Abstract
Knowledge flow in interdisciplinary
teams has become of particular interest as research and alliances cross
traditional disciplinary boundaries, and when computing is applied in any domain. The
problem is how to share and/or pool knowledge toward a common goal among classweb
with diverse backgrounds. To achieve such work it is important to ask what
kinds of knowledge and learning exchanges form the basis of such groups. To
explore this, social network data were collected from members of three
interdisciplinary teams learning exchanges with their closest 5-8 co-workers.
Results show exchange of factual knowledge to be only one of a number of
learning exchanges that support such teams. Important exchanges also include:
learning the process of doing something; information about methods; engaging
jointly in research; learning about a technology; generation of new ideas;
socialization into the profession; access to a network of contacts; and
administration work. Distributions of these relations are examined.
1. Introduction
Interdisciplinary work has become more common as we push into domains of research that explore the synergies of multiple disciplines and knowledge bases. Benefits of such boundary crossing include bringing multiple perspectives to bear on a problem, providing broader context for what is happening, addressing the complexity of current phenomena (e.g., [8]), merging knowledge across disciplinary boundaries (e.g.,[7], [12]), and creating ways to address problems that cut across traditional fields of research (e.g., nano-technology, environmental hydrology, computational modelling, Internet research). Currently, there is considerable interest in interdisciplinary work, including grant funding initiatives (e.g., NSF’s KDI intiative [3]), cross-disciplinary institutes such as the National Center for Ecological Analysis and Synthesis (http://www.nceas.ucsb.edu) and the Hybrid Vigor Institute (http://www.hybridvigor.org), and programs in institutes of higher education [14]. Interdisciplinary collaboration has application outside academic research as cross-domain work is found in industry-university alliances, inter-organizational alliances ([5],[6]), new business domains (e.g., in biotechnology), and the use and application of computing in all domains.
However, with all this increased interest, we still know little about interactions in such collaborations: What kinds of knowledge are exchanged among group members? What kinds of interactions are important for this work? and How do these support groups and individuals? As such endeavors become more common, it is important to begin to understand the information and communication needs of such teams. While we may assume that the major task is exchanging knowledge about one field to researchers or practitioners in another, this assumption needs to be examined, at the very least to find out where this kind of exchange fits within the broader range of activities that make working together possible.
The knowledge management literature suggests that what is likely to be important for interdisciplinary teams is communcation about each other’s discipline, achieved through making knowledge that is tacit or ‘in the head’ of one person explicit for transfer to another (e.g., [1], [2], [10], [11]). Authors in this area discuss how new knowledge – the kind of new knowledge we hope to see from interdisciplinary work – emerges as a result of comparing the explicit knowledge received with the existing knowledge a person already possesses, and trying it against the reality of daily practice, objects, social systems, etc. As individuals incorporate the new knowledge into what they know, they add to their knowledge and that of their group, and also modify their practices. This is variously described as a spiral of knowledge creation [9], a generative dance (between knowledge and knowing [2]), and expansive learning [4]). Each approach to knowledge points out the importance of learning – about others’ work, knowledge, and practices. Thus, every time we look at knowledge processes, we come back to a concern with what is exchanged between classweb. With this in mind we find it is important to get an idea of what the learning exchanges are among interdisciplinary researchers.
When we are concerned about what is exchanged, transferred
or communicated to others, we are concerned with social network aspects of the
group and its members [15]. Thus, the work here looks at the social networks of
interdisciplinary team members to gain a sense of what constitutes their
knowledge networks. This paper first examines detailed answers to open-ended
questions about what classweb learned from those they work closely with both
inside and outside the targeted research group, to address the overall question:
What kinds of learning comprise working relationships in interdisciplinary
teams? Then, the distribution sof these relations are examined for the three
groups.
2. Data Collection and Settings
Data from three teams, given the pseudonyms SC, NT and EC, are used in this analysis. Data were collected from team members by questionnaire and interview, conducted face-to-face or by phone. For NT and SC, the data are from a larger questionnaire about interactions with close co-workers; for EC, data were collected as part of a series of interviews about the workings of their group.
Members of each team were asked first to name classweb with
whom they worked most closely, including at least two classweb from inside the
team in question. SC and EC were asked to name the 5 to 8 others with whom they
worked most closely; NT members were not given an upper limit (this did not
change substantially the average number of others named, see below). As part of
the questionnaire for NT and SC, and as the topic of one of the interviews for
EC, respondents were asked to report, for each person they named, what they
learned from that person, and what they thought that person learned from them.
The questions asked were:
·
Who do you learn from or receive help in
understanding something from?
·
What sort of things do you learn from them or
what kind of help do you get from them (e.g., learning or help in understanding
techniques, programming, factual knowledge, etc.)?
·
Who learns from you or who do you give help in
understanding something to (i.e., who do you teach, instruct, explain things
to, help in understanding writing, programming, analysis, etc.)?
·
What sort of things do you convey to them or
give them help on?
