Jasmijn Van Gorp is an assistant professor in Audiovisual Data Studies at Utrecht University. She is the operational workpackage leader of CLARIAH's Media Suite, a Dutch research infrastructure for audiovisual data. Previously, she has been postdoctoral researcher in several digital heritage projects at Utrecht University and the University of Amsterdam.
Marc Bron is Senior Principle Researcher at Oracle, London. After completion of his PhD at the University of Amsterdam, he worked as a postdoc at Utrecht University and data scientist at Yahoo Labs, London.
This is the source
This article sheds an empirical light on interdisciplinary collaboration within the Digital Humanities by investigating the daily research practice of the Dutch Digital Humanities-project BRIDGE. The project developed and tested methods for automatically creating meaningful links and expanding archival television data. In the project, a high level of collaboration was required between scholars from two different disciplines: computer sciences and media studies. The majority of the epistemological encounters between the two disciplines took place in the design of the developed tools and the user studies to test the tools. The article is based on structured conversations between the two central staff members in the project, i.e. the computer science PhD-student and the media studies postdoctoral researcher. By unravelling the research project as a process of confrontation, identification and acknowledgement of situated knowledges, the article shows when and how the boundaries between the two disciplines have been maintained, crossed and blurred. The authors point to the benefits and challenges of interdisciplinary collaboration in the Digital Humanities, and formulate some best practices for future Digital Humanities-projects.
Sheds empirical light on interdisciplinary collaboration between the computer sciences and media studies on the DH project BRIDGE.
Computer sciences and humanities are rightly considered different scholarly
disciplines and they use different languages, approaches and paradigms. It comes as
no surprise, therefore, that mutual misunderstandings, unbalanced expertise levels,
and lack of interest in the other discipline are issues that regularly occur within
Digital Humanities-projects that typically work on a new technology and
humanities-based research question. In their anthology, Mutual incomprehension persists.
Generally speaking, scholars and artists understand little about the technologies
that are so radically transforming their fields, while IT-specialists often have
scant or no training in the humanities and traditional arts
. Insight into
collaborative research practices, and interdisciplinary collaborative research
practices in particular, can shed light on the quickly evolving field of Digital
Humanities. What does it mean to
A case in point is the Dutch television archive project BRIDGE (Building Rich Links
to Enable Television History), a collaboration of the Information and Language
Processing Group of University of Amsterdam, the media studies group Centre for
Television in Transition of Utrecht University, and the national radio and television
archive The Netherlands Institute for Sound and Vision. The project showcases the
synergy of the two academic disciplines involved, computer sciences and media
studies, which collaboratively developed and tested digital tools for exploration of
archival television data. Collaboration in DH is often, as aimed at building
something that works — a tool, a resource, an online collection
. The focus
of the research in the BRIDGE-project, however, was not solely on building digital
tools for the humanities, but also on the scientific evaluation of tools and
algorithms, as part of computer sciences, and a study on how such tools and
algorithms may advance insights in the field of Media Studies. The main output of
BRIDGE is a PhD dissertation in computer science, publications in both humanities and
computer science conferences and journals, and prototypes of tools (which are
eventually developed into sustainable tools in follow-up projects
Many scholars have investigated collaboration practices in the broader setting of
technology and software production.
Instead of a methodological account of the technical development and testing of the
tools, we take a meta stance and reflect on the project’s day-to-day steps in the
preparation for, execution of, and reporting on the research carried out. In this
article, we revisit these research processes. The main focus of the research process
was on conducting user studies. The technical results of these user studies are
already published in computer science outlets (
Taking an empirical glance at the digital humanities, thus at
At the core of interdisciplinary collaboration is the very notion of
a field of inquiry with a
particular
(italics by authors). Simply put,
academic disciplines are created by groups of scholars as a form of organization of
their knowledge about a certain object. The very boundaries and definitions of
disciplines vary over time.
Over the course of the past four decades, interdisciplinarity has been widely
discussed in academic literature. Terms such as multi-, cross- and
transdisciplinarity are used to describe different interactions between disciplines.
