Network analysis of knowledge building on rural extension in Colombia Received for publication: 18 June, 2016. Accepted for publication: 30 November, 2016. Doi: 10.15446/agron.colomb.v34n3.58500 1 Grupo GaMMA, Facultad de Ciencias Agrarias, Universidad de Antioquia. Medellin (Colombia). holmes.rodriguez@udea.edu.co 2 Centro de Investigaciones Económicas, Sociales y Tecnológicas de la Agroindustria y la Agricultura Mundial (CIESTAAM), Universidad Autónoma Chapingo. Texcoco, Mexico. Agronomía Colombiana 34(3), 393-402, 2016 Network analysis of knowledge building on rural extension in Colombia Análisis de redes de generación de conocimiento en la extensión rural en Colombia Holmes Rodríguez1, Carlos Julián Ramírez-Gómez1, Norman Aguilar-Gallegos2, and Jorge Aguilar-Ávila2 ABSTRACT RESUMEN Based on the analysis of scientific papers published on rural extension in Colombia since 2010, an interpretive descriptive study was conducted to identify the level of collaboration be- tween authors and institutions in the creation, systematization and dissemination of knowledge in rural extension. Informa- tion was gathered from a search in bibliographic databases to identify papers published in rural extension. 50 papers were found. They were organized in a database, and using social network analysis, a review of relational structures and indi- cators derived from the scientific collaboration between the authors and institutions involved in the publication conducted. Authors from 28 different institutions have participated in the 50 papers identified, 70% of them have been published by re- searchers working in the same institution. The findings of this study support the conclusion that actors building knowledge on rural extension in Colombia have a limited intra and inter- institutional articulation, making it urgent to strength public policies and incentives to foster relationships between research groups and institutions. A partir del análisis de la publicación de artículos científicos sobre extensión rural en Colombia desde el 2010, se realizó un estudio de carácter descriptivo interpretativo para iden- tificar el nivel de colaboración entre autores e instituciones en la generación, sistematización y difusión de conocimiento sobre extensión rural. La información se recopiló a partir de la búsqueda en bases de datos bibliográficas para identificar los artículos publicados sobre extensión rural. Se localizaron 50 artículos, los cuales se ordenaron en una base de datos y con el uso del análisis de redes sociales se revisaron las estruc- turas relacionales e indicadores derivados de la colaboración científica entre los autores e instituciones involucrados en la publicación. En los 50 artículos identificados, han participado autores de 28 instituciones diferentes; el 70% han sido publica- dos por investigadores que pertenecen a la misma institución. Los hallazgos de este estudio permiten concluir que los actores que generan conocimiento sobre extensión rural en Colombia presentan una escasa articulación intra e interinstitucional lo cual hace apremiante el fortalecer las políticas públicas y los incentivos para fomentar los relacionamientos entre los grupos de investigación y entre las instituciones. Key words: knowledge management, researchers’ networks, social network analysis, rural development. Palabras clave: gestión del conocimiento, redes de investiga- dores, análisis de redes sociales, desarrollo rural. Rivera and Sulaiman (2009) indicate that extension was originally conceived as part of a “knowledge triangle”, formed by research, education and extension. However, today it is addressed in a more comprehensive manner and is valued by various actors participating in rural develop- ment, not only in the context of improving productivity, but also for its contribution in strengthening bonds between farmers, researchers, agricultural education institutions, and other actors in society (Faure et al., 2012), that some- how form what could be called an “innovation system”, through actors interacting in a process of generation, dissemination and use of knowledge in order to increase agricultural production looking for economic and social changes (Hellin, 2012). Introduction According to Christoplos (2010), “extension” can be under- stood as the systems that facilitate the access of farmers, their organizations, and other market players to knowledge, technologies and information. Extension encourages their integration with research members, teaching, agro-indus- try and other institutions; and contributes to the design of practices and technical, management and organizational skills. At an international level, extension services among other factors, are recognized as key points for the develop- ment of the agricultural activity (Kilelu et al., 2014; Klerkx and Leeuwis, 2009; Muñoz and Santoyo, 2010). 394 Agron. Colomb. 