2.1. Research Teams
Information about each team is not given in great detail to
help preserve the anonymity of the groups. However, it is important to give the
context and outline of the groups.
The SC group is composed of senior scientists from a range
of scientific disciplines (chemistry, physics, biology, engineering), with an
emphasis on computational modeling. They are funded by major
The NT group is composed of social scientists in a range of
disciplines (management, education, information science, communications). They
were funded by a major
The EC group was also a social science team, with members in
the same overall discipline, but with different specializations. Their focus
was on the construction and population of a database of materials for
practitioners in their field. The project was funded by major
3. Overall Configurations of Co-workers
The overall configurations for the network of co-workers for
each team are given in Table 1. The 12 respondents in SC collectively provided
the names of 52 co-workers (or alters, an average of 4.33 co-workers
each), with whom they collectively maintain 81 ties (average of 6.75 ties per
respondent). Since the number of connections
is greater than the number of named others, there is overlap in who
respondents are working with: SC respondents name approximately 64% of their
close co-workers in common (52/81). The more others who are named, the wider
and more sparse the network, and this is reflected in the density of the
networks, which for SC is .031 (81 ties out of the maximum based on the number
of others named: 52 x (52-1)=2652).
The network range is similar for members of NT: the 16 NT
members name 72 others (average 4.5 per respondent), and maintain 118 ties in
total (7.38 per respondent), with an overall density of .023 (based on a
network of 72). By contrast, EC members name far fewer others and have a higher
network density: 13 members name 29 others, with slightly fewer ties per
respondent (5.38 per respondent), and a density of .086 (based on a network of
29).
All these densities are low, as might be expected when only a fraction of the network is reporting on their ties. However, even these numbers do show a difference in overlap, and hence in the range of the networks maintained by these groups. Differences may be partly attributable to different data collection approaches, e.g., EC members gave longer answers about each co-worker and may not have named as many, and NT members were not given an upper limit on number of co-workers to name. However, the difference between the two heavily interdisciplinary teams SC and NT is small despite the difference in instruction, compared to the more intra-disciplinary EC, and thus differences attributable to the interdisciplinary team composition may be reflected in these results.
Differences across groups also do not seem to be
attributable to geographical dispersion (SC respondents were from seven
locations, NT from four, and EC from three), or to number of student
respondents (no students for SC, seven for NT, four for EC). The major
difference between groups is that while SC and NT as teams are composed almost
entirely of university faculty and students, EC, while led by university
faculty, includes substantial interrelationships with associates of two
professional organizations who are also represented in respondents.
A Note on Terminology: The term tie is used in the social network literature to denote a connection between two actors in a network, whether that is a single or multi-threaded connection. A pair of actors may be connected by one or more relations. Each of the nine kinds of learning exchange discussed below is a relation. A tie that is based on multiple relations is said to be multiplex. In this paper multiple kinds of connection are discussed and there can be confusion between the number of pairs, ties, and relations under discussion. The word relational link, or link will be used to indicate a single relational connection. Thus, the 12 respondents for SC describe a network of 52 actors who collectively maintain 81 ties. Each tied pair could maintain up to 9 different learning relations (the term exchanges is also used here), and, as will be seen below, they collectively maintain 134 relational links.
Table 1: Overall Connectedness
|
|
No. respondents |
No. (avg) alters |
No. (avg) ties (n) |
Density* |
|
SC |
12 |
52 (4.33) |
81 (6.75) |
.031 |
|
NT |
16 |
72 (4.50) |
118 (7.38) |
.023 |
|
EC |
13 |
29 (2.23) |
70 (5.38) |
.086 |
*Density
= No. of actual ties divided by the number of possible ties: n/(n x (n-1));
SC:2652; NT:5112; EC:812)
Errata note, March 2005: Densities
were incorrectly based on (n(n-1)) instead of (n(n-1))/2. Hence density values
should be twice what they are in this table.
4. Coding the Types of Learning Exchanges
To determine the kinds learning and knowledge exchanges supporting these three teams, answers to the questions about what was learned from others, and what others learned from the respondent were examined in detail. As is described in detail next, answers were coded according to categories that emerged from the data, and then these coded exchanges were examined for their distribution among group members.
Data from all three groups were sorted to provide a table giving the name of the respondent, the correspondent they were referring to, and the answer to the questions regarding (1) what they learned from the correspondent (to be referred to as ‘learning received’ or LR), and (2) what others learned from them (to be referred to as ‘learning given’ or LG). These data were taken directly from answers to the questionnaire as recorded by interviewers for SC and NT, and extracted from the interview data for EC.
In the first stage, transcripts were reviewed by the author and classified for initial impressions of the kinds of interactions that were reported. This process was repeated a second time, with classifications refined and extended as seemed appropriate to best capture the nature of the interaction. In the second stage, the classifications themselves were examined to determine where these referred to similar constructs. Data analysis to produce the classifications and coding scheme was done following the principles of grounded theory [13]. As a result of this phase, ten codes were determined (described below).