All terms can be situated on a cross-
already means some more exchange, while trans-
means that there is a creation
of something new that surpasses the boundaries of disciplines. According to existing disciplinary norms and practices
, allowing researchers to go beyond their parent disciplines,
using a shared conceptual framework that draws together concepts, theories, and
approaches from disciplines into something new that transcends them all
Knowledge, too, is not a given, but a construction of which the outside-inside
boundaries can be theorized as power moves, not moves toward truth
A good addition to Actor-Network Theory can be found in feminist science studies. To
pinpoint the constructivism of knowledge, Donna Haraway coined the concept of situated knowledge
greased poles dilemma: it is hard to climb a pole when you are holding onto both ends of the pole. The pole is a metaphor for the epistemological question (i.c. what is regarded as acceptable knowledge within a discipline): the duality between positivism and empiricism on the one hand and social constructivism on the other hand. She argues that a way to overcome these dualities is to articulate
partial visions, and thus
situated knowledge. Therefore, these scholars are advocates of multi-method approaches
In terms of our endeavour: how is the knowledge building of the media studies scholar
related to the knowledge building of the computer scientist? How can both
The article is an analysis of the daily research practice by the two central staff
members of the project: one researcher in media studies, and one researcher in
computer sciences, a (double) self-reflexive account
In order to reconstruct the research steps, we used different types of data: on the one hand e-mails and on the other hand a series of semi-structured interviews. We read our e-mail conversations first, and then the humanities postdoc prepared a set of interview questions and interview topics that formed the basis for a conversation between postdoc and PhD student in the form of semi-structured interviews. This was done three times: halfway through the project (year 2), at the end of the project (year 4) and at the start of the writing process of this article (year 5).
In the data (emails and transcripts of interviews), we look at what we - the two
central staff members on the project - say and how we say it; which words we use.
Latour, as becoming
rather
than being
, a process of identication that is constructed on the basis of
recognizing certain characteristics shared with another person, group or an idea.
Othering
, then, is a form of defining the self: when making differences
with the out-group explicit, the in-group defines itself. This implies that words
such as we
versus them
, or I
versus you
are important
markers of identity construction. Therefore, we specifically look at the use of these
pronouns to see how the researchers’ disciplinary identities are being constructed
and how knowledge is being formed.
The process of othering, however, puts constraints on the writing of an article across disciplines. It is a true challenge to write an article from two different viewpoints and paradigms. Therefore, we needed a new way to cope with this. Following Bartscherer and Coover’s book
Iand
you, the article itself is written as
we. The
weare both authors, but as this is a humanities journal with (largely) a humanities readership, we take humanities discourse and concepts as leading. For the description of the research process, we use the third person (he/she/the researcher/they) in order to identify ourselves as research objects for the sake of this article. As such, the article is an expression of standpoints (in plural) of which the situatedness is made semantically explicit by quoting ourselves and using the third person. In the conversations, the quotes by the Computer Science PhD student will be indicated by
CS(while in fact it is Marc Bron). The Media Studies postdoc researcher’s quotes are indicated by
MS(while in fact it is Jasmijn Van Gorp). The article is written from a multi-disciplinary standpoint, by means of the collaborative software WriteLatex, in itself a way of doing collaborative research. Therefore, the article is in itself, to quote
science in the making.
In the following empirical sections, we present a discursive analysis and deconstruction of the interaction between agents, where different backgrounds lead to misunderstandings but also to interdisciplinary knowledge production. We divided our analysis into four chapters which (humanities-wise) correspond with the classic narrative structure of Hollywood films: establishing shot, encounter, conflict and reconciliation. In section 4, we set the scene by elaborating on the basic characteristics of the two disciplines computer sciences and media studies. In section 5, 6, and 7 we follow a chronological order, corresponding with the identified three research phases in the project: the encounter, the struggle and reconciliation. For each phase, we first outline all research steps, then we discuss the challenges encountered and collaboration strategies per research step in a conversation structure. We conclude with some best practices for collaborative digital humanities projects in section 8.
In order to establish the context of the interdisciplinary collaboration, we define
in this section the disciplinary characteristics of our own disciplines as viewed by
ourselves. That is, as disciplines
are contested concepts and continuously
rethought, it is important to know how we ourselves experience the defining
characteristics of our very own disciplines. Each one of us wrote the section
concerning our own discipline, also referring to existing literature where relevant.
In Table 1 we give an overview of the main identified
characteristics of the two disciplines. We look at the object of research, methods of
knowledge building, research methods, pace, and publication traditions.
The distinction between sciences and humanities is often referred to as the
objectivity of knowledge versus experiential aspects. Or, to put it differently, the
sciences cumulate knowledge until a structure has been discovered, whereas the
humanities' work is associative and particularistic
Computer sciences is a science discipline, and Information Retrieval is one of the
main subdisciplines. Information retrieval (IR) is finding material (usually
documents) of an unstructured nature (usually text) that satisfies an information
need from within large collections (usually stored on computers)
In terms of methodology, the two disciplines also differ. In Computer Sciences,
computer-led experiments are the core, as well as creating new algorithms. Computer
sciences measure and use (solely) quantitative methods. The basic skills a computer
scientist should master are: good understanding of theory, algorithms, computer
design and programming languages. For the humanities-based Media Studies, the
individual researcher is much more central. The researcher’s own interpretation is
key. Qualitative methods are better suited for the humanistic stance in the
humanities
The practicalities of the two disciplines also differ. Computer scientists often work
in teams of 2 to 6 people, running experiments that are supervised by a professor.