34(3) 2016 In this regard, Universities and National Research Centers as well as public and private institutions have a strategic role within the process of building codified knowledge that is regularly measured through the publication of scientific research products and patents (Rivera-Huerta et al., 2011). Furthermore, in the case of the agricultural sector, these ac- tors have the mission of strengthening extension, improv- ing their innovation capabilities, through interactions and coordination to create new information articulated with the demand (Klerkx and Leeuwis, 2008; Spielman and Birner, 2008), building networks for strengthening relations and ties, which increase the production and dissemination of knowledge (Aguilar-Gallegos et al., 2015, 2016; Brunori et al., 2013; Vega de Jiménez and Rojo, 2010). This approach replaces the linear view of knowledge cre- ation and innovation, by an interactive process between different actors in agricultural innovation systems (Mu- ñoz and Santoyo, 2010). In fact, in the case of Colombia, this approach was adopted by the Ministry of Agriculture and Rural Development (MADR) setting up the National Subsystem of Agricultural Technical Assistance (SSATA), tied to the National System of Agricultural Science and Technology (SNCTA), in order to coordinate actors to improve the development and dissemination of knowledge (MADR, 2012). At the international level, there is a consensus through university researchers and their research groups about scientific collaboration between universities and other actors playing a key role in the progress of knowledge (De Stefano et al., 2013), since networking allows sharing ideas, methodologies and approaches, which may help provide solutions to common problems. However, in Colombia the empirical evidence indicates a low articulation between actors; therefore, it is important to conduct studies in or- der to determine levels of cooperation and coordination between them. According to the findings of the Mission for Rural Transformation (Misión para la Transformación del Campo, in Spanish) (2015), the promotion of networking and development of capabilities, should generate knowledge management strategies to achieve greater impact in rural areas, improving performance of sectors and alternatives to strengthen rural producers. In this context and in the case of Colombia, this paper aims to analyze the level of collaboration between actors in the creation and dissemination of knowledge on rural exten- sion, using social network analysis in order to guide public policy to strengthen rural extension under an innovation system approach. Methodology An interpretative and descriptive study was conducted based on the analysis of co-authorship of scientific papers on rural extension, which is a partial indicator of scientific collaboration (Katz and Martin, 1997) between research- ers, especially in the publication of papers (Lopaciuk- Gonczaryk, 2016). A social network analysis was used, following other studies of this type (De Stefano et al., 2013; Russell et al., 2009; Valderrama-Zurián et al., 2007; Zazo et al., 2015). Documents considered by Faure et al. (2012) as “gray literature”, such as information booklets, sector reports, books, theses or other documents without peer review were not included in the analysis. Data collection and editing Information was gathered through bibliographic databases search (Dialnet, Ebsco, Redalyc, Redib, Scholar, SciELO, Scopus, Sciencedirect) to identify all papers published on rural extension in Colombia, based on the definition of rural direct technical assistance provided under Colombian law (Congreso de Colombia, 2000), which includes social, environmental, economic and technical issues; between 2010 and 2015, because during that period, the MADR revived interest in improving the quality and coverage of the ATDR service. Information of Author (Aut), institution of affiliation (Ins), year of publication, and name of the journal were linked to each paper (Art), making the ne- cessary adjustments for homonymy and synonymy results (Calero et al., 2006). Network analysis and collaboration indicators A social network analysis (SNA) was used as a tool for observing, studying and understanding the relational structures derived from scientific collaboration between authors and institutions involved in the publication of papers on rural extension; the relationship analyzed in this study is the participation of the different authors in the different papers. SNA allows to identify the positions of actors within the network, which partly determines the limitations and opportunities that those actors and the network have in general (Borgatti et al., 2013). To analyze the participation of authors in each paper, 2-mode networks were used; to analyze collaboration between authors, 1-mode networks were used (Borgatti et al., 2013; Wasserman and Faust, 1994). Adapting the indi- cations of Valderrama-Zurián et al. (2007), the number of relationships between authors participating in a paper is calculated as m!