Who is in your social network and how do they influence its behaviour? Which events affect it positively or negatively? You need to find out if you’re carrying out social media campaigns.
The research that ignited some fire in my belly today is from a computer scientist (Andrew J. Scholand at Sandia National laboratories) and two psychologists (Yla R. Tausczik and James W. Pennebaker from the University of Texas). It’s called “Social Language Network Analysis” (SLNA) and this is a new methodology for identifying the different types of relationship between people in a social network. They elegantly call it “socially situated relationships between individuals”. The tools they use come from social language processing and network analysis. They are looking to map the expertise level of each individual in the population and also their level of friendship /social support within the network.
In social networks as well as in life we are influenced by the believes, trends, and events in our communities. Sometimes we have more impact in particular contexts and less in others, or perhaps we are more influenced or less influenced by the others depending on the topic. The roles we play change and affect the community as a whole. They say:
“We actively leverage relationships to both make sense of the world and select optimally among our available choices in a socially situated way, with the salience of various groups waxing and waning in different contexts. In any given group, however, the informal organization that structures and defines processes such as sensemaking is often not explicit or even consciously recognized by participants. Within organizations, similar amorphous behavior and decision shaping concepts (such as ‘culture’ and ‘norms’) are recognized and discussed, but do not have computable formulations. Driven by research questions that seek to bring to the fore these intangible yet powerful influences, we have developed a new quantitative approach that leverages the ability of social language processing to identify psychological, social, and emotional undercurrents in interpersonal communication with the structural insights of network analysis. We call this approach social language network analysis (SLNA)”.
They have applied this to the real world application of organizations wanting to improve employee performance and retention but you could just as easily apply this to online communities. In fact this could be a nice health check for those types of social network.
Social network analysis (SNA):
“Social network analysis (SNA) measures and represents the regularities in the patterns of relations among entities. SNA is predicated on the concept of the relational tie as an essential building block, focusing on social structure via a collection of methods.”
SNA has been around for well over a century and has been used to analyse complex relationships between members of a social system of any type. It takes into consideration everything from the structure, to the relationships between individuals right down to each individual themselves. It also looks at attitudes, behaviours and beliefs in such systems. It is most of all concerned with how the ties between the individuals (so the edges of the graph), affects the individuals and the relationships between them. Influence in the network may come from a select number of nodes for example. The attributes of the individuals are less important than their ties. This way you can work out why the network is behaving in a particular fashion or identify different influences within it.
Social language processing:
“Social language processing is built on the idea that language conveys information beyond the literal meaning of the words used. Empirical studies have shown that the way in which people use language can reveal information about their thoughts and emotions. ”Linguistic Inquiry and Word Count (LIWC) was designed to measure word use in psychologically meaningful categories. LIWC has been successfully used to identify relationships between individuals in social interactions, including relative status (e.g. deception) and the quality of close relationships”.
As they point out, here the goal is to determine traits such as age, sex, social status, mental health etc…of the individuals in the network. Different types of words give information about these things, for example emotive words, pronouns, social words and so on. However “SLNA uses language variation to elucidate internal structure rather than using an external definition.”
SLNA:
“By using the linguistic content of communication in SLNA’s quantitative models we can discover relationship subtleties missed by content-agnostic SNA analyses.”
There are 3 steps to this process:
- Preprocessing: all the data (text, could be a conversation or a tweet) is assigned to each individual
- Processing: all text associated with links is converted to a quantitative metric (of your choice e.g. ratio)
- Post-processing: graph processing (visualisation)
Metrics are many and you can use whichever you want. For example they used the PageRank algorithm to determine an individual’s perception of the expertise hierarchy in their specific group. You could choose to measure levels for friendship in the group using SLNA for example or couple that data with events and see how different experiences change things within the group.
Your social media campaign:
I do believe that counting the number of followers is not ever going to tell you very much about what is going on in your networks. A social media campaign must have specific goals, very much like a website has a specific reason for being. If the goal is to spread a message, then you need to check that your network is healthy in this respect. Do you have the right balance of individuals and will they pass on your ideas? Maybe your social media campaign is about getting people to start behaving in a new way (start buying your product instead of a competitor’s for example) then you need to send signals through that network and analyse the effect. Who are the influencers and how can you best cater to their needs?
We’re just scratching the surface here, there’s a lot to think about and a lot of data to gather and process to find out the things that really matter.