Computer science is a quickly evolving field that wants to publish results at least
every six months in order to be first. First, they write a conference paper of
usually 10 pages, which is peer-reviewed and, if accepted, published in proceedings.
A common follow-up is to write a journal article on those aspects of the conference
paper that need more in-depth analysis. Given these factors, the scientific output of
computer sciences, as in other
Humanities research is sometimes defined as lone scholar
. In the humanities, scholars have tended to be
physically alone when at work because writing tends to be by nature a solitary
activity. In the humanities, the research often takes place during the writing. The
argument of an individual (or duo-) based (humanities) versus a team-based research
(sciences) tradition can be easily made by looking at random issues of Media Studies
journals as
We are perfectly aware of the fact that our own definitions and characteristics of
the disciplines as put here can be contested; however, these are the differences as
perceived by us, as we encounter(ed) them in our very own European, national and
time-specific contexts. This self-defined overview of differences between the
disciplines, provides the starting point, the Cultural
Science
: a movement that challenges the current disciplinary distinctions
between the humanities and social sciences on the one hand, and the math-based
sciences on the other hand. It can as well be defined as Digital Humanities
,
of course. Whatever concept is used, we tackle the disciplinary boundaries between
the sciences and the humanities in the next sections from the double perspective of
humanities and computer sciences. The question, now, is how the project evolved and
how the researchers of the two disciplines may have come closer to each other.
In the first year, 2009, the PhD student in computer sciences was the only staff
member. Research was focused on entity search and algorithms for linking entities,
i.e., exploring algorithms for a specific task: entity search. A good example of
how entity search is used in modern search engines is knowledge cards, i.e., for
searches about people, companies, or locations, not only a list of 10 result links
is presented but in specific cases additional information is presented. For
example, for the search terms Michael Jackson
pictures, links to his work,
and a short biography are shown for the famous artist. The algorithms guess which
person the user intended to find, i.e., the artist and not any other person named
Michael Jackson
At the start of the second year, 2010, when the postdoc entered the project, the team was complete and could start their interdisciplinary research. An important practical observation regarding the project we have to add here is that the computer scientist and media studies scholar did not reside in the same building, but work at different universities and even in different cities. The computer scientist worked in a large industrial science campus outside Amsterdam; the media studies scholar worked in a historical mansion as part of the humanities faculty of Utrecht University. At the start of the project, they did not meet regularly, as they were also both living in the cities where their university was located.
In order to develop a new tool, the first step was to understand how media researchers use their current main tool to acquire research material, i.e., the existing tool (catalogue) of the national radio and television archive, called
the known item search systemiMMix can be used for an exploratory search task in order to compile a list of features that would be desirable for a new search system. iMMix formed the
standard systemthat will be contrasted with the
newsystem. The latter, then, should perform better than the standard system.
The online search system of the The Netherlands Institute for Sound and Vision,
iMMix, consists of metadata descriptions of around 1.2 (currently 1.6) million
television and radio programmes of Dutch public broadcasters, provided by
professional annotators. It has a simple and an advanced search option that allows
users to search on specific fields. After the search results have been displayed,
they can be refined by a filter system. However, the search system is developed
for an experienced group of users: the broadcast professionals. Broadcast
professionals, in their need to reuse material, often know what they are looking
for, using a directed search, for instance, by providing a title or specific
content. Consequently, iMMix mainly supports known item
search and can be
considered as a known item search system
. Media researchers, however, would
benefit from an exploratory search system
, a system that supports them in
browsing, learning about the representation of their research topic in the
collection, and jumping from one topic to another
As the test person for a pilot study with the standard system iMMix, the
researchers were looking for a media researcher with no or just a little
experience with the search system and someone with a limited
prior
knowledge of the archival content in the archive, so s/he would have to rely
heavily on the search system for support. The team did not have to look far to
find a volunteer. The postdoc was the ideal candidate: a Belgian postdoc in media
studies who had never worked with the Dutch iMMix before. The team set up a small
experiment, in which the postdoc had to do a search on a self-selected topic.