/(m-n)!n!, where m is the number of authors in the article and n the number of elements of groups. This 395Rodríguez, Ramírez-Gómez, Aguilar-Gallegos, and Aguilar-Ávila: Network analysis of knowledge building on rural extension in Colombia analysis approach was also used to analyze the collabora- tion between the authors’ institutions of affiliation. Based on Freeman (1979) and Borgatti et al. (2013), the indicators used for the network analysis were as follows: 1. Degree: number of links or relationships that a node has. Thus, the higher the degree, the higher the level of collaboration of an author. 2. Betweenness: frequency measurement of a given node when it is on the shortest path connecting other pairs of nodes. It was only measured for 1-mode networks to make reference to the relative im- portance that an author has in connecting other authors. 3. Density: measure of cohesion that makes reference to the number of existing links on the network in relation to possible links, expressed as a percentage. It was only considered for 1-mode networks. Additionally, homophily - the tendency to bond with indi- viduals who have characteristics similar to ours (Lazarsfeld and Merton, 1954) - was calculated. It was only calculated for the 2-mode network using the E-I index (external and internal links) from Krackhardt and Stern (1988), classify- ing institutions into three types: public, private and others (NGOs, trade unions and independent). Calculation of indicators and network observation were performed us- ing Ucinet (Borgatti et al., 2002) and NetDraw (Borgatti, 2002) software. Results Scientific collaboration during the period 2010-2015 115 authors participated in the 50 papers analyzed. Some of them contributed with more than one participation, for a total of 103 different authors (Tab. 1). 74% of the papers have two or more authors. Within the period analyzed, collaboration increased, as in 2010 less than half of papers were written in co-authorship and by 2015 that figure reached nearly 86%. A growing trend was found in the increase of both pub- lications and authors, additionally, there are increasingly more authors involved (Fig. 1). However, the number of new authors involved in publications on rural extension in Colombia is decreasing. The largest increase was seen in 2011, when 15 new authors joined the 19 existing ones. In the last year (2015) there were only 14% new authors. This situation may be considered normal, as the data is cumula- tive; however, there are few authors contributing with more than one collaboration, as three authors have published three articles; six have published two, and the remaining 94 have only participated in one paper. This could indicate the insufficiency of critical mass discussing this topic. Scientific collaboration network between authors It was found that authors have in general little participa- tion in several publications. There are few authors having two or three links to papers; i.e., author 007 (Aut-007), who participates in three papers (003, 027 y 031) (Fig. 2). Likewise, it can be observed that there are more papers with two or more authors than papers published by a single author. 74% of papers have been published in co-authorship, although the participation of the same author in several papers is lower. From the 2-mode network, consisting of authors and papers, it was possible to obtain a 1-mode network (Bor- gatti et al., 2013) and therewith a representation of direct collaboration between authors (Fig. 3 ). It was found that 11 authors have individually published a paper; without any collaboration. They have published 13 papers, since an author (Aut-001) published two papers individually, and another author (Aut-017) published one individually and other in collaboration. The maximum number of co- authorship in a paper was 5, in two different cases. Three papers have been published by four authors. The most frequent collaborations occur between two and three au- thors. In both cases 16 papers have been published by that number of authors. TABLE 1. Scientific collaboration in the production of papers 2010-2015. Year No. of papers No. of papers written in co-authorship (%) Total No. of authors Authors per paper Maximum No. of authors in a paper New and different authors Increase in new authors (%) 2010 13 6 (46.2) 21 1.6 3 19 — 2011 7 6 (85.7) 16 2.3 3 15 78.9 2012 8 5 (62.5) 19 2.4 5 17 50.0 2013 8 7 (87.5) 21 2.6 4 19 37.3 2014 7 7 (100.0) 20 2.9 5 20 28.6 2015 7 6 (85.7) 18 2.6 4 13 14.4 Total 50 37 (74.0) 115 2.3 5 103 396 Agron. Colomb. 