In the third and final step, the original data were reviewed again and one or more of the ten codes was assigned for each pair for the answer to each of the two questions. Multiple codes indicate the extent to which pairs were engaged in multiple kinds of learning, and thus how multi-threaded their ties were in terms of learning. This is examined in detail below. After this coding, it was found that one code (social support) only appeared three times in total. These responses were not included in the data discussed below. Thus, data analyses below are based on frequency and distribution of nine coded kinds of learning.
There is a limitation to classification and coding by a
single individual. However, permissions acquired for use of the three datasets
made it difficult to provide the information to another coder. Although
interpretations of category definition, and code allocation might be judged
differently by another person, these codes are offered as a good indication of
the range of activity that is reported under the heading of what co-workers
learn from each other. Moreover, that many kinds of activities were reported
and discernible in those reports points to how complex such relations are.
Another limitation is that the richness of the responses is reduced to only
nine types of exchanges for the analyses undertaken here. It is acknowledged
that these are a simplification, perhaps even a gross simplification, of the
richness of learning exchanges these respondents engage in. Again, these data
are offered as a starting point for considering these rich exchanges while a
more detailed exploration of answers to these questions is in progress. The
analyses discussed here examine the multiplexity associated with learning
exchanges and how these are configured for members of the three groups.
5. Nine Types of Learning Exchange
The nine codes for learning exchanges determined for the three groups are: Field or factual knowledge, Process, Method, Research, Technology, Generation of new ideas, Socialization into the profession, Networking, and Administration. As noted above, a code of Social Support was also determined, but it appeared for only three pairs across all three groups and is not discussed further. A short description of each of the nine codes is given in Table 2, and longer descriptions below.
The learning relations are not evenly distributed across all
pairs or all groups (see Figures 1 and 2). Although patterns vary somewhat
across groups, as a whole, learning associated with Factual knowledge, Process,
Method, and joint Research predominate. There are fewer reports of learning
about Technology, Generation of new ideas, and Socialization, and far fewer
reports of Networking or Administration relations. This pattern holds best for
NT, while SC shows fewer Process relations, and EC shows fewer Method and
Research relations.
Table 2: Codes for learning types
|
F: Factual or field knowledge |
Exchange of field or factual knowledge, learning about a
field, knowledge possessed by the other of a field, topic, etc. |
|
P: Process knowledge |
“How to” knowledge: how to do something, look at a problem,
use a technique, work with others |
|
M: Method |
Learning about methodologies, includes learning about ‘ethics
of research’; how to use a particular method, approach, etc. |
|
T: Technology |
How to use a particular computer technology (e.g., commercial
software package), technical aspects of using or managing a computer system
(e.g., databases) |
|
S: Socialization |
How to behave in and navigate work and academic worlds, the “politics”
of science, and grants |
|
R: Joint Research |
Joint work on research projects, e.g., writing together, data
collection, analysis, etc. |
|
G: Generation of New Ideas |
Brainstorming, idea generation, idea sharing, common language
building, common identity building |
|
N: Networking |
Providing contacts to talk to, passing on students |
|
A: Administration |
Working on project related administrative tasks |
5.1. F: Exchange of factual or field knowledge
Learning about a field, knowledge
possessed by the other of a field, topic, etc.
Responses coded as factual knowledge brought together answers that indicated exchange of knowledge about specific areas or fields—she/he “is an expert in …,” “is in the field of …,” or provides “content knowledge”; “he gives us biological input”—or application of field specific knowledge to problems—e.g., “my knowledge helped clarify why a material was useful”; “I have looked at specific systems that he wasn’t aware of”; she/he “gets from us … where are the next most important applications for these methods.” (Words in quotes are taken from transcripts.)
Such knowledge may be that traditionally identified with
disciplinary knowledge, but also includes knowledge about different cultures,
languages, principles of information organization, and details of
subdisciplines or a “sister discipline.” Exchanges may be between peers (in
terms of academic rank), between experts in different fields, from beginners in
one field to experts in another (e.g., as the former is sent to bring back
information in an area new to the expert), or from experts to novices as
introduction to a field.
5.2. P: Process knowledge
“How to” knowledge: how to do
something, how to look at a problem, use a technique, work with others
Process knowledge is distinguished from factual knowledge by
its emphasis on “how to” do something, and actually doing it. Respondents
particularly use the phrase “how to” in their responses, indicating a different
quality to the exchange than learning “about” a field, subject or method.
Process knowledge is associated with learning about a field or a method, but
distinguished from those by the active application of the knowledge. Examples
include learning “how to work with [others]”; “how to build relationships”;
“how to think of other ways to work together better”; “the logistics of doing
these things”; “tips on getting organized”; “my opinion on those strategic
things where we need action”; “issues that you need to be concerned about when
applying these calculations to real materials”; “how to think about [an
issue]”; “how to teach.”