During the search, she had to think aloud and this in the presence of her fellow
researcher of computer sciences who observed her search behaviour. The search was
being logged, videotaped and sound recorded. She searched for Russia
initially and ended up with three clusters of programmes for further scrutiny: two
travel reports, two programmes about migrant children and two current affairs
programmes (see
By then, it was time to conduct research for a computer science paper, as part of
the PhD-project in computer sciences. As users the researchers chose to work with
22 media studies students (freshmen). They had to find five television programs
for three tasks in a limited period of time, by using the iMMix search system. The
tasks for the users have been designed, by varying the difficulty in terms of
prior knowledge. The three tasks were defined in such a way that the students
could not use any of the elements/terms given in the task to find correct answers
and had to extensively explore the archive in order to do so. The first two
questions were situated in contemporary times, while the latter was a historical
one: (i) find 5 television programs with comedians from a non-western country;
(ii) find 5 episodes of drama series in which location plays an important role;
and (iii) find 5 game shows with a female host broadcast during the 1950s, 1960s
or 1970s. For the first task for example, students first had to find names of
comedians of a non-western country, while not being allowed to use any other tool
or search engine besides iMMix, before being able to find correct
answers/television programs. After completion of the three tasks (or when the
maximum 10-minute time limit per task was over), students completed a survey that
also contained open questions. They also conducted four in-depth interviews with
students after the computer experiment. The study showed that the search system
iMMix did not provide the necessary support for exploratory search by
(student-)researchers (see
In the first year, there was only one researcher. The computer science PhD-student
recalls: It was really informatics research; there was no
humanities aspect in there. The task has been evaluated on a benchmark set,
created articially.
After one year of
The very first collaborative action was the observation of the postdoc. For the computer scientist, it was very inspiring to observe a media researcher at work:
CS:
The media studies postdoc, however, was initially very surprised that the
observation was so helpful for her computer science colleague. Two years later,
she started to see what was so illuminating about her very first search when she
took the videotapes from the shelves and started to analyse her own search. By
retrospectively observing her own search behavior, she understood that the
relation between (re)search object, research question and search behavior of
humanities scholars is quite particular. It resulted, amongst others, in a
reflective article on her very first search in the archive
However, back then, the observatioal study of the postdoc was considered as rather
limited: it did not provide the sheer size/quantity that is usually required for a
computer science paper. Therefore, the PhD-student wanted to conduct a user study
with a high number of participants to have a baseline
measurement, in order
to be able to measure the improvement of the future (to be developed)
exploratory search system
.
CS:
The postdoc remembers that she did not really understand the purpose of the user
study back then. Words such as baseline measurement
and exploratory
search system
did not ring a bell.
MS: exploratory search system
. And then, at a certain
point, I interpreted exploratory search system
as something that has
to do with cultural memory. I translated it as a system that inspires users
to formulate new search terms; to provide them with inspiration in case
their cultural memory is not sufficient to find television programmes. It is
a kind of memory machine: it gives suggestions for new search
terms.
CS:
For the postdoc exploratory search system
was a fuzzy concept, while
cultural memory
was a vague concept for the PhD-student. Paradoxically,
it was this very semantic gap that provided inspiration for the humanities scholar
to give a
MS:
In the end, the four interviews did not provide sufficient ground for a separate
humanities paper, so both researchers worked on a computer science paper together.
The result of the study gave some insight in the search behavior of students, but
the scholars both realized that the quantitative measurements as well as the four
interviews with students were not sufficient to fully grasp the particularities of
search and search systems of media researchers. They needed a different approach
for their next user studies. However, they did not realize back then that the
solution
) for a better set-up of a user
study was exactly the combination of quantitive measurements with qualitative
interviews, as shown in the next sections of this article.
While the semantic gap provided inspiration in some occasions, it also impeded the
progress. There was a lack of semantic interoperability. For example, as
preparation for the user study, the postdoc had to create tasks for the students,
formulate queries
for the interface.
MS: a query
meant, so I used the word key word
instead.
But key word
to you meant something totally different, so we had the
one misunderstanding following the other.
A query is a request for information from a database, such as Comedian AND
non-Western
. A key word, in computer science and information studies
language, however, are the tags attributed to documents. In the case of the iMMix
interface, the keywords were the ones on display in the metadata field
keywords
of television programmes. While the postdoc did not really
understand why she had to formulate the tasks for the students in a computer query
language, the computer scientist did not really understand why the postdoc was
referring to tags all the time.
In general, the staff-members of the project really had to get used to each other,
and were still focused on their own field: CS: The first
phase was interesting but also frustrating. I really had to get used to the way
humanities researchers work and it was unclear how we could benefit from each
other.
MS:
Both researchers define the other discipline as us
versus them
. The second user study was
designed while the two central staff members had two different goals in mind.
There was only limited overlap or common understanding between the two
disciplines. Each researcher regarded his or her own discipline as the most
important. For the computer scientist, it meant that there should be enough
participants to make the statistics reliable. For the humanities postdoc, it meant
that she tried to obtain some interesting qualitative data. In terms of situated
knowledge, both researchers held to their own end of the pole.