34(3) 2016 120 100 80 60 40 20 0 201620142013 20152012201120102009 N um be r Year Cumulative articles Cumulative authors Aut-091 Art-006 Aut-069 Art-012 Aut-070 Art-014 Aut-092 Art-030 Aut-096 Art-042 Aut-098 Art-035 Aut-097 Art-005 Aut-099 Art-004 Aut-086 Art-032 Aut-026 Art-037 Aut-027 Art-013 Aut-051 Art-035Aut-052 Art-045 Aut-055 Art-022 Aut-054 Art-024 Aut-053 Art-025 Aut-087 Art-019 Aut-089 Art-018 Aut-088 Art-041 Aut-072 Art-020 Aut-085 Art-046 Aut-084 Art-021 Aut-083 Art-050 Aut-090 Aut-011 Aut-013 Aut-035Aut-062 Aut-063 Aut-064 Aut-021 Aut-043 Aut-044 Aut-071 Aut-014 Aut-012 Aut-019Aut-025 Aut-006 Aut-057 Aut-058 Aut-056 Aut-020 Aut-022 Aut-034 Aut-033 Aut-067 Aut-066Aut-065 Aut-012 Art-026 Art-023 Art-043 Art-048 Art-029 Art-016 Art-009 Art-015 Art-040 Art-010 Art-007 Art-008 Art-001 Art-011 Art-017 Aut-013 Aut-028 Aut-040 Aut-041 Aut-042 Aut-047 Aut-073 Aut-076 Aut-075 Aut-074 Aut-005 Aut-037 Aut-039 Aut-036 Art-038 Art-034 Art-047 Aut-038 Aut-068 Aut-103 Aut-102 Aut-094 Aut-095 Aut-093 Aut-081 Aut-082 Aut-080 Aut-100 Aut-101 Aut-059 Aut-060 Aut-061 Aut-078 Aut-077 Aut-079 Aut-031 Aut-030 Aut-024Aut-023 Aut-016 Aut-017 Aut-049 Aut-050 Aut-048 Aut-018 Art-049 Art-002 Art-036 Art-028 Art-027 Art-003 Art-031 Art-033 Art-044 Aut-003 Aut-004 Aut-002 Aut-006 Aut-045 Aut-046 Aut-029 Aut-007Aut-032 Aut-008 Aut-006 Collaboration has been based, almost entirely, on the publication of one article. This occurs because interaction between pairs of authors is limited to a single time (weak ties). Only two pairs of authors (Aut-007 and Aut-008; Aut- 010 and Aut-011) have collaborated twice (strong ties) in the publication of two different papers. In both cases, other authors have participated in the publication of those papers. There are few authors who manage to connect different collaborations. This is the result of publishing different papers with different authors. Take Aut-040 for example, who manages to connect authors 041, 042 and 047. This author published one paper with the first two, and another one with the latter. Only five authors of this type (black nodes) were found across the entire collaboration network. Network indicators show that 103 authors have managed to establish 200 links between them. Therefore, network density is low, as well as the degree average of each author (Tab. 2). The last indicator shows that, on average, each author has collaborated with nearly two authors. This is supported by the fact that collaboration between two and three authors is quite common. However, the collaboration network is fragmented as there are 40 components. The best connected component links only six authors. This is also the reason why the network diameter is low. In fact, the last indicator is achieved through any of the five nodes FIGURE 2. 2-mode network of collaborators. Authors (circles); papers (triangles). FIGURE 1. Paper publishing and author par ticipation trends. 397Rodríguez, Ramírez-Gómez, Aguilar-Gallegos, and Aguilar-Ávila: Network analysis of knowledge building on rural extension in Colombia that serve as intermediary between different collaborations (Fig. 3, black nodes). That is, although 74% of the 50 papers have been written in collaboration, there is not a real dense collaboration network between all the authors. Likewise, there are no authors having a significant centrality; network centrality is only 3%. TABLE 2. Basic indicators of the collaboration network between authors. Indicator Value N 103 Links 200 Average degree 1.942 Density (%) 3.807 Components 40 Diameter 2 Average distance 1.130 Network centrality (%) 3.060 Scientific collaboration network between institutions The participation of authors from 28 different institutions was found. 70% of papers analyzed were published by re- searchers belonging to the same institutions, that is, there was no inter-institutional collaboration. The remaining 30% were published in institutional collaboration. It is worth mentioning that inter-institutional collaboration was analyzed, that is, collaboration between different institutions, but intra-institutional collaboration: between different departments, faculties, specialties of the same institution, etc., was not considered, as it is common for authors to include only the main institution to which they are affiliated. Participation analysis of authors from 28 institutions (circles) with 50 papers (triangles) shows the institutions of affiliation of the authors with most publications (Fig. 4). Authors from Ins-001 have participated 17 publications, out of which 8 have been in collaboration with other institu- tions. The participation of each of the authors’ institutions of affiliation and their collaboration with other institutions, that is, papers with two or more links, can be observed. This is significant since, just as there are institutions with prolific authors, there are also others with fewer publishing authors, and most of these publications are made in col- laboration. Take the case of Ins-025 who has four papers, three out of which were published in collaboration. There Aut-013 Aut-018 Aut-019 Aut-020 Aut-021 Aut-022 Aut-025 Aut-035 Aut-072 Aut-001 Aut-092 Aut-068 Aut-102Aut-103 Aut-060 Aut-059 Aut-061 Aut-096 Aut-097 Aut-098 Aut-036Aut-037 Aut-038Aut-081 Aut-082 Aut-080 Aut-076 Aut-075 Aut-074 Aut-073 Aut-101 Aut-100 Aut-039 Aut-063 Aut-062 Aut-064 Aut-067 Aut-010 Aut-011 Aut-012 Aut-065 Aut-066 Aut-079 Aut-078 Aut-077 Aut-044 Aut-043 Aut-083 Aut-085 Aut-084 Aut-028 Aut-050 Aut-048 Aut-049 Aut-051 Aut-052 Aut-054 Aut-055 Aut-053 Aut-045 Aut-029 Aut-046 Aut-095 Aut-087Aut-088 Aut-089 Aut-056 Aut-058 Aut-057 Aut-094 Aut-090 Aut-094 Aut-093 Aut-086 Aut-008 Aut-009 Aut-0047 Aut-042 Aut-041 Aut-040 Aut-007 Aut-099 Aut-032Aut-026 Aut-031 Aut-033 Aut-027 Aut-017 Aut-016 Aut-006 Aut-003 Aut-004 Aut-005 Aut-002 Aut-070 Aut-071 Aut-069 Aut-030 Aut-031 Aut-014 Aut-015 Aut-023 Aut-024 FIGURE 3. 