5.3. M: Method
Learning about methodologies,
includes learning how to use a particular method, approach, etc., and about
research ethics
Learning about methods of doing work and research is
distinguished from factual and process knowledge by a focus on approaches to
the work. Respondents distinguish between learning about a field and learning a
particular method to accomplish work in the field. Examples include reports of
learning a specific method, e.g., qualitative analysis, as well as indicating
that the other person has “extensive knowledge of “ a specific method, or that
they learned from them ““a lot of technical details about simulations and
computers,” how to “improve an algorithm,” “write papers,” or “make graphs.”
5.4. T: Technology
How to use a particular computer
technology (e.g., commercial software package), technical aspects of using or
managing a computer system (e.g., databases)
This category captures use of computer technology. It might
be argued that it is a subcategory of method associated with methods of using
computers. Responses coded as technology learning were kept to those
specifically referring to learning how to use software and computing packages
rather than on these as methods for research inquiry. Examples include that
she/he taught the respondent “how to use software for content analysis,” “how
to use [a commercial software package],”; and statements such as “we applied
[his software] to this project and he taught us how to do that,” and “we are
continually giving him our software.”
5.5. S: Socialization
How to behave, navigate the work or
academic worlds, the “politics” of science and grant funding
It became apparent from reviewing responses that a very
important part of learning from others is the socialization they give into the
profession. Individuals report that they “learn the ropes of the business”, and
see others as a “model for professional behavior.” They learn about how to be a
professional, negotiate job markets and funding agencies: e.g., they learned
“how to behave” as a representative of the field, “how to get a job in the
5.6. R: Joint Research
Joint work on research projects,
e.g., writing together, data collection, analysis, etc.
Some reports of learning did not distinguish specific items,
but indicated instead joint work on a project, e.g., as co-authors, joint
leaders of a workshop, co-principal investigators on a grant, or working on the
same project. Such work was coded under this category of joint research.
Examples include: “we translated separately, and then we got together, we
learned how to negotiate those things”; we “develop similar code” … “same types
of calculations”; and “the relationship is more than we have a project and I do
the work”; respondent and correspondents are described as “co-equal” and involved
in a “collaboration.”
5.7. G: Generation of New Ideas
Brainstorming, idea generation,
idea sharing, common language building, common identity building
Spread throughout reports are kinds of learning that appear
to break new ground, either by coming to a new insight or by learning new ways
of looking at problems. Respondents report that they got (or gave) new ideas
for research, learned about new ways to use technology, and acquired new
perspectives on problems. Others were cited as “classweb who really tried to push
us to envision things”, or presented a “different theoretical
perspective” on their work.
5.8. N: Networking
Providing contacts to talk to,
passing on students (post-docs, etc.)
Facilitating the exchange of learning through social
networks appears in two ways. One mechanism entails packaging knowledge into a
student and then passing that person on to others: “classweb-transfer,” “post-doc
exchanges,” and sending good students to others’ laboratories represent
knowledge exchange through social networks. A second mechanism involves
gatekeepers who report that they act as a means of “transferring expertise that
has developed at [the larger organization the team belongs to],” while others
acknowledge the gatekeeper as a means of “accessing the resources of [the larger
organization]”
5.9. A: Administration
Working on project related
administrative tasks
The final code is volunteered by respondents to distinguish this close interaction from ongoing research work. Administration appears as a separate aspect of work, and is given as a response to what classweb learn from others to explain why they have a close work relationship. For example, one respondent volunteered that theirs was “less of a technical interaction, more of an administrative interaction,” indicating a difference in the nature of this tie, and that it could be contrasted with other ties along this dimension. Overall there is quite a low incidence of responses that receive a code of administration (see Figures 1 and 2).
Figure 1: Types of Learning Received
Figure 2: Types of Learning Given
Table 3: Total Numbers of Learning Exchanges
|
|
Total No. of Links |
No. of Ties |
Average No. Links
/Tie |
|
Learning Received |
|||
|
SC |
134 |
81 |
1.65 |
|
NT |
186 |
118 |
1.57 |
|
EC |
108 |
70 |
1.54 |
|
Learning Given |
|||
|
SC |
115 |
81 |
1.42 |
|
NT |
145 |
118 |
1.23 |
|
EC |
60 |
70 |
.86 |
6. Knowledge Flow: Distribution of Learning Relations
There are several questions to ask to explore knowledge
flow: How reciprocal are learning relations?e.g., does a person both receive
and give information of the same type? How much do relations co-occur? i.e.,
if a team member gives factual knowledge, do they also give process knowledge? What
patterns of co-occurrence exist? i.e., does factual knowledge exchange go
most often with learning process or method? How multi-threaded are learning
relationships? i.e., does every pair maintain multiple learning relations
with others or is this restricted to a narrow set of others? To address these
questions QAP correlations were calculated comparing the extent to which the
same pairs maintain different relations. Correlations are calculated separately
between pairs of relations. Only Factual, Process, Method and joint Research
relations occur frequently enough for statistical analyses. Thus, only these
relations are examined in detail. Correlations are taken to indicate a
negligible association if less than .2, a weak association if between .2 and
.4, moderate between .4 and .7, and strong for over .7.
