SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PROPOSED METHODOLOGICAL FRAMEWORK
  5. SIGNIFICANCE AND CONTRIBUTION
  6. Acknowledgements
  7. REFERENCES

The production and creation of knowledge is not dependent on any individual and isolated entity; instead, knowledge is diffused, exchanged, and circulated among various entities. To study the knowledge flow and transfer within and across different disciplines can help us better understand science and scientific collaboration. This paper presents a methodological framework to study knowledge flow, including a knowledge hierarchy, the construction of knowledge flow network, and indicators that can be used to measure disciplinarities.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PROPOSED METHODOLOGICAL FRAMEWORK
  5. SIGNIFICANCE AND CONTRIBUTION
  6. Acknowledgements
  7. REFERENCES

The production and creation of knowledge is not dependent on any individual and isolated entity; instead, knowledge is diffused, exchanged, and circulated among various entities. Knowledge flow, in the past twenty years, has become more inter-sectoral, more inter-organizational, more interdisciplinary, and more international (Lewison, Rippon, & Wooding, 2005; Wagner & Leydesdorff, 2005; Autant-Bernard, Mairesse, & Massard, 2007; Ponds, Van Oort, & Frenken, 2007; Buter, Noyons, & Van Raan, 2010).

Similar to many important concepts in economics and bibliometrics, the transfer of knowledge is an unobservable phenomenon (Jaffe, Trajtenberg, & Fogarty, 2000). As an alternative, researchers rely on proxies to measure the concepts of interest. The quantitative studies of knowledge flow usually use citations as the research instrument. Citations between scientific articles imply a knowledge flow from the cited entity to the citing entity (Jaffe, Trajtenberg, & Henderson, 1993; Van Leeuwen & Tijssen, 2000; Nomaler & Verspagen, 2008). Using the trading metaphor (Stigler, 1994; Lockett & McWilliams, 2005; Cronin & Meho, 2008), knowledge flow has been explored as the intellectual trading among different disciplines.

The quantitative studies of interdisciplinarity were made available by researching on citation networks aggregated at the field level. Researchers usually choose a subset of representative journals or the full sets of journals from a field based upon the ISI's classification of journals, and then measure the extent to which the chosen field cites the publication of other subject categories. Network-based indicators have also been proposed to measure how interdisciplinary different research fields are. Examples include entropy (Zhang et al., 2010), integration and specialization (Porter, Roessner, Cohen, & Perreault, 2006; Porter, Roessner, & Heberger, 2008), diversity and coherence (Rafols & Meyer, 2010), percentage of multi-assignation (Morillo, Bordons, & Gomez, 2003), and relative openness (Rinia et al., 2002).

Previous endeavors on inter-sectoral, inter-organizational, and interdisciplinary knowledge flows laid sound theoretical and methodological foundations to the inquiry of knowledge flow studies. Nonetheless, these studies only involved a few disciplines as the research target, and consequently were not able to provide a holistic view of the developments and interactions of scientific disciplines. To our best knowledge, currently there is no study on finding the knowledge flows covering all science and social science disciplines. Our study is thus motivated to conduct a more comprehensive examination of the scientific trading, and obtain a bird's-eye view for the developments and interactions of various scientific disciplines.

PROPOSED METHODOLOGICAL FRAMEWORK

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PROPOSED METHODOLOGICAL FRAMEWORK
  5. SIGNIFICANCE AND CONTRIBUTION
  6. Acknowledgements
  7. REFERENCES

Knowledge hierarchy

How to effectively organize scientific knowledge is one of the key concerns of academic databases. Some databases use subject headings to classify papers (such as ACM Digital Library), while some others use subject areas to cluster journals (such as Scopus). The academic database Scopus has a well-defined journal classification schema called All Science Classification Codes (ASJC). The schema is composed of minor subject areas, major subject areas, and top-level divisions. A journal is usually assigned into one or several minor subject areas. In total, there are around 300 minor subject areas. These subject areas are grouped into 27 major subject areas, and these major subject areas are further grouped into 4 top-level divisions: Life Sciences, Physical Science, Health Sciences, and Social Sciences & Humanities. We refer to this schema as knowledge hierarchy and visualize it in Figure 1. In the proposed study, we will be focusing on the analysis of the middle four knowledge hierarchies: journals, minor subject areas, major subject areas, and top-level divisions. The limitation of this classification scheme is that such scheme may not be purely based on research specialties and topics; factors as practicality and managerial decisions may also contribute to this outcome. Comparisons with other field level clustering results, for instance, Map of Science (Boyack, Börner, & Klavans, 2009), science overlay maps (Rafols, Porter, & Leydesdorff, 2010), are highly recommended.