1-mode network of collaboration between authors. 398 Agron. Colomb. 34(3) 2016 is also Ins-007 with three papers, all of them in collabora- tion. Finally, there are institutions whose authors do not publish in collaboration (bottom right of the graph) or whose authors have published several papers, but none in collaboration; for example, Ins-002. Some points of interest were found when turning the 2-mode network (Fig. 4) into 1-mode between institutions (Fig. 5). Nine institutions of affiliation appear isolated, since the papers published by their authors were not written in collaboration. There is a very central institution of affilia- tion (Ins-001), who managed to establish links with other 10 institutions from the 8 papers published in collaboration. This fact is related to the importance of the institution in Colombia. Nevertheless, most of the collaborations have oc- curred between two institutions of affiliation. There is only one paper with collaborations from three institutions, and another one from four. Only in one case, between Ins-001 and Ins-025, two papers were published in collaboration (thicker line). This implies that inter-institutional research needs strengthening. Just as there are authors who manage to link other ac- tors, i.e. intermediaries between collaborations, several institutions were also found to play this role with their participation in two or more papers. The diamond-shaped circles represent the institutions that play this role within the network. Out of the 28 institutions, only 6 actors of this type were found. The calculation of indicators of the previous network revealed that the 28 institutions have established 42 collaborative links; therefore, each one of them has an average of 1.5 links (Tab. 3). Network density also serves as an indicator to measure the level of articulation and the number of existing links, which is 11.1%. This type of indicators would serve as a baseline to analyze in the future the evolution of articulation within the network. In comparison to the network of authors, this one ref lects a greater articulation, as there are fewer components. A fact to note is that the best connected component manages to link directly or indirectly 60.7% of all institutions, but this is achieved with a higher network diameter, that is five steps, and also for that reason, the average distance is 2.6. Network centrality is 33.9%, which is very visible because of the importance that the institution of affilia- tion Ins-001 has within the network. Ins-018 Ins-026Ins-025 Ins-028 Ins-027 Ins-004 Ins-029 Ins-016 Ins-012 Ins-006 Ins-009 Ins-014 Ins-013 Ins-023 Ins-024 Ins-002 Ins-008 Ins-007 Ins-015 Ins-011 Ins-017 Ins-031 Ins-010 Ins-005 Art-020 Art-001 Art-006 Art-015 Art-049 Art-023 Art-026 Art-017 Art-019 Art-018 Art-025 Art-024 Art-021 Art-007 Art-010 Art-041 Art-006 Art-005 Art-037 Art-039 Art-038 Art-031 Art-034 Art-027 Art-003 Art-013 Art-030 Art-014 Art-028 Art-036 Art-033 Art-032 Art-046 Art-047 Art-008 Art-022 Art-011 Art-002 Art-016 Art-035 Art-050 Art-048 Art-029 Art-040 Art-044 Art-012 Art-043 Art-045 Art-042 Ins-030 Ins-019 Art-009 Ins-003Ins-001 FIGURE 4. 2-mode network of collaboration. Institutions (circles); papers (triangles). 399Rodríguez, Ramírez-Gómez, Aguilar-Gallegos, and Aguilar-Ávila: Network analysis of knowledge building on rural extension in Colombia TABLE 3. Basic indicators of the collaboration network between institu- tions. Indicator Value N 28 Links 42 Average degree 1.500 Density (%) 11.111 Components 11 Diameter 5 Average distance 2.569 Network centrality (%) 33.900 Although the network of institutions of affiliation is denser, it is important to mention that this is partly because the network is smaller, and on the other hand, because it is the result of collaboration between the authors that have published the papers. In fact, if strategies to increase col- laboration and the number of authors in each paper were implemented, it would result in a greater collaboration between institutions. It was found a slight level of homophily on the entire net- work, with an E-I index of -0.048 (Tab. 4). Public institu- tions have more links between them than with the other types of institutions, however, homophily is more prevalent as the E-I index is -0.286. Besides, there is a higher density of links between them. Meanwhile, there are no links be- tween private institutions, i.e. they tend to heterophily (E-I index of 1.000). It is interesting to see that their collabora- tive links have been established with public institutions and not with any other type. The group of other institutions has links between them and with public institutions, but not with private ones. However, as density of links is higher to the inside instead of to the outside, then the E-I index of this group is also negative (-0.143). DISCUSSION The results of this study indicate that in Colombia knowl- edge building networks on rural extension are limited; the calculated indicators show a low participation of authors and institutions in collaboration networks; neverthe- less, the causes of this phenomenon should be deepenly studied. In this sense, authors like Ahrweiler and Keane (2013) find that promoting a networking approach implies complexity, related with the existence of behavior, action and communication models; actors who may present com- patibilities and incompatibilities, communicative interest or perhaps different strategic perspectives. Such is the case of homophily as an attribute present in all the structures of the studied networks, and where the phenomenon is evident between universities and other public and private institutions; this behavior could inf luence structural links, Ins-004 Ins-029 Ins-016 Ins-018 Ins-003 Ins-005 Ins-026 Ins-012Ins-007 Ins-015 Ins-008 Ins-025 Ins-027 Ins-028 Ins-010 Ins-011 Ins-030 Ins-031 Ins-002 Ins-017 Ins-009 Ins-006 Ins-023 Ins-013 Ins-024 Ins-019 Ins-014 Ins-001 FIGURE 5. 1-mode network of collaboration between institutions. Color legends: blue nodes: public institutions; red nodes: private institutions; green nodes: other type (NGOs, independent and trade unions). Legend of shapes: diamond nodes: intermediating actors; circular nodes: non- intermediating actors. 400 Agron. Colomb. 34(3) 2016 generate processes of social selection, among others. This may hinder processes of building and dissemination of knowledge, as already addressed in other studies (Isaac, 2012; McPherson et al., 2001). The lack of networking reveals a limitation of institutions of higher education in strengthening innovative capacity of the agricultural productive sector and the direct rural tech- nical assistance service due to the lack of generation and dissemination of knowledge on rural extension. On that subject, other authors state that universities are strategic players in the innovation system, thanks to their knowledge of local reality derived from their regional presence. They also have an additional strategic capacity for research, through the theses of their undergraduate and graduate students (Fonseca and Rugeles, 2004); nevertheless, at pres- ent, large part of the knowledge generated is not properly systematized and ends up falling into the “gray literature” (Faure et al., 2012). For that reason, the number of scientific papers published is low in comparison with the number of students. Therefore, it is important to implement strategies for the codification of generated knowledge. The centrality indicator shows the lack of authors that have an important centrality. However, at the level of institu- tions, results show a greater connection, probably due to the smaller number of these compared with the number of authors, a better connection, a higher diameter and a higher centrality indicator around the National University of Colombia. This implies an opportunity to implement strategies to promote networking, because the more central actor, as suggested Barrientos-Fuentes and Berg (2013), could generate feedback mechanisms between research institutions, to improve the development and dissemina- tion of innovations on rural extension. Thus, the consolidation of knowledge building networks on rural extension between authors and institutions should be part of the strategies in the implementation of innovation systems within the agricultural sector, in a way that spaces can be generated for collaboration between institutions, and also effective strategies to support networking research programs for generation and dissemination of knowledge on rural extension; because as De Stefano et al. (2013) highlight, scientific collaboration between universities and other actors is important in the progress of knowledge. In this sense, knowledge building networks can be impor- tant for agricultural development. Several Initiatives begin to emerge as the Red Nacional de Extensión Rural [National Network of Rural extension], led by the University of An- tioquia; and the Red de Estudios Rurales [Network of Rural Studies], led by the University of Tolima. Both can serve as platforms for systematizing and publishing successful experiences on rural extension, that can be imitated in other departments of the country, as in the case of the University of Tolima (Ibague, Colombia) with the intern- ship program for service delivery of ATDR, financed by the regional government (Serrano et al., 2015). Accordingly, given the observed shortage of scientific pro- duction in matters of extension and technical assistance in Colombia, it would be relevant to conduct a cause analysis, including one of human capital skills in the use of tools for systematization and publication of papers, because many institutions engaged in rural extension does not publish papers. Nevertheless, there is a growing increase in research activities carried out by associations of pro- ducers and universities, in the