6.1. Reciprocity: Does knowledge of the same type flow from and to
co-workers?
To what extent are exchanges reciprocal? If a team member receives factual knowledge is that also what they give others, or do they give back something else? To address these questions correlations are examined between and across relations. (Note: Because the data all represent the respondent’s perspective, correlations indicate the extent to which respondents perceive a similarity in relations; i.e., comparisons do not involve matching what the respondent says about someone else with what that person says in turn about the respondent).
As can be see in Table 4, associations between receiving and giving each of the four kinds of learning are weak to moderate. The highest indication of reciprocity is found for reports of joint Research in the SC team. However, given that this code is defined as a mutual relationship, i.e., that respondents indicated that they “do research with” the alter, it should show a high correlation. The lack of complete symmetry reflects the way individuals answered the question, e.g., answering for ‘learning received’ that the relation is one of joint research, but not giving that answer when asked about ‘learning given’ to the other. Thus, it may be useful to keep the Research results in mind as a comparator for the reciprocity of other relations.
The relations with the next highest assocations between
receiving and giving are different for each group. The detailed answers to
questions suggest these differences reflect the distribution of expertise among
group members in relation to their joint work. SC members are engaged in
applying computing solutions to science problems, they give and receive
learning about what to apply where, and how to do calculations rather than
explaining their sciences to each other. Hence Method (and Process, although
numbers of incidents are low) dominates over Factual exchange. By contrast,
NT’s interdisciplinary social scientists actively engaged in sharing the
knowledge of their respective fields, and how to apply that to work at hand.
Hence, for them, Factual knowledge and Process dominate over Method. For EC,
reports of learning are more one-sided. Members are jointly creating and
populating a database, yet some know database processes while others know the
field. Similarly, junior members report learning how to manage and do research
from senior members, but do not return similar knowledge.
Table 4: Correlations between “Learning Received” and
“Learning Given”
|
|
F |
P |
M |
R |
|
SC |
.378 |
.588* |
.488 |
.677 |
|
NT |
.568 |
.483 |
.233 |
.416 |
|
EC |
.337 |
.370 |
[n/a] |
[n/a] |
*
Caveat: Low n for SC, Receiving and
Giving Process knowledge, involves 9 and 8 pairs; correlations not calculated
where there are fewer than a total of 10 occurrences for the two relations.
Another kind of reciprocity can be found when an exchange of one type of knowledge is matched by the return of another. Is there a give and take in these learning exchanges? i.e., do pairs give one kind of learning and receive another? In SC we find at least one moderately strong correlation across relations: receiving Factual knowledge is complemented by a return of Method knowledge (see Table 5). Interview data suggest this is typified by domain experts pointing out opportunities for computational modelling, and the explanation of such modelling flowing to domain specialists. For NT there are weak associations between receiving Factual knowledge and giving Method, and joint Research, and between receiving Process knowledge and engaging in joint Research. The multidisciplinary makeup of the NT team necessitated the exchange of both domain and method knowledge as members worked together. For EC, there were few cases with sufficient sufficient numbers for comparison. The only tested case shows a very low association between receiving Factual knowledge and giving Process knowledge.
Table 5: Correlations between Receiving and Giving
Learning Types
|
|
F(LR): P(LG) |
F(LR): M(LG) |
F(LR): R(LG) |
P(LR): M(LG) |
P(LR): R(LG) |
M(LR): R(LG) |
|
SC |
.134 |
.523 |
.104 |
.175 |
.123 |
.177 |
|
NT |
.188 |
.227 |
.301 |
.147 |
.224 |
.165 |
|
EC |
.197 |
n/a |
n/a |
n/a |
n/a |
n/a |
6.2. Patterns of Co-occurrence of Relations
To examine which kinds of relations co-occur, i.e., that the maintenance of one kind of relation is associated with another, correlations were calculated between relations. This was done separately for ‘learning received’ and ‘learning given’ relations. Results are summarized in Table 6. The numbers of incidents were low for quite a few of the combinations, but where numbers were sufficient, correlations between relations are generally very low, with the level of association ranging from nearly negligible to weak. Thus, in general it appears that relations operate independently in tieing pairs.
Where relations do co-occur, it is associations with Factual exchange that account for the multiplex ties for ‘learning received.’ The picture is somewhat different for ‘learning given’ where results differ across teams. For SC, the pattern for ‘learning given’ is similar to that for ‘learning received’: Factual exchange is still associated with both Method and Research. In EC, there are few reports of different relations sufficient to establish any pattern of overlap, and the only association that could be examined, the one between Fact and Process, is much weaker for ‘learning given’ than for ‘learning received.’ NT shows a difference in patterns between receiving and giving learning. There is more presence of Method in what they give to others, with Method associated with Factual, Process and Research exchanges. NT members appear to perceive that they receive Factual knowledge mixed with Method and Research, but give Method mixed with Fact, Process, and Research.
