Research article: Online influencers

There are many attempts in the industry (and many apps) to identify online influencers. My main concern with them is that they operate with a seat-of-the-pants operational definition of the concept of “influencer.” What are, exactly, the behaviors that characterize an influencer? And how do you know that the behaviors a certain app is measuring (e.g. number of followers and number of retweets on Twitter) are actually measuring social influence and not something else? We have seen that, according to such measures, Sockington the cat is more influential than Chris Brogan.

This is where academic research can help. I’m browsing the latest issue of Human Communication Research and came across this article:

Huffaker, D. (2010). Dimensions of leadership and social influence in online communities. Human Communication Research, 36(4), 593-617

[note: David Huffaker completed his Ph.D. at Northwestern and now he is a researcher at Google, according to his website. This paper was part of his dissertation work.]

The article set out to identify the communication traits of online leaders (aka influencers, but influencer is not a word, so it can’t be used in an academic publication). These communication traits are of two types: (1) linguistic characteristics; and (2) social interaction patterns.

To identify influencers’ communication traits, Huffaker used both automated textual analysis and social network analysis.

Drawing on previous literature on leaders in the offline world and opinion leaders, Huffaker proposes the following abilities that define online leaders: The ability to:

  • trigger feedback
  • spark conversations within the community
  • shape the way other members of a group discuss a topic

in other words, they…

  • set agendas for discussion by causing or facilitating dialog on a particular topic
  • frame discussion by shaping the way a topic is talked about

The study was designed to examine the relationship between 3 characteristics of online leaders and online leadership itself. If there is a strong relationship, this means that the 3 characteristics are good indicators of online leadership. So, the 3 characteristics were the independent variables, and online leadership was the dependent variable:

Independent variables:

  1. communication activity (measured as: number of posts a person has contributed to a group; number of replies a person contributes to messages initiated by other group members; length of participation in the community, as an indication of credibility)
  2. social networks (measured as: expansiveness – the number of times a person replies to different group members; reciprocity – frequency of a person’s participation in a back-and-forth dialog with another person; brokering – being the link between two otherwise unrelated groups)
  3. language use (measured as: talkativeness – the average length of messages contributed by a person; linguistic diversity – the number of unique words found in a message; assertiveness – frequency of certain words indicating assertiveness, such as “always” and “never;” affect – frequency of words that represent emotional language, such as “nice,” “ugly,” “happy,” etc.

Dependent variable: online leadership, operationalized (measured) as:

  • reply trigger – the ability to inspire responses
  • conversation creation – the ability to spark a long dialog between users
  • language diffusion – measured as the number of words used by the author of a message that were repeated by other users in subsequent replies (so, if A triggers a discussion using the word “inappropriate,” if the same word is used frequently in other posts on the topic, this indicates high language diffusion, and therefore, social influence

The author used linguistic analysis(LIWC)  and social network analysis (UCINET) software and performed the analyses on a random sample of 16 Google Groups on various topics. The sample included 33,540 users and 632,622 messages written between June 21, 2003-January 31, 2005.

The measures of the independent and dependent variables were analyzed using correlations, regression analysis, and hierarchical linear modeling analysis (I wish I could explain these to you, but I can’t – especially not the last one) to test a series of hypotheses that are neatly summarized by the author in this one sentence:

“Users who generate the most message replies, comments, or conversations, or spread the most word choices [aka online leaders, the dependent variable – MV’s note] were expected to exhibit more communication activity and tenure in the community, more network centrality and brokering behaviors, and language that exhibits talkativeness, affect, assertiveness, and linguistic diversity [measures of the independent variables, MV’s note].”

After testing relationships between these variables, the following emerged as characteristics of online leaders:

Online leaders:

  1. post a lot of messages and a lot of replies (high communication volume)
  2. have been part of the group for a long time (online tenure)
  3. engage with several different members of the group
  4. engage in back-and-forth dialog with members of the group
  5. tend to write longer messages than other group members
  6. use a richer vocabulary that other community members. Tthis is linked in the research literature to cognitive complexity – i.e. how smart one is – and therefore, to credibility. Or, it is possible that the richer, more colorful language draws readers in.
  7. are assertive
  8. express affect and emotion (attitudes) – which, along with assertiveness, may be an indicator of leaders’ passion for the topic
  9. are able, through behaviors 3 and 4 above, to create supportive, loyal relationships with other community members and between them. Leaders are important to the success of the group as a whole because they are instrumental in creating and maintaining relationships within the community.

The only characteristics that was not associated with leadership was brokerage.

Of course, these findings are valid for discussion groups. We don’t know yet if they apply to other types of online communities.

So, what do you think? Do these sound to you as reasonable characteristics of online leaders? Will this study change the way you identify online influencers?


How social media change organizing

I gave this presentation in TECH 621 today – I’m pretty proud of the way I synthesized and organized (what I thought were) the most important ideas from Clay Shirky‘s book “Here Comes Everybody.”

I’m not sure how well it went over in class – students seemed tired, and we didn’t have time to discuss as much as we might have liked to. So I’m posting here and inviting students and readers to continue the conversation in the post’s comments. If you have read the book, I believe you’ll appreciate this synthesis. If you haven’t, I’m not sure how much sense it makes…


Questions? Comments? Cabbage jokes?

Reading notes: Twitterville

Twitterville is a collection of stories about Twitter written by a twetizen who is enchanted with the Twitter village. It is a business book as much as it is a piece of anthropology – by reading stories about a place, we infer its values, social norms, and culture.

Most of the stories are wonderful, uplifting, and show the positive side of Twitter. They are not, I think, your everyday Twitter stories – they are the extraordinary events that stand out in a place’s history. I’m glad someone took the time to document and save them. I remember living through most of them, and it felt great to read these accounts of recent Twitter history. Israel is an excellent story teller, and if I didn’t envy his warm, fluid, friendly, yet clear and simple writing style so much, I’d go on and on praising it :).

I loved reading the book, and enjoyed every page of it. I can imagine critics complaining that the book is overly positive – that it portrays Twitterville as a better place than (they think) it is. Israel’s Twitter enchantment doesn’t bother me, primarily because, like a respectable ethnographer, he spells out his biases clearly and repeatedly. He explains his point of view and enables the reader to decide how to interpret the content. As a qualitative researcher, I do not believe in the myth of objectivity. I think the best we can do is explain our biases, so readers can make informed decisions about interpreting our writing. I see very little of this in popular literature, and I hope more authors will adopt this practice.

… and Israel’s enchantment with Twitter doesn’t bother me, because I can relate to it and I share his point of view. I was initially amused by the claim that Twitter can lead to… world peace. But as I read the last chapter, I realized that, as a firm believer in the power of communication to make and break our world, I too, think, that conversation is the best solution – and that it can, indeed, help us make peace.