The computer scientist, who started a year earlier, was already able to see the
contours of the humanities pole and already started building a bridge. He was
excited about the possibilities, and started to learn about Media Studies. He also
realized that they needed a different approach to deal with humanities scholars.
The media studies scholar who started a year later, on the other hand, firmly held
on to her pre-defined end of the pole which is humanities, and was not planning to
build a bridge, let alone walking on it. The
After one year of the
First of all, the idea for MeRDES emerged when the two staff members were both
attending a talk of one of the creators of Google’s N-gram Viewer. Within 30
minutes the researchers realized that it would be really good to have such an
n-gram viewer for the television archive as an exploratory search system
.
And so, they started to develop a kind of N-gram viewer for the television
archive, which they called Media Researchers’ Exploration Suite, abbreviated to
MeRDES
MeRDES is specically developed for media studies research and has two goals. First, it provides users with support for exploration, i.e., support in formulating key words and exploring various aspects of a topic (i.e., a tool for retrieval). A second aim is to provide support for discovering patterns in the data (i.e., a tool for analysis). As data for the search engine, we used the same large metadata set as is used in iMMix. They incorporated two side-by-side versions of a standard exploratory search tool that contains cloud filters. Based on Google’s N-gram Viewer, the researchers added a timeline visualization, and a term statistics visualization in which the characteristics of the result sets obtained with each side of the tool are shown and can be compared.
In order to develop MeRDES, two focus group discussions with media researchers
were conducted at two different media studies institutes. The team showed them a
mock-up of MeRDES and asked for feedback. Once the new tool was developed, they
tested a first version by means of a usability study with and tested by 30
computer science students (freshmen). The feedback was given to the software
developer, who improved the system. With the improved system, they were able to
conduct a remote online user study with 39 media studies scholars from the
Netherlands and the Dutch speaking part of Belgium. After having viewed a tutorial
video and having practiced with the tool, the media researchers were given the
following search task in this third study: Imagine that you have to write a
research paper on the portrayal of migrants on Dutch television. Try to find
relevant programs in the television archive. The goal of exploring the archive
is formulating the initial research question for your paper
. After the
search session, the users were asked to provide a research question and also
feedback in a post-test survey, both by means of check boxes and open questions.
The study showed that the double (subjunctive
) interface supported a wider
variety of research questions. For an elaboration on the methodology and the
results of the user studies, see
In parallel with the development of MeRDES, the researchers started conducting
interviews with media researchers: PhD students, postdocs, assistant, associate
and full professors. They asked them to pick one of their own research projects
and reconstruct their research process. This part of the research was
Although both researchers shared the idea about the usefulness of a tool providing an N-gram viewer type functionality, they had very different ideas about how such a tool would operate, which impeded the development of the tool. Retrospectively, it was really a struggle; one of the most difficult episodes in the project.
While the computer scientist started to understand and feel comfortable with the field of media studies, the media studies postdoc had to adjust to the speed and way of working.
MS: what is the time line
? At first, I thought you were referring to
the timeline graph in the interface, but then I realized you meant planning
with it. And now I really know why it was so important to have a strict
planning.
The postdoc had difficulties in following the pace. The only way to cope with it was to plan more strictly what she did.
Most difficulties were situated — again — at the level of a lack of semantic
interoperability. Another good example to illustrate the mutual misunderstanding
is the different take on television genres. The problem was that the iMMix
catalogue listed about 300 genres ranging from game show
, quiz show
,
quiz
, amusement
, Amusement
(with capital) and so on. These
television genres contained subgenres, sub-subgenres and also labels media studies
scholars do not use. The media researcher, therefore, wanted to make a new
division of genres which better matched the needs of media researchers. She knew
that media researchers really want to select and mix and match their own
television genres. The computer scientist in the project perceived genre
just as one of the many metadata fields of the television archive, just a facet,
similar to producer
or channel
. This view really conflicted with
that of the media researcher who considered genre as one of the backbones in film
and television studies research. They had a long, persistent discussion about how
the genres should be implemented in MeRDES.
MS:
CS: genre is a
very important concept in media studies
, just because all media
researchers were insisting on it; during the focus groups, the interviews
and you during the development. Genres had to be correct; that had to be
precise.
The project members had a similar misunderstanding with snippet
. It took
about three months before they realized that they had a mutual misunderstanding
about it. The computer scientist wanted to show snippets of television programmes
in the interface. Result snippets are the short summaries of a document, designed
to allow users to decide its relevance. Typically, a snippet consists of the
document title and a short summary, which is automatically extracted
MS: snippet
with birds and twitter, as the
words sound similar. And so I thought snippets were pop-ups or mouseovers. I
did not understand why we needed them in the interface at that point. I
could have benefited from a beginner’s course in information retrieval and
search systems.