diversification of institu- tional structures and the focus of agricultural research. This latter, in the case of the Agronomía Colombiana journal of the National University has led to an increase of 13.64% in the publications in the field of agricultural economy and rural development in the period 2003-2012 (Ligarreto, 2013). Finally, it is considered important to conduct studies to determine the way in which that knowledge is used by analyzing the effect or impact of such publications, either by the number of citations received or the operability of their proposals. However, this research contributes to con- struct baseline indicators that could be used to assess in the future the improvement in the articulation of researchers and institutions addressing this subject. TABLE 4. Density matrix and E-I index by group of institutions. Group n Density E-I index by group E-I index of network Inst. Public Inst. Private Other Inst. Inst. Public 13 11.500 7.700 2.900 -0.286 -0.048Inst. Private 7 7.700 0.000 0.000 1.000 Other Inst. 8 2.900 0.000 7.100 -0.143 401Rodríguez, Ramírez-Gómez, Aguilar-Gallegos, and Aguilar-Ávila: Network analysis of knowledge building on rural extension in Colombia Conclusions The findings of this study lead to the conclusion that ac- tors that build knowledge on rural extension in Colombia have a poor intra and inter-institutional articulation. For the foregoing, strengthening public policies and incentives becomes urgent to foster relationships between research groups and between institutions. In this way, it will contrib- ute to the consolidation of the new collaboration networks that can serve as platforms for dissemination and especially for the use of knowledge on rural extension, strengthening the role of researchers, universities and research centers in shaping the territorial systems of innovation. The strengthening of mechanisms to connect researchers in matters of extension and technical assistance should start from recognizing their theoretical and methodological capacities. In this regard, the SNA approach can become an important tool for monitoring the impact of implemented actions that seek to strengthen collaboration for building knowledge on rural extension. Although it is a challenge in gathering information and continuous analysis, this type of longitudinal analysis would be a very useful tool for both universities and public institutions responsible for guiding science, technology and innovation policies. Acknowledgments The authors thank the Committee for the Development of Research (CODI) for the financial support and to Strategies for Sustainability 2016 of the GaMMA Group, University of Antioquia, Medellin, Colombia. Literature cited Aguilar-Gallegos, N., E.G. Martínez-González, J. Aguilar-Ávila, H. Santoyo-Cortés, M. Muñoz-Rodríguez, and E.I. García- Sánchez. 2016. Análisis de redes sociales para catalizar la innovación agrícola: de los vínculos directos a la integración y radialidad. Estudios Gerenciales 32(140), 197-207. Doi: 10.1016/j.estger.2016.06.006 Aguilar-Gallegos, N., M. Muñoz-Rodríguez, H. Santoyo-Cortés, J. Aguilar-Ávila, and L. Klerkx. 2015. Information networks that generate economic value: A study on clusters of adopters of new or improved technologies and practices among oil palm growers in Mexico. Agric. Syst. 135, 122-132. Doi: 10.1016/j. agsy.2015.01.003 Ahrweiler, P. and M. Keane. 2013. Innovation networks. Cognit. Stud. Econ. Soc. Sci. 12, 73-90. Doi: 10.1007/s11299-013-0123-7 Barrientos-Fuentes, J.C. and E. Berg. 2013. Impact assessment of agricultural innovations : a review. Agron. Colomb. 31(1), 120-130. Borgatti, S.P. 2002. Netdraw network visualization. Analytic Tech- nologies, Harvard, MA. Borgatti, S.P., M.G. Everett, and L.C. Freeman. 2002. Ucinet for Windows: software for social network analysis. Analytic Technologies. Harvard, MA. Borgatti, S.P., M.G. Everett, and J.C. Johnson 2013. Analyzing social networks. SAGE Publications, London, U.K. Brunori, G., D. Barjolle, J. Ingram, and L. Klerkx. 2013. CAP reform and innovation: the role of learning and innovation networks. EuroChoices 12(2), 27-33. Doi: 10.1111/1746-692X.12025 Calero, C., R. Buter, C.C. Valdés, and E. Noyons. 2006. How to iden- tify research groups using publication analysis: An example in the field of nanotechnology. Scientometrics 66(2), 365-376. Doi: 10.1007/s11192-006-0026-z Christoplos, I. 2010. Cómo movilizar el potencial de la extensión agraria y rural. Foro mundial sobre servicios de asesoramiento rural. FAO, Rome, Italy. Congreso de Colombia. 2000. Ley 607 de 2000. In: https://www. minagricultura.gov.co/Normatividad/Leyes/ley_607_00.pdf, consulted: January, 2015. De Stefano, D., V. Fuccella, M.P. Vitale, and S. Zaccarin. 2013. The use of different data sources in the analysis of co-authorship networks and scientific performance. Soc. Networks, 35(3), 370-381. Doi: 10.1016/j.socnet.2013.04.004 Faure, G., Y. Desjeux, and G. Pierre. 