the digital humanities at stanford: mixed
networks, mixed messages
Who's working on digital humanities at
Stanford? As technologists, our impulse is to record and analyze this
network, yet as humanists, we can see the problems with this project
before we even start. Network analysis has not often been applied to
humanistic topics because our data sets and data collection are
willfully untidy: no two people, no two relationships, no two objects
are alike, so how can we reduce them to a simplistic set of nodes and
edges? Yet we're intrigued by the possibility of one more way to look at
human relationships, so we created these mixed network graphs — graphs
of unlike objects — and analyzed them by asking questions rather than
computing answers. Below are three visualizations that reveal different
aspects of the mixed network. We created these multiple perspectives in
order to shed light on the decisions that go into the creation of a
network visualization.
The data were gathered in late 2010 and early 2011 from academic
websites by Elijah Meeks and Molly Wilson. Elijah's data-gathering
ranged across the digital humanities landscape of Stanford, while
Molly's focused on the Mapping the Republic of Letters project. Elijah
and Molly could have chosen alternate ways to collect data — personal
interviews, time on task, email networks — but chose to look at the way
the digital humanities community at Stanford is representing itself. The
visualizations are thus a representation of the digital humanities' own
web presence.
I. Who's There?
First, let's pick out the people involved. Anything that does not
represent an actual person here is greyed out, and the people nodes are
colored according to their status within the university: faculty,
graduate students, staff, and others (usually, people who are not a part
of Stanford). Larger nodes simply mean more connections to other nodes.
●
Faculty
●
Graduate Students
●
Staff
●
Other People
◎
Degree (undirected)
- Several of the largest projects seem to attract primarily one
type of contributor. There does seem to be such a thing as a faculty
project node, a graduate student project node, or a staff project node.
Does this have an analogue in the real world, or is it an artifact of
the way people are credited as contributors?
- Faculty (red nodes) are connected to the greatest number of
nodes. There are three large green nodes but no large blue nodes. Are
faculty actually involved in the most projects, or does this show how
connecting faculty names with a project gives the project greater
visibility?
- Many of the nodes in this graph do not represent people; more
specifically, few of the highly connected nodes are people. This is
partially explainable by the way the data were gathered; we scraped
websites and CVs rather than interviewing people about their personal
relationships. Nevertheless, is there a real-world analogue to projects
being central to people's working relationships?
II. Project Neighborhoods
It's easy to accept that a node represents a person or project, but
the concept of a link is harder to swallow. The visual symbol of a
connecting line can mean founding a project, funding a project, or even a
single promise to drop by for a meeting. For our first foray into
problematizing links, let's turn to four Stanford projects that we'd
expect to see represented especially well: the Bill Lane Center for the
American West, the Literature Lab, the Spatial History Project, and
Mapping the Republic of Letters. In each graph, the color starts at the
white node and spreads one degree in all directions, showing what graph
theory calls an "ego network". Hover over the graphs to see the ego
network expand to two degrees. Node size corresponds to betweenness
centrality: how likely is it that paths between nodes go through the
given node?
●
Bill Lane Center for the American West network
●
Lit Lab network
●
Spatial History Project network
●
Mapping the Republic of Letters network
●
Center of ego network
◎
Betweenness centrality
- In some cases, adding another degree to the ego network makes
the ego network much larger (e.g. Mapping the Republic of Letters). In
other cases, it doesn't make as much of a difference (e.g. Bill Lane
Center). Becoming connected to a project node with many other
connections — in network theory, a node with high degree — can make an
ego network explode.
-
What is the relationship between the first-degree network and the
second-degree network, and how does it play out in projects? A project
with a small first-degree network but a large second-degree network may
be more nimble and streamlined, with resources at hand but not in the
way. On the other hand, it may be isolated, with valuable resources too
far away to be useful. A project with a large first-degree network may
be thriving, or it may be paralyzed with more hangers-on than it can
handle.
III. Hubs and Authorities
Now that we're examining edges, we can think of networks as belonging
to two broad categories. The first is an undirected network, where each
link represents a reciprocal, symmetrical relationship between two
nodes — most social networks are seen as undirected. The second is a
directed network, where each link has a source and a target, such as the
world wide web. Our digital humanities network is directed, and we
chose to make larger entities the targets of smaller entities. For
example, an edge between a person and a project treats the person as the
source and the project as the target. To preserve the integrity of
directed edges, we didn't record many links between two people or
between two projects.
Here, the nodes' size is based on an algorithm called "hubs and
authorities", originally developed to analyze relationships between
websites. A hub is a node with many links pointing outward from it, and
an authority is a node with many links pointing to it. The graph on the
left represents hub rankings, and the graph on the right shows authority
rankings. The four projects from above are colored as before, but in
addition, nodes with hub or authority values greater than the median
value are highlighted; hover over the graphs to find out what those
nodes represent.
- A strong authority, for us, is a large entity that many smaller
entities are associated with, and a strong hub is a smaller entity
that's associated with many larger ones. Our network has many more
strong authorities than strong hubs. Where are all our hubs? Given the
cynical view that academia trades on big names, we might expect a few
faculty or staff to have their names on many projects, especially
because our data collection came from websites and publicity materials.
However, the data collection doesn't reveal a cohort of individuals who
are involved in massive numbers of large projects. This might allow us
to call Stanford's digital humanities community fairly distributed, with
nobody who holds a disproportionate amount of influence. On the other
hand, the reason we don't see more hubs could also have to do with our
data collection. We spent more time on project websites than on personal
websites, and we made projects the center of our investigation.
-
Given that all of our major hubs and authorities are projects, not
people, it's worth thinking about what a hub project looks like and what
an authority project looks like. A hub project (e.g. Humanities
Research Network) probably is involved with many other projects that we
saw as larger or more comprehensive. It is a resource, and a well-used
one. An authority project (e.g. Parker Library on the Web) is likely
large and makes use of many resources, and most of the projects it's
involved with are projects we saw as smaller. For us, a funding agency
counted as smaller than the project it funded, so an authority project
may also be one funded by many sources.
-
Two of Stanford's flagship digital humanities projects are the only
two nodes that are both strong hubs and strong authorities: the Spatial
History Project and Mapping the Republic of Letters. This means that
they are linked to larger projects and smaller ones, and if we didn't
know anything about the projects themselves, we might imagine that these
were midsize projects with a relatively low profile. However, these are
two of the most well-known projects in the network, possibly due to
their many edges directed both inward and outward. Should we be making
projects strong hubs as well as strong authorities?