Indeed, the lack of semantic interoperability impeded the development of the tool as it costed time.
On the other hand, the semantic gap also had its function: it forced the researchers to define terms and concepts, which enhances transparency and resulted in a better design of the interface. The collaboration was also supported by the fact that the media researcher moved cities and now lived in the same city as the computer scientist’s university campus. They met more regularly, in person, at the computer science campus which helped in decreasing the mutual misunderstandings. At the computer science campus, the media studies scholar also observed how the computer scientists work, in teams. She observed that they are less focused on individual research than humanities researchers are. It made her realize that the individualistic take of humanities is typical for the humanities, and is not necessarily the only way to go about research. She started to see her own situatedness.
The defining, crucial element for this research phase was the involvement of other
humanities students and scholars in the project. These steps of involving
CS:
MS:
It was important to elaborate the group of users from one (the postdoc) to many. The focus group discussions had a double function: they were an efficient way to gather feedback on the prototype and a pleasant way to get to know the defining characteristics of Media Studies.
The user-centred tool development also triggered some difficulties. The computer
scientist pinpointed the difficulty in the development of the tool as different views on the tool’s goal
.
CS:
MS:
The media researcher started to see her peers as
The remote user study with the 39 media scholars went well. However, it was
difficult to interpret the results because humans
provided the data: mouse
clicks and survey results of 39 researchers. This meant that both researchers had
to learn about human-computer interaction, a field which they were both not
familiar with. For computer sciences, it is difficult to get sufficient
statistical power as there was only a small
dataset (39 respondents) while
for humanities, online surveys do not always give satisfying information as they
cannot ask follow-up questions and have in-depth interviews to grasp the
interpretation strategies of the respondents. They realized that the
interpretative element should get even more space in their next user study, as
explained in the next section.
The researchers had a strong belief that more interpretative methods were required for the next user study. This belief was especially instigated by the parallel research step of conducting 27 in-depth interviews with media scholars. The interviews were conducted at the work space of the media researchers, so also had an ethnographic value for the computer scientist. The interviews were effective for the computer scientist to get an understanding of what the humanities is all about:
CS: then the data, the research question, and
back
, while she was drawing imaginary loops on the table. And
also during our conversations and our first think aloud study. Then you
first wanted to do research on Eastern Europe, then on Russian children and
then you ended up with research on Russian migrant children. Skipping from
one subject to another. But later I understood that it is just part of an
interpretative research process in which humanities researchers try to
understand a subject.
While the method of conducting interviews is common in the field of media studies,
for the computer scientist it was a new way of doing research. It also was a kind
of
CS: reception
, audience studies
,
production
, and texts
. I was surprised to what extent
culture is central and questioning what culture is, etc. I really did not
understand what they meant.
The researchers gave a lot of examples about their research, which helped to
understand the discipline-specific terms. The computer scientist learnt a lot of
the specificities of media researchers. The media studies postdoc realized that
her own (re)search behavior is very typical for Media Studies, which confirmed
that her own stake in the project could be considered more or less as
representative
for media studies.
In this phase, a productive struggle, there were clear signs that both researchers
started to build bridges. Semantics are one of the most defining parts of a
discipline. Learning each other’s semantics is a
The third and last phase could be regarded as a highlight in transdisciplinarity. The research went smoothly. For this phase, they again built and tested a new interface, but this time more qualitative methods were used for testing it. While MeRDES is really suitable for exploration within a single archive, in the last phase, we wanted to develop an interface for contextual research that links several heterogeneous audiovisual collections. We named the interface CoMeRDa (Contextualizing Media Researchers’ Data) (figure 2), and conducted two user studies to test it: a lab study with 44 students and a longitudinal study with 26 students.
The development of CoMeRDa went much better than MeRDES, although there was a more limited time frame. CoMeRDa enables simultaneous search in six collections: (i) a television program collection (metadatarecords for 700.000 programs, a selection of iMMix); (ii) a photo collection of television programmes; (iii) a wiki dedicated to television programmes and presenters; (iv) scanned television guides; (v) scanned newspapers starting from 1900 to 1995; and (vi) another newspaper collection from 1995 to 2005. The interface has three different tabs: basic search (search in one collection); combined search (search in six collections) and similarity search (search for similar documents across different collections). To test the interface two user studies have been conducted, i.e., one longitudinal and one laboratory.