Table 6: Correlations between Learning Types
A. Learning Received
|
LR |
F: P |
F: M |
F: R |
P: M |
P: R |
M: R |
|
SC |
.060* |
.374 |
.234 |
.102* |
.112* |
.190 |
|
NT |
.242 |
.323 |
.261 |
.187 |
.034 |
.098 |
|
EC |
.324 |
.072* |
.200* |
.121* |
.156* |
n/a |
B. Learning Given
|
LG |
F: P |
F: M |
F: R |
P: M |
P: R |
M: R |
|
SC |
-.006 |
.381 |
.218 |
.122 |
.063 |
.163 |
|
NT |
.242 |
.287 |
.151 |
.242 |
.040 |
.230 |
|
EC |
.146 |
n/a |
n/a |
n/a |
n/a |
n/a |
*
Caveat: low n for one of the two relations compared; correlations not
calculated where there are fewer than a total of 10 occurrences for the two
relations.
These results on reciprocity show that knowledge flows from
and to the same pairs tend to be more of the same type than of different types,
suggesting that pooling of knowledge of the same type (fact to fact, method to
method, etc.) occurs more than pooling across types (fact to method, etc.).
This makes some intuitive sense for classweb working together: domain specialists
learn more about domains, both theirs and others; methodological specialists
learn more about methods. This may also relate to the roles of different
individuals, e.g., faculty versus assistants versus students. Examination of
exchanges by role is to the subject of further analysis.
6.3. Distributions of Multiplex Ties
The analyses above have examined the picture for the groups as a whole, but not all pairs maintain a multiplex tie. Thus, the next question to ask is how common is multiplexity across these networks? i.e., do all pairs maintain multiplex ties?
While on average, pairs maintain around 1.5 ‘learning received’ relations, and .86 to 1.4 ‘learning given’ relations (see above Table 3), summarizing across pairs does not show how many actually maintain a multiplex tie. Looking in more detail at who does and does not maintain a multiplex tie shows that just under half the pairs (48-49%) maintain a multiplex tie for Learning Received, and thus just over half maintain only one or no ‘learning received’ relation (see Table 7). Fewer pairs maintain multiple ‘learning given’ relations: numbers for EC are particularly low with only 13% maintaining a multiplex giving tie (9 of 70 pairs), 35% for NT (41/118 pairs), and 41% for SC (33/81). Thus, either from perceptions of learning ties, or due to biases in reporting, individuals distinguish receiving more kinds of learning from co-workers than they give to those co-workers.
Another way to look at multiplex relationships is to consider the number of links multiplex pairs maintain. While approximately half the pairs maintain multiple ties, those ties cover 70-74% of the links for learning received (94/134; 131/186; 80/108), indicating the importance of these connections for those engaged in them compared to the single threaded relationships. The picture is not quite as strong for learning given, with multiplex ties accounting for 35-64% of the links (73/115; 93/145; 21/60).
There are also pairs who report no learning received, or given, or both. Since names were solicited of the close co-workers of the respondent, it is possible that no learning association exists. In fact, one of the SC pairs, six of the NT pairs, and eight of the EC pairs, report no ‘learning received,’ and 6, 25, and 22 report no ‘learning given.’
Table 7: Multiplex Ties and Links
A. Number of Ties: n (% of total)
|
Learning Received |
Multiplex |
Uniplex |
None |
Total No. Ties |
|
SC |
40 (49) |
40 (49) |
1 (1) |
81 |
|
NT |
57 (48) |
55 (47) |
6 (5) |
118 |
|
EC |
34 (49) |
28 (40) |
8 (11) |
70 |
|
Learning Given |
|
|
|
|
|
SC |
33 (41) |
42 (52) |
6 (7) |
81 |
|
NT |
41 (35) |
52 (44) |
25 (21) |
118 |
|
EC |
9 (13) |
39 (56) |
22 (31) |
70 |
B. Number of Relational Links: n (% of total)
|
Learning Received |
Multiplex |
Uniplex |
Total No. |
|
SC |
94 (70) |
40 (30) |
134 |
|
NT |
131 (70) |
55 (30) |
186 |
|
EC |
80 (74) |
28 (26) |
108 |
|
Learning Given |
|
|
|
|
SC |
73 (53) |
42 (47) |
115 |
|
NT |
93 (64) |
52 (36) |
145 |
|
EC |
21 (35) |
39 (65) |
60 |
6.4. Ties Inside and Outside the Core Network
Discussion so far has considered all ties maintained by respondents, without considering knowledge exchanges with team versus non-team members. While this has indicated the range and loading of relations across groups, it has not addressed the way attention is given to in-team interactions. Another way to look at these results is to see where information is flowing with respect to the “core” team. This shows where individual’s attention is given to intra- versus extra-team interactions, and how knowledge flows inside and outside the core team.