thumbnail image

Figure 1. A six-layer knowledge hierarchy

Download figure to PowerPoint

The construction of knowledge flow network

The raw data input is a journal-to-journal citation matrix (network), with the cell values denoting the number of citations from the citing journal to the cited journal. As each journal is associated with one (or several) minor and major subject categories, two field-to-field citation matrices (Figure 2a) can be aggregated based on the journal-to-journal citation matrix: one is on the minor subject areas and the other is on the major subject areas.

thumbnail image

Figure 2. An example of a knowledge flow network

Download figure to PowerPoint

Measurement

Knowledge flow is operationalized by citation flows among various research entities. In order to characterize the knowledge flow between disciplines, we use Dijkstra algorithm (Dijkstra, 1959) to search for shortest path between two fields in the knowledge flow network. The idea of the Dijkstra algorithm is to find a path between two nodes so that the sum of edge weight reaches the minimum in a weighted directed network.

Intuitively, the distance and traffic follow a reverse relationship. Therefore, we propose a measurement to define the distance between two fields in the knowledge flow network (Figure 2b).

  • equation image

In Figure 3, two examples (knowledge flow from E to A and knowledge flow from A to E) are given based on the sample knowledge flow network illustrated in Figure 2a. The length of a knowledge path is operationalized as the number of fields involved in a knowledge transfer. Therefore, in Figure 3, the knowledge path from Field E to Field A is two, and the knowledge path from Field A to Field E is four.

thumbnail image

Figure 3. Examples of finding shortest knowledge paths between different fields

Download figure to PowerPoint

thumbnail image

Figure 4. Proposed indicators to measure knowledge flows

Download figure to PowerPoint

Evaluative indicators

For an effective evaluation, we proposed several indicators, including average shortest path length, average shortest path weight, occurrence in shortest path, etc. For each subject area, the average shortest path length measures the average shortest path between the chosen subject area and all other subject areas. This measurement has directions: if it starts from the cited subject area, the average shortest path length denotes how easily its knowledge can be accessed by others (i.e. science/social science, classes, and subject categories); if it starts from the citing subject area, the average shortest path length denotes how easily it can access other's knowledge. The average shortest path weight is the accumulative value of the average shortest paths, and thus it is a measure of how distant a subject area is from other subject areas. The occurrence in shortest path denotes how important a subject area is to other subject categories' knowledge transfer. It is an indicator related to betweenness centrality. A subject area with higher occurrence in shortest path may have higher betweenness and thus plays a role of interconnecting various knowledge sources. Above mentioned indicators can be applied to minor subject areas, major subject areas, and the top-level knowledge divisions. At journal level, we use indicators number of citing (or cited) subject areas to measure the extent a journal cites (or been cited by) other journals. The indicators and their meanings for the hierarchies are presented in Figure 4.

SIGNIFICANCE AND CONTRIBUTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PROPOSED METHODOLOGICAL FRAMEWORK
  5. SIGNIFICANCE AND CONTRIBUTION
  6. Acknowledgements
  7. REFERENCES

Previous work on discipline interactions mainly focused on mapping science, without considering directions of knowledge flow (e.g., Boyack, Börner, & Klavans, 2005), or knowledge path (e.g., Kiss et al., 2010). In the proposed study, we intend to find important knowledge paths among all scientific disciplines, and provide empirical results on the significant patterns of knowledge transfer and dissemination. The proposed knowledge hierarchy is effective in organizing scientific knowledge. This hierarchy can be easily adopted in other related studies by other scholars. The proposed indicators quantify patterns on knowledge flow and dissemination, providing additional insights into interdisciplinary studies. These indicators are also valuable for scientific evaluation and science policy making.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PROPOSED METHODOLOGICAL FRAMEWORK
  5. SIGNIFICANCE AND CONTRIBUTION
  6. Acknowledgements
  7. REFERENCES

This project is supported by the Elsevier Bibliometric Research Program (EBRP): http://ebrp.elsevier.com/

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PROPOSED METHODOLOGICAL FRAMEWORK
  5. SIGNIFICANCE AND CONTRIBUTION
  6. Acknowledgements
  7. REFERENCES