across all accounts is #FF. But digging further corporate hashtags tend to highlight products, corporate positioning, or making searchable the companies as producers. Whereas ordinary people’s hashtags seem to reflect the community – mainstream media, consumer interests. Yes Twitter gives ordinary people a voice, but they are still positioned as audience, as consumers when you look at those tweets.
I also looked for links in tweets, corporate accounts use much more links and their use it on the rise. There is a rise of the amplified talk. Originally in 2010 I saw links as ways to signify authority as recommenders, as endorsers. But different things happening now. General trend in links – Twitter is more multimodal – photos and videos increasingly important. And Twitter is increasingly multi-platforms – Facebook groups, Google Plus, Pinterest, Tumblr, Instagram, Daily Booth, VintageCam, YFrog, Whosay, Mobile Apps. Posts to multiple sites or connecting sites. Images are used to indicate products, what they are selling. But that’s not all that’s happening. In 2010 ordinary users tended to share clear links to articles in their field, their own blog – you could tell what their profession was. Corporations point to own web sites, promotional offers. Real collapse of professional/personal now taking place. Now ordinary uses point to some professional identity links, but also general life, photos, interests (e.g. fashion etc). Corporate use is a little different, some corporate professional links, but also sharing of images by customers/users, of experience images etc.
The last sort of tweets being shared are modified retweets, You see that celebrities their use is declining. Their use is slightly declining for ordinary users. But their use is increasing by corporations. And how does that happen? It’s about sharing compliments, feedback, things that promote their brand. They show they are engaging but in a very specific and careful way.
When I look at distribution of tweet types in 2012 there seemed to be few changes but Corporate use is radically different – many more addressed messages. Why is this happening? Up to this point I had looked mostly at updates, so it was time to explore addressed messages. I started with concordance techniques from corpus linguistics. I looks for the words that appear much more in just those addressed messages compared to all those other messages. I used the remaining dataset as the reference dataset, addressed messages as the sample. And certain words appear much more often, such as “hi”, “thanks” etc. They don’t just occur often, they often occur together. For instance “@username Hi [name], sorry for your frustration. Please follow/DM us additional details regarding this and we can try to help. Thanks.” So we are seeing the rise of customer care here. But it’s not just corporates who apologise…
Difficulty of apologising…
- Reluctant apologies – [cue Big Bang Theory clip of Penny reluctantly apologising to Howard. And being told to get over herself by him. Then him bursting into tears].
- Punk apologies – [cue music video]
- Politicians apologise – [cue Nick Clegg Apology Song video]
Even cats say sorry… even Whales say sorry when Twitter is over capacity…
Apology as a “post event speech act” (Spencer-Oatey 2008) – I’m following this understanding. This is recognition of something going wrong, acknolwedging that, reconciling parties. Enables future interaction and restoration of equilibrium (Ogiermann 2009). But research literature looks in linguistics tend to be about private apologies but there is a need for more work on public apologies, of apologies in large corpuses.
There is huge use of “sorry/apologise” here – aggregated data for both (and americanised spellings) show huge use of these terms in corporate tweets. Semantic components of an apology based on Bloom Culford(?) in 1989. Semantic components:
- Illocutionary Force Indicating Device (IFID) e.g. we’re really sorry
- Taking responsibility
- Explanation or account
- Offers of repair
- Promise of forbearance – not to make same mistake again.
In the 1200 tweets there was only one case of taking responsibility, and only one case of promise of forbearance, both were in ordinary accounts. Maybe commonsense reasons – liability, appropriateness, responsibility or role of person speaking.
You see lots of Illocutionary Force Indicating Devices. But there are very different approaches. Companies avoid restating the problem in 66% of apologies. The reverse happens in ordinary accounts – 58% of their apologies. My favourite of the apologies was “I’m sorry for the slugs in your strawberries”. It is good to acknowledge what you are apologising for but that is very risky, you risk raising the profile of the issue or validating etc.
10% of corporate apologies give an explanation, 27% of ordinary users did. When companies did explain their apology they shifted blame: denied the offence – telling the user they got it wrong; place blame with third party; factors beyond the company’s control (e.g. legal requirements, weather, etc). And on the rare occassions companies do accept responsibility they do that in a very specific way. They use linguistic constructions that made it very hard to see responsibility, e.g “sorry for the ongoing issues caused by the Booking Office cluser, there is a staff shortage in the area and we are working on it”. You need to show yourself in best light is the theme here, and a good way to do that is to make offers of repair. When company does that they happen in a very specific way with an awareness of the multi-party nature of the interactions. Offer of repairs tends to be something monetary or tangible – but not the tweeter doing that.
So these apologies are embedded in wider interaction. You see this in the way that questions occur in corporate apologies. 22% of corporate apologies and 13% of ordinary apologies include a question. Another aspect is the use of imperative, they are telling the customer or giving a command. It happened in 33% of corporate apologies, not at all in ordinary apologies. So often further contact initiated by company – e.g. “standby for a message” or further contact required by customer – e.g. drop us an email. That latter type are often hedged. But these are risky, they don’t close the loop, they risk the customer not responding, following up etc.
Openings and closings tend to be quite specific. Companies tend to use “Hi” and end in “Thanks” and a signature. Ordinary people do not. 37% of companies include a signatures, none of the ordinary accounts to. More interestingly perhaps, 19% of apologies posted by companies include greetings, that “hi”, but again no ordinary accounts to. They seem to be trying to build rapport, but that marks them out as different from ordinary users. So companies using this startegy mark their social distance, and show structures derived from email, not from conversation. But you do see alternative openings. Discourse markers (5% in company accounts, 27%(?) in ordinary accounts) – several flavours, so can, for example be associative expressiveness e.g. “Oh, I’m sorry” etc. Emoticons are also used to intensify negative sentiment or to upgrade positive sentiment – for offers of repair say. But sometimes it doesn’t match well. Some mismatch of negative responses – smiley to mitigate negative response. And sometimes it’s about promoting rapport (especially in line with future interactions).
So, what does this mean. The reason companies apologise in this way is to avoid face-threatening damage to reputation – e.g. avoiding restating problem. Mitigate face-threatening damage… [sorry, couldn’t keep up there]. So the implications? apologising is important strategy in use as part of customer care. There are repeated, distinctive patersns suggest a particular genre shaped by purpose of interactions and positions and roles of participants. So, the application here? Well it’s interesting for it’s own shape, you see the patterns you may not otherwise see. But challenge for myself is how do we use the work we do as linguistics in a way that helps other people. Well one of the things I’m doing is talking to London company who are creating social media analytics software, to create customer care software to make this work better. Perhaps not always pushing interactions into other spaces, closing loops, showing responsiveness etc.
Final thoughts – obvious limitations here. I haven’t taken into account participant perspectives; haven’t looked at whole interactions, just the apologies, it’s not the whole iteration; small number of accounts considered and not neccassarily noting location and cultural differences between Uk and US say. Possibilities though – what do you want to do with your work? Where can other perspectives by useful?
Unfortunately we finish without time for questions. I wanted to ask whether the rise of hashtags didn’t reflect the adoption and maturity of use by companies, or the use of hashtag campaigns. Hashtags also create links unlike search terms so have added value. And I wanted to ask about the issue of collapse – there has been a Twitter corporate strategy to boost use by media, by celebrities, as part of advertising campaigns all of which encourage collapse. Wider use and adopting of Twitter beyond professional spheres also have a big impact on collapse here, of the types of interactions, of the merging of followers etc.