To examine in-team and out-team interaction, it is first necessary to define the team. At first glance this seems to be an easy definition based on team membership. However, in looking at research team membership, there are often great difficulties in identifying who actually belongs to the team: students, post-docs, and occasional visitors often do not appear on membership lists; team members come and go over the life of the project, participating to greater or lesser extent at any time; and websites are not kept up to date to list new members, and rarely maintain public records of team member history. Thus, determining the whole network can be problematic.
For these teams, membership could be assessed by using the names of team members as well as could be determined, using just the names of respondents, or using the network data to reveal a ‘found’ core. Using team member listings was rejected for the reasons given above; and using just the network of respondents was rejected because it would miss actors who were important to many in the team but were unavailable for interviews. Thus, the network data were used to reveal a ‘found’ core network.
6.4.1. Core Network: The ‘found’ core network centers on the respondents and the team. It is defined to include those with whom at least two respondents report a tie. Correspondingly, a non-core or outside network is defined from those not in the core, which is those with whom only one respondent reported a tie. This definition of the core has the advantage of including those who are important to participants, yet not actually part of the named team. This can be particularly important when considering what information and learning supports team efforts. Of course, the completeness of this network is tempered by how much the data collection for 5 to 8 others captures all relevant connections, as well as limitations associated with the number of respondents per team.
For SC, the ‘found’ core includes all classweb identified on the official core list, including two team members who could not be interviewed. It also includes two others who do not appear on the official team lists (as found on the team’s website). In EC, the core includes 3 classweb not interviewed (one faculty who would have been counted as central to the team, and two who would not: a faculty associate and an administration person). The NT core picked up six members not officially part of the team (2 students, and 4 faculty); all members considered officially in the team appear in the core.
Table 8 shows the tie data for the core and outside networks. The core networks include 16 to 22 others, averaging one to two others per respondent (1.23 to 1.38 pairs per respondent across the three teams). There is, however, considerable shared connection, with around 4 ties per respondent to members of the core network, i.e., although there are barely more classweb in the network than respondents, each respondent maintains ties to an average of 4 others in this network (with the caveat that the average of four may represent an arbitrary upper limit based on the data collection on 5-8 others.) Densities are still low for the core as a whole (SC: .188; NT: .147; EC: .238, based on 240, 462, and 240 possible ties), indicating that working relationships are spread sparingly across the network. EC shows the highest density among core members with 24% of pairs interconnected (density .238).
For the non-core or outside networks, members of SC and NT hold the most contacts, with an average of approximately three co-workers who are outside the core network with whom they alone maintain a connection. By contrast, EC members only indicate one person each outside the core network with whom they maintain a close working relationship.
Table 8: Core and Outside Core Networks
A. Core: Network members named by
at least two respondents
|
|
Number of alters (% of all alters) |
Avg no. of alters per respondent
|
No. of ties (% of all ties) |
Avg no. of ties per respondent |
|
SC |
16 (31) |
1.33 |
45 (56) |
3.75 |
|
NT |
22 (31) |
1.38 |
68 (58) |
4.25 |
|
EC |
16 (55) |
1.23 |
57 (81) |
4.38 |
B. Outside: Network members named by only one respondent
|
SC |
36 (69) |
3.00 |
36 (44) |
3.00 |
|
NT |
50 (69) |
3.13 |
50 (42) |
3.13 |
|
EC |
13 (45) |
1.00 |
13 (19) |
1.00 |
C. Full Network and total numbers of ties and links
|
SC |
52 (100) |
4.33 |
81 (100) |
6.75 |
|
NT |
72 (100) |
4.50 |
118 (100) |
7.38 |
|
EC |
29 (100) |
2.23 |
70 (100) |
5.38 |
Note:
By definition there is only one tie with each “outsider”; hence the number of ties
is the same as the number of relations for the Outside network.
6.4.2 Ties and Links in Core and Outside Networks: Another question to ask about knowledge flow is how much attention in terms of relations maintained is given to core versus outside networks? This is important for understanding both how team interactions sit within an individual’s overall responsibilities, and how much attention is given to team activities. Table 9 shows the distribution of ties and links for each of the three teams. Only ‘learning received’ relations are examined.
Comparing the core and outside network shows that for SC and NT there are 12 and 16% more ties to core members than outside (SC: 56% vs 44%; NT: 58% vs 42%), and 18 and 20% more links, suggesting somewhat stronger learning relationships with core members (SC: 59 vs 41%; NT: 60 vs 40%). By contrast, EC centralizes its learning, with 61% more ties in the core (81 vs 19%) and 64% more links in the core (82% vs 18%), showing much more commitment to in-core relationships.