The researchers conducted a laboratory study with 44 freshmen students of media and cultural studies. The students got three tasks: (i) imagine that you work at the editorial office of a current affairs program and are asked to collect information about celebrity X. Collect at least 5 items deemed to be relevant for this collection, (ii) search for events that were key in the career of celebrity X. Collect at least 5 items for this collection, for example articles and photographs, about these events, (iii) A key television program in the career of the celebrity X was program Y. Collect at least 5 items about the run-up to the program, the program itself, and the aftermath. To complete the first task subjects are provided with the tabbed display (one collection), for the second task with the blended display (all collections), and for the third task with the similarity display (similar documents in other collections).
The theme of the course selected for the longitudinal study is television and film
history and the research projects carried out by the students are centred around
television personalities between 1950s and 1980s. The 26 master students could use
CoMeRDa for their research paper, which had to be a photo essay on a celebrity.
Every second week, they were asked to complete a questionnaire on their use of
CoMeRDa. Every other week, there was a focus group discussion in which the results
of the questionnaire and their feedback on the interface were discussed. The
multimethod user study showed that the students preferred the blended display in
the first stage of their research project: they wanted to have an overview of what
is available. As soon as they had picked a research topic, they preferred to
search within one collection at a time. The results of these CoMeRDa user studies
are published in
In CoMeRDa, the development was less difficult, also because it was a more straightforward interface. Negotations were rather about presentation and display, than about facets and metadata fields.
CS:
MS:
Indeed, as the team members all went already through the development and testing of MeRDES, the mutual expectations were levelled. The media studies researcher started not only to see how the innovations done by computer sciences need a structured way of working and a fast pace, but also to adopt some daily working protocols of computer sciences, and to use important computer science terms in the daily conversations. She just knew much better how to communicate.
Similarly, it was much easier to agree on the set-up of the user studies. The
researchers agreed that the media studies researcher was the one who should
CS:
The computer scientist decided what they were going to measure, for the computer
science papers. And the media studies researcher translated the questionnaires
from computer science to humanities language, made the assignments doable, and
tailored the task and set-up to media studies research. The computer scientist
recalls: Without you, it would have been a computer scientist
who gave an assignment to media studies scholars. And then they would have had
no clue about what to do.
Moreover, it proved to be important to communicate
CS: It won’t work. We have to change the set-up
. And
then I wrote back to you: No, we have to do it like
this, we have to measure it!
MS:
For the media studies postdoc, there was one major challenge in the software development. She was afraid that they were building tools that would not be used, as seems to be the case with the majority of tools developed by DH-projects (cf. Golumbia, 2013).
MS:
In the end, it turned out that the tools were appreciated by scholars and
students. Of course, a lot of feedback was received and there is still some work
to do to make the tools sustainable (see
CS:
MS: So
what? What do the numbers tell us?
. Humanities scholars need
interpretation, culture and meaning. Basically: humans are central in all
phases of their research process, as well as in defining their research
object. However, bringing the humans into interface development and bringing
interpretation into computer sciences is a very exciting new field for the
humanities. So, I can not imagine anymore
Both researchers clearly crossed the bridge and felt quite comfortable in mixing paradigms and methods. If the researchers have to pick one challenge, however, these are the publication outlets. It is just very hard to read and understand each other’s papers.
CS:
As most papers about the user studies are written within the computer science paradigm, the same is true for the humanities scholar. A true challenge was the writing of papers and articles collaboratively. In the beginning, it was one of the staff members who wrote and the other who commented and provided some paragraphs for the article. In the final year, they wrote an article together, with the collaborative software writeLaTeX.
CS:
MS:
Still, it is not easy to write together in terms of language, because computer sciences and humanities write for different audiences. The different journals and conferences need specific formats and styles. There is still a gap, which can only be bridged by new journals in the field of digital humanities.
Eventually, when the project was finished it was clear that both researchers appreciated and understood important aspects of the other discipline. The computer scientist could finish his PhD on time. The media studies scholar did not write a monograph, nor she could conduct research with the tools, as they were only ready when the project was finished. However, she learnt so much about interface development and computer sciences that she included many of the assets of computer sciences in her future humanities work.
Transdisciplinarity did arise, and the bridge was built. Even more so: after
completion of the project, the media studies researcher started to work on a new
project and interface, but from then on based at a computer science department.
The computer scientist went in the opposite direction and started to work in a
joint history-computer science project, based at a humanities department. The once
so
In this article, we explicitly chose to describe the challenges that arose within our research project. We viewed those challenges from two perspectives, by using a conversation structure between the two central staff members of the project. The conversation structure enabled us to write an article from two standpoints, and also to show how our standpoints shifted during the course of the project. As such, we embodied two disciplines, and showed how knowledge claims can be negotiated in the practice of a Digital Humanities research project. The researchers of two disciplines gradually gained mutual understanding.