2012. New challenges in agricul- tural advisory services from a research perspective: A literature review, synthesis and research agenda. J. Agric. Education Ext. 18(5), 461-492. Doi: 10.1080/1389224X.2012.707063 Fonseca, S. and L. Rugeles. 2004. De lo agropecuario a lo agroindus- trial: Una visión desde la ciencia, la tecnología y la innovación. CeniRED; Colciencias, Bogotá. Freeman, L.C. 1979. Centrality in social networks: conceptual clarification. Soc. Networks 1(3), 215-239. Doi: 10.1016/0378 -8733(78)90021-7 Hellin, J. 2012. Agricultural extension, collective action and in- novation systems: lessons on network brokering from Peru and Mexico. J. Agric. Education Ext. 18(2), 141-159. Doi: 10.1080/1389224X.2012.655967 Isaac, M.E. 2012. Agricultural information exchange and organiza- tional ties: The effect of network topology on managing agro- diversity. Agric. Syst. 109, 9-15. Doi: 10.1016/j.agsy.2012.01.011 Katz, J.S. and B.R. Martin. 1997. What is research collaboration? Res. Policy 26(1), 1-18. Doi: 10.1016/S0048-7333(96)00917-1 Kilelu, C.W., L. Klerkx, and C. Leeuwis. 2014. How dynamics of learn- ing are linked to innovation support services: Insights from a smallholder commercialization project in Kenya. J. Agric. Edu- cation Ext. 20(2), 213-232. Doi: 10.1080/1389224X.2013.823876 Klerkx, L. and C. Leeuwis. 2008. Matching demand and supply in the agricultural knowledge infrastructure: Experiences with innovation intermediaries. Food Policy 33, 260-276. Doi: 10.1016/j.foodpol.2007.10.001 Klerkx, L. and C. Leeuwis. 2009. Shaping collective functions in privatized agricultural knowledge and information systems: the positioning and embedding of a network broker in the Dutch dairy sector. J. Agric. Education Ext. 15(1), 81-105. Doi: 10.1080/13892240802617445 Krackhardt, D. and R.N. Stern. 1988. Informal networks and orga- nizational crisis: an experimental simulation. Soc. Psychol. Quart. 51(2), 123-140. Doi: 10.2307/2786835 402 Agron. Colomb. 34(3) 2016 Lazarsfeld, P. F. and R.K. Merton. 1954. Friendship as a social pro- cess: A substantive and methodological analysis. pp. 18-66. In: Berger, M., T. Abel, and C.H. Page (eds.). Freedom and control in modern society. D. Van Nostrand Co., New York, NY. Ligarreto, G.A. 2013. Considerations on the research and dissemina- tion of agricultural knowledge by the Facultad de Agronomía. Agron. Colomb. 31(2), 243-252. Lopaciuk-Gonczaryk, B. 2016. Collaboration strategies for publish- ing articles in international journals – A study of Polish scien- tists in economics. Social Networks 44, 50-63. Doi: 10.1016/j. socnet.2015.07.001 MADR. 2012. Subsistema nacional de asistencia técnica agro- pecuaria. In: http://www.adelnarino.org/2011/web/descargas/ Subsistema_Asistencia_Tecnica_Agropecuaria.pdf; consulted: August, 2015. Mcpherson, M., L. Smith-Lovin, and J.M. Cook. 2001. Birds of a feather: Homophily in social networks. Ann. Rev. Sociol. 27, 415-444. Doi: 10.1146/annurev.soc.27.1.415 Misión para la Transformación del Campo. 2015. Estrategia de ciencia, tecnología e innovación agropecuaria y de acompa- ñamiento integral. Nuevas Ediciones, Bogotá. Muñoz, M. and V.H. Santoyo. 2010. Del extensionismo a las redes de innovación. pp. 31-69. In: Aguilar, J., J.R. Altamirano, and R. Rendón (eds.). Del extensionismo a las redes de innovación. CIESTAAM, Universidad Autónoma Chapingo, Texcoco, México. Rivera, W.M. and R.V. Sulaiman. 2009. Extension: Object of reform, engine for innovation. Outlook Agric. 38(3), 267-273. Doi: 10.5367/000000009789396810 Rivera-Huerta, R., G. Dutrénit, J.M. Ekboir, J.L. Sampedro, and A.O. Vera-Cruz. 2011. Do linkages between farmers and academic researchers inf luence researcher productivity? The Mexican case. Res. Policy 40(7), 932-942. Doi: 10.1016/j. respol.2011.05.001 Russell, J.M., M.J. Madera J., and A. Shirley. 2009. El análisis de redes en el estudio de la colaboración científica. Redes. Rev. Hisp. Análisis Redes Soc. 17(2), 39-47. Serrano, R., L.E. Guzmán, and M. Jimenéz. 2015. Análisis de un modelo de apoyo a la asistencia técnica agropecuaria caso: municipio de Fresno-Tolima. p. 46. In: Proc. 1er. Congreso Internacional de Estudios Rurales, http://redestudiosrurales. wix.com; consulted: March, 2016. Spielman, D. and R. Birner. 2008. How innovative is your agri- culture?: Using innovation indicators and benchmarks to strengthen national agricultural innovation systems. The World Bank. Washington, DC. Valderrama-Zurián, J.C., G. González-Alcaide, F.J. Valderrama- Zurián, R. Aleixandre-Benavent, and A. Miguel-Dasit. 2007. Redes de coautorías y colaboración institucional. Rev. Esp. Cardiol. 60(2), 117-130. Doi: 10.1157/13099458 Vega de Jiménez, M. and Y. Rojo. 2010. Red: estructura para generar innovación. Rev. Cienc. Soc. 16(4), 699-708. Wasserman, S. and K. Faust. 1994. Social network analysis: methods and applications. Cambridge University Press, New York, NY. Doi: 10.1017/CBO9780511815478 Zazo, Á.F., S. Ardines, and E. Castro. 2015. Redes de colaboración de las unidades de investigación de la Universidad de Panamá: investigación, desarrollo e innovación. Redes. Rev. Hisp. Análisis Redes Soc. 26(2), 84-117.