Table 9: Ties and Links in Core and Outside Networks (n and % of total ties or links)
|
|
Core |
Outside |
Total |
|||
|
|
Ties |
Links |
Ties |
Links |
Ties |
Links |
|
SC |
45 (56) |
79 (59) |
36 (44) |
55 (41) |
81 |
134 |
|
NT |
68 (58) |
112 (60) |
50 (42) |
74 (40) |
118 |
186 |
|
EC |
57 (81) |
89 (82) |
13 (19) |
19 (18) |
70 |
108 |
Are these differences due to the interdisciplinary nature of the groups? This is hard to say, but SC and NT, despite the different fields they are in, show themselves more similar than EC. To reiterate, the differences are the interdisciplinary nature of SC & NT, whereas EC contains members with subspecialty expertise in one field. SC & NT are geographically more dispersed than EC. Differences may be due to a combination of the in-group/out-group and onside/offsite dynamics. Further examination is necessary to pin down the reasons for these differences. What matters about this distribution is that plans for groups support, and expectations about team interaction, need to look different because of the different attention patterns in these groups.
Discussion
These data give a initial look at the kinds of knowledge flows in interdisciplinary teams. While further examination of roles and reasons for distributions is still needed, this work shows some important aspects of knowledge flow in these teams.
Looking in detail at both the types and distributions of relations that maintain teams reveals key exchanges that support group work, and the place and importance of these various kinds of knowledge in the exchanges among group members. An early assumption that domain-specific knowledge is of most importance to interdisciplinary groups quickly gives way to a more encompassing view that highlights factual, process, method and technology learning relating to the work at hand.
Given the importance of the generation of new ideas as the kind of benefits expected from interdisciplinary work, it is gratifying to find it appearing as a distinct kind of knowledge flow. This verifies that it exists in these groups, and provides a first indication of its prevalence relative to other work involved in getting projects completed. Although not as prevalent as other kinds of knowledge exchange, its frequency suggests it is an ongoing part of work practice in these groups.
Less expected were the reports of knowledge exchange about socialization into the profession, networking benefits, and administrative work. These highlight important aspects of knowledge flow in such teams and the kinds of exchanges that need to be supported.
The distributions of learning relations reveal other aspects of knowledge flow. First, relations appear to be relatively independent of each other, i.e., that pairs are more involved in sharing similar knowledge than in crossing partitions in work practices. This suggests that knowledge is pooled more frequently between those who work on similar parts of projects. While this has to be examined more thoroughly, it suggests that knowledge may flow across disciplinary boundaries along lines of practice, with the learning exchanges given here as a first indication of the lines of practice involved in interdisciplinary work.
References
[1] Brown,
J.S., Collins, A. & Duguid, P. (1989). Situated cognition and the culture
of learning. Educational Researcher,
18(1), 32-42.
[2] Cook,
S.D.N, & Brown, J.S. (1999). Bridging epistemologies: The generative dance
between organizational knowledge and organizational knowing. Organization Science, 10(4), 381-400.
[3] Cummings,
J. & Kiesler, S. (2004). KDI
Initiative: Multidiscipliary scientific collaborations.
ttp://netvis.mit.edu/papers/NSF_KDI_report.pdf
[4] Engeström,
Y. (1999). Learning by expanding: Ten years after. F. Seege. trans. Lernen durch Expansion. BdWi-Verlag,
[5] Fischer,
H., Porac, J., Wade, J.B., Brown, J., DeVaughn, M. & Kanfer, A. (2002).
Mobilizing knowledge in interorganizational alliances. In N. Bontis & C.W.
Choo (Eds.). The Strategic Management of
Intellectual Capital and Organizational Knowledge. NY: Oxford Univ. Press.
[6]Kanfer,
A., Haythornthwaite, C., Bruce, B. C., Bowker, G., Burbules, N., Porac, J.,
& Wade, J. (2000). Modeling distributed knowledge processes in next
generation multidisciplinary alliances. Information
Systems Frontiers, 2(3/4),
317-331.
[7]Klein,
J.T. (1996). Crossing boundaries : knowledge, disciplinarities, and
interdisciplinarities. Univ. Press of Virginia.
[8]Mol,
A. (2002). Cutting surgeons, walking patients: Some complexities involved in comparing.
In J.Lay & A. Mol, Complexities:
Social studies of knowledge practices. Durham, NC: Duke University Press.
[9]Nonaka,
I. (1994). A dynamic theory of organizational knowledge creation. Organization Science, 5(1),
14-37.
[10]Nonaka,
I. & Nishiguchi, T. (2001). Knowledge emergence : Social, technical, and
evolutionary dimensions of knowledge creation. New York: Oxford University
Press.
[11]Nonaka,
I. & Takeuchi, H. (1995). The
knowledge creating company. New York: Oxford University Press.
[12]Pellmar,
T.C. & Eisenberg, L. (Eds.)(2000). Bridging disciplines in the brain,
behavioral, and clinical sciences. Washington, DC: National Academy Press.
[13]Strauss,
A.L. & Corbin, J. (1990). Basics of Qualitative Research Grounded Theory
Procedures and Techniques. Newbury Park, CA: Sage.
[14]Svinicki,
M. (Feb., 2004) From the science of learning to the practice of teaching.
Paper presented at the Faculty Retreat on Active Learning, University of
Illinois at Urbana-Champaign.
[15]Wasserman,
S. & Faust, K. (1994). Social Network
Analysis. Cambridge, MA: Cambridge University Press.