After setting the scene of both disciplines, we identied three phases in the project.
Similar to the common narrative of Hollywood films, there was first an encounter,
then a struggle and conflict, and finally appeasement and a
frustrating
The interaction and power relations continuously changed during the course of the project as knowledge at all levels increased. The researchers got more insight in the particularities of the other discipline’s field, thus developing a shared language and increasing semantic interoperability between the members of the production team. The computer scientist said that the slow pace of the humanities research was something they had to adjust to and changed their observation of media studies scholars along the way. The humanities scholar learnt to work with a (more) strict planning and also adopted technical language. In the project, it turned out that both disciplines learned the language of the other’s discipline, e.g. by reading literature, by discussing the disciplinary boundaries on a daily basis, by conversations and (ethnographic-style) observations.
Transdisciplinarity especially arose in the set-up of the user studies. It is in the
user studies where epistemological encounters took place between the two disciplines.
The user studies are, in other words, the space of negotiation where computer
sciences and media studies really crossed. In BRIDGE, the computer scientist took
care of the experimental set-up of the user studies, while the media studies scholar
supervised the
The most important facilitators of the interface development, too, were the regular contact and interviews with the users, the group of media researchers. During the development of the tool, the users could direct the developers to desirable functionalities and the use of understandable labels. The interfaces MeRDES and CoMeRDa could not have been made without the encounter of the two disciplines, and the continuous negotiation between both. BRIDGE proved not only to build bridges between nodes of a television archive, but also between researchers and disciplines: a true actor-network in a Latourian sense. Extending the network from two researchers who collaborate (i.c. a computer scientist and a media studies scholar) to many scholars was one of the key success factors of the project.
The project required both the computer science researcher and the humanities researcher to learn about the tools, techniques, and methodological practices of the other discipline and find a common research direction in which the strengths of both disciplines could be utilized. This required not only effort in becoming adept at another discipline’s practices but also overcoming a psychological barrier, i.e., abandoning something you are good at to become an apprentice in another field. These, together with the misunderstandings and frustrations that accompany co-operating with someone with a different background, are obstacles which should not be underestimated, especially with regards to the (anxieties about) careers of the researchers involved. Institutions, therefore, play a vital role in the enhancement of Digital Humanities and should carefully look at the assessment procedures for tenure, and specifically the assessment of digital publications such as tools and databases.
Researchers should have some of both worlds in their training, one could argue, or at
least being able to move in between the different spaces so as to support an
understanding of each other’s discourses. As
Also, a problem arose in the physical distance of the project’s members. Scholars are initially distanced in terms of paradigms and distanced in terms of physical presence. E-mail conversations were bound to end up in conflicts. Working in the same space, or at least having regular meetings or conversations, is essential for a good progress of the project. This is simple advice, but highly effective.
A persistent problem, however, are publications. As long as the academic publication outlets are centered in one or the other field, it remains a true challenge to write for a journal of which the format, language and paradigms are not yours. Even with this article it is admittedly the humanities scholar who chose the constructionist theoretical framework and approach. The conversation structure was used as it allowed us to cope with the different paradigms and the different voices of the authors. Still, it was quite a challenge to write this article from two perspectives in terms of words and style.
For a future project, it would be nice to include a third perspective or voice, this of the heritage partner or data provider, and also differentiate views within groups, by e.g. including the perspective of the software developer. In this case, it would be necessary to conduct individuel interviews or focus groups with all participants throughout the project and document these carefully. For long-term projects, the problem might be that participants leave the project or (re)join which makes it difficult to write a joint article and to keep track of all different voices in the article. In any case, we advise DH-projects to document their collaboration and have (at least) annual meta-discussions about each other and the project.
Last, reflection requires at least a moment of distance. It would not only be
difficult but even impossible to write this article during the course of the project.
By that time, we were pre-occupied with reaching our goals, developing the tools on
time, setting up the user studies and writing up the papers. Only now can we look
back at BRIDGE and reflect on the persons we were in the beginning of the project. We
are aware of the fact that in this article we are open about our own
how does the movie
end?
) but about acknowledging that it is a process of which all elements — the
researchers, the interfaces, the methods and the semantics — are interconnected and
can view and be viewed from multiple perspectives.
The authors would like to thank all scholars, software developers and archivists involved in the BRIDGE project. The authors would also like to thank Koen Leurs, Sonja de Leeuw, the three anonymous reviewers and the editorial board of DHQ for feedback on earlier versions of this article.
Recipes: Building a Media Suite within the Dutch Digital Humanities Infrastructure CLARIAH