***** To join INSNA, visit http://www.insna.org *****

A timely question, for me :)

On Jan 10, 2006, at 9:48 AM, Antonella Cascio wrote:

> ***** To join INSNA, visit http://www.insna.org *****
>
> Dear All,
> We are two students at Politecnico di Milano (Italy), working on
> our master
> thesis. We want to map conversation threads (specifically from forums
> residing in the intranet of the company we're studying) in order to
> visualize communication flows and run analysis on them. We know it's
> something that has been done many times before but we can't find a
> way to
> operationalize the very first steps in this process: how to create a
> communication matrix from the forum istself. For instance: how to
> code the
> message that starts a thread? As a communication of the one-to-all
> type?
> and
> what about a reply to a message? is this considered a message from the
> poster to the thread initiator, from the poster to all other
> posters in the
> same thread or from the poster to everybody?

I think that it matters what you wish to analyze because Forums are
quite different from the social networks that were the archetypical
data for the development of most of the measurements. For example
consider just betweenness centrality, it is a flow concept that makes
sense in situations where the dyad is a private communication,
however I don't think it makes sense in a forum situation where the
dyad of 'replied-to' is quite different. We can't, for example,
interpret dyads-in-public as a route for information flow.

In our work[1,2] we analyze bug-tracking forums and were interested
primarily in measurements of contribution, rather than information
flow (they are, of course, related but remain distinguishable). We
decided to analyze the dyads as contributions, where outdegree
centralization becomes a measure of the inequality of contribution;
we have tried to stay away from measures whose interpretation relies
on flow (such as betweenness centrality).

We therefore were able to avoid coding one-to-all semantics, sticking
with reply-to-sender. Thus the first message (only one half of a
dyad) is coded as the direction of the second message, or not coded
at all if there were no replies. The second dyad is between the
sender of the 3rd message and the sender of the 2nd message. Yes, it
is simple but suits the research question (we hope!).

We have considered weighting the replies on the basis that senders to
bug forms are often added to a 'group reply' so that all posters
earlier in the list get all the replies. However while this is true,
it is also arbitrary because some bug tracking forums post all
message to, say, the developer-list, so that everyone sees them. The
'lived experience', if you will, of each forum is different. For
example, if one were analyzing MySpace blogs, it would be tempting to
measure a post as a communication to all friends; but one has no data
on whether the post was actually read by each friend[3]. One could
reformulate one's research question to deal with this (create a node
labeled "Jane's Friends"?) or perhaps a question can work with dyads
with the semantic of "attempted communication" or "potential
communication".

In short I think there cannot be an over-arching answer, it is going
to depend on the semantics of the dyads, your research questions and
the specific practices of the forum you are examining. So, what are
your research questions? Of course I am equally sure that there is
more to be said and look forward to hearing from others ;)

Cheers,
James Howison, PhD Candidate
School of Information Studies
Syracuse University
http://james.howison.name

[1] Kevin Crowston and James Howison (2005) "The social structure of
Free and Open Source software development" in First Monday, volume
10, number 2 (February 2005). http://floss.syr.edu/publications/first-
monday-sna/

[2] Crowston, K. & Howison, J. (In press). "Hierarchy and
centralization in Free and Open Source Software team communications".
Knowledge, Technology & Policy (in press). http://floss.syr.edu/
publications/ktp2005.pdf

[3] Certainly one can infer that the post was read if the friend
posts a replay, but what of those that don't post a reply, did they
read it? Perhaps one's research questions can cope with a dyad that
has the semantics "B was moved to reply by A's post" but how could
one know that a non-link didn't mean "B was too annoyed to reply to
A's post" rather than "B did not read A's post" or "B was
uninterested in A's post" or, in a Knowledge Management context, "B
already knew what A was suggesting".

One could continue with this, for example when analyzing IRC messages
it is common for clients to be set to alert (say Bounce in the dock,
or pop-up notification) when a user's nym or word of interest is
mentioned. Should one then assume that they read every message with
their nym in it? What about immediately surrounding messages? Or is
the act of attempted communication enough?

It's almost enough to make one turn to survey research ;)

_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.