RE: <nettime> Markets, Hierarchies, Networks: 2 questions

["Shannon Clark" <shannon@jigzaw.com>]

I'll take a stab at answering this - but also will issue an invitation
to everyone on the list to join me in person in San Francisco May 7-9 to
explore this much further and in greater depth - and I definitely would
organize additional smaller (i.e. less expensive) options on other dates
to further the conversation as well.

Social Network Analysis is growing in popularity in the past few years -
as a result a lot of people are starting to use it in various ways, not
always accurately or theoretically justified (or in keeping with good
practices around privacy which is a highly contentious issue). That all
said, it is my personal thesis (and which I agree with other recent
posts by John Hopkins on this point) that the way forward with the study
of ALL types of networks is massively multi/interdisciplinary approaches
- that look at the approaches in technology, in biology, in "pure" math,
in computer science, in many different social sciences and emerge from
that work.

It is also important to realize that when you are studying networks
which are in a social context, you are looking at an estimation, an
approximation of the organizations - usually at a point in time (or over
some person). As Valdis Krebs (see http://www.orgnet.com) said at a
one-day conference we were at (and at which he was speaking) last week,
these tools are best seen and used as a way to suggest actions and get a
sense of the structures - but at the end of the day people are people
and won't arbitrarily, for example, make a connection just because you
introduced them to each other.

He will be announcing a new release of his social network
mapping/visualization software later this year (probably in the next
month or so) in that he has suggested some variations to the "standard"
(in social network analysis circles) measures, as well as some new ways
to use and engage with the data.

In his case, he typically maps connections between people, typically
assuming bi-directional links (i.e. if I am linked to Brian, Brian is
also directly linked to me) and usually with the nodes being people
(less often in his typical use cases organizations as a single entity).
What his software then does is a) visually display the data in an
automatically clustered and clear manner b) allow for layers of data to
be used (i.e. to show a network with changes over a period of time, or
with the addition of new parties to the data set) c) perform a new set
of analyses on the data - including factors such as how many people
within the network each person in the network typically can "see" -
where "sight" might be defined as those parties I a)m connected to or
the parties I am connected to are connected to (possibly also who those
parties are connected to - the demo went by pretty quickly). This is
somewhat a new approach.

But it is important to recognize a few things.

1. A Heirarchy, if what you are looking at is representing how people in
theory are organized (or perhaps how people are "supposed" to work and
interact) can be represented within this type of data - it is just one
particular form of a network (and one that usually does not "measure
well" compared to other structures

2. These diagrams are never complete or accurate - they are dependant
deeply on how data is collected as well as a lot of non-statistical
factors - such as how you define what a "link" is (i.e. am I "linked" to
all the parties on  nettime? To all the people contributing directly to
this particular thread? To just those people whom I later meet and
interact with directly - i.e. not through the list?). And not
infrequently activity and life changes the networks - people leave/join
an organization, people meet outside of work at their children's
schools, or people's partners know each other - etc - in short
connections as well as activity "over" those connections may not (and in
some importance always are not) captured fully by the data collection. 

So at best the diagrams and the statistical analysis of the data just
are one part of a toolkit for understanding the organization. And
different structures, measures, and ways of interacting "within" that
network may be best suited for different organizations and purposes.

But all that said - there are a lot of very exciting developments in the
study of networks - especially when you step outside of "just" social
networks and look more broadly at all types of networks. In areas as
diverse as computer science, physics, and transportation studies a lot
of pretty interesting work and increasingly interdisciplinary approaches
are being tried.

A few that come to mind - I read recently of an approach to modeling the
spread of computer viruses which employed techniques developed in the
study of epidemics. At MeshForum last year Dr Evind Almaas of the
University of Notre Dame (where he works with Dr. Albert Lazlo-Barabassi
in his lab) spoke about techniques he is using to study the networks
within biological systems, specifically the interactions within cells at
nearly a protein level. His PhD, however, is not in biology but is in
Physics. 

I started MeshForum exactly to help explore these overlaps - but also to
explore where the language and different definitions are hindering
interaction and learning of alternative approaches and perspectives from
other fields of academic study, from the arts, or from different
industries. This year at MeshForum I am focusing the conversations on
visualization techniques and on the complex case of very large scale
social networks.

Large scale social networks are not well understood. Most social network
analysis is conducted on relatively small datasets (not uncommonly in
the less than a few hundred people), in a few cases on somewhat larger
datasets (the employees of a company typically) but the tools tend to
have problems at much greater than a few thousand nodes. And the
underlying math is usually focused on "shortest path" identification,
however as Valdis points out the actual people within a specific network
do, in fact, use other paths than the shortest - and not just because
they do not know which is the "shortest" path.

But large scale social networks also present challenges in many
dimensions - until recently the examples were fairly hard to study
(perhaps some religious organizations, some very large scale unions,
some political organizations - but none of those are all that easy to
study - nor are they entirely alike many of the emergent "networks"
today). Any visualization that seeks to render relationships between
millions of people visible will be highly complex (and unlikely to be
understandable on just 2 dimensions). It is also the case the in large
scale social networks there is some degree of data that shows that some
initial theories about "network effects" are not telling the whole (or
perhaps only) story. 

i.e. just because groups can form, does not mean that they either form
or last - Christopher Allen (who will also be speaking at MeshForum 2006
has posted about testing theories about "Dunbar Numbers" by using data
from online massively multiplayer games (such as Worlds of Warcraft) to
look at what actual, real group sizes are formed and which last and
"succeed" - see
http://www.lifewithalacrity.com/2004/03/the_dunbar_numb.html - as he
states, the common usage claims that "groups are limited to 150 or less"
- his data and research suggests that the more practical number is much
different, and lower in most cases)

For more information on MeshForum 2006 see http://www.meshforum.org and
our wiki at http://www.socialtext.com/meshforum. For four free podcasts
from MeshForum 2005 see
http://www.itconversations.com/series/meshforum2005.html in particular
Dr. Anna Nagurney and Dr. Noshir Contractor's presentations are relevant
to this conversation. Dr. Nagurney hold multiple professorships at U.
Mass Amherst (this year she is a Ratclife Fellow at Harvard) - in the
business school and in schools of engineering. She studies networks
(indeed has published multiple books on them) but her interest is around
especially the intersections and interactions between networks - for
example the impact of changes within the telecommunications netowrk(s)
on the transportation networks of a region (i.e. does increased
telecommuting options show up in the traffic patterns)

Dr. Contractor is in the School of Communication at UIUC. His talk
offers an overview of his decade plus of research into the creation of
knowledge and social networks - offering a very different focus and set
of questions than Dr. Nagurney. 

Hope this helps continue this great conversation - and as I said, I hope
that many people on Nettime can join me in San Francisco, either in May
for MeshForum or for other future continuations of this discussion.

Shannon




#  distributed via <nettime>: no commercial use without permission
#  <nettime> is a moderated mailing list for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: majordomo@bbs.thing.net and "info nettime-l" in the msg body
#  archive: http://www.nettime.org contact: nettime@bbs.thing.net