Comparing the diffusion and adoption of linked data and research data management services among libraries

Introduction

Libraries have experienced numerous cycles of innovation and are likely to experience a faster pace of innovations in the future due to technology developments and increasing economic pressure. If consistent patterns in library innovations are identified, libraries and relevant stakeholders will be informed and well-prepared for future innovations. Libraries will be able to anticipate the types of challenges they will encounter in future innovation adoption and the kinds of strategies they could use to overcome obstacles. Relevant stakeholders, such as funders, professional associations, and vendors, will be able to design interventions that facilitate innovation adoption in libraries.

Consistent patterns occur when phenomena happen repeatedly in multiple innovations and across many libraries. They can be identified through investigating multiple innovations either longitudinally or cross-sectionally. This study aims to identify consistent patterns in the diffusion and adoption of innovations among libraries through investigating two recent innovations: linked data and research data management service. These two innovations are chosen for the following reasons. First, both innovations have diffused among libraries for over a decade and have reached enough saturation to investigate their diffusion and adoption patterns. Second, they represent two different kinds of innovations. linked data is primarily a technology innovation, although it can be used to power various kinds of user services. Research data management service is primarily a service innovation, although its implementation can have technological components. In addition, linked data originates outside of the library community and is being adopted in many fields. Research data management service is a new service that has emerged mostly in research libraries, although some other parties, such as data archives, university research administration and information technology, are also involved. The commonalities between the diffusion and adoption of these two very different innovations likely indicate consistent patterns that apply to many future innovations.

Theoretical framework

Diffusion of innovation theories are used as the theoretical framework to analyse the diffusion and adoption of the two innovations. Diffusion of innovation studies have been conducted for several decades in many different disciplines. A large body of theories has accumulated regarding various aspects of the generation, diffusion, implementation, and consequences of innovations. Many of these theories have been summarised in Rogers (2003). This study uses the following tenets from these theories: adopter categories, diffusion models, communication channels, innovation decisions, and adoption processes. According to Rogers, when an innovation spreads to almost all the potential adopters in a social system, adopter distributions over time approach normality, and adopters can be categorised into five groups: innovators, early adopters, early majority, late majority, and laggards. In this study, libraries are defined as adopters. There are two different models of the diffusion of innovations. In a centralised diffusion model, ‘an innovation originates from some expert source (often a research and development organization)’ and then ‘diffuses as a uniform package to potential adopters who accept or reject the innovation’ (p. 395). In a decentralised diffusion model,

innovations often bubbled up from the operational levels of a system, with the inventing done by certain lead users. Then the new ideas spread horizontally via peer networks, with a high-degree of reinvention occurring as the innovations are modified by users to fit their particular conditions. (p. 395).

Communication channels through which innovations are diffused can be categorised as (1) interpersonal (face-to-face exchange between individuals) versus mass media (radio, television, newspapers, and so on, that enables one or a few individuals to reach an audience of many); and (2) localite (which links a potential adopter with sources inside the social system under study) versus cosmopolite (which links a potential adopter with sources outside the social system). Rogers also pointed out that opinion leaders in the interpersonal diffusion network have a great impact on the rate of adoption.

Rogers (2003) described the authority innovation-decisions within organizations, which means the decision to adopt an innovation in an organization is made by a few authoritative individuals who possess power, high social status, or technical expertise. He pointed out that the process of innovation adoption in organizations includes five stages. The first stage is agenda-setting, which includes ‘(1) identifying and prioritising needs and problems and (2) searching the organization’s environment to locate innovations of potential usefulness to meet these organizational problems’ (p. 422). The second stage is matching, which means a problem from the organization’s agenda is matched with an innovation in order to determine how well they fit. The third stage is redefining ad restructuring. At this stage, the innovation is re-invented to accommodate the organization’s needs and structure more closely, and when the organization’s structure is modified to fit with the innovation. The fourth stage is clarifying, which ‘occurs as the innovation is put into more widespread use in an organization, so that the meaning of the new idea gradually becomes clearer to the organization’s members’ (p. 427). The last stage is routinising, which occurs ‘when an innovation has become incorporated into the regular activities of the organization’ (p. 420). Rogers also illustrated an innovation-decision process, which models how individuals or organizations pass from gaining initial knowledge of an innovation, to making and confirming an adoption decision. This process also includes five stages: 1. Knowledge (the potential adopter is exposed to an innovation’s existence and gains an understanding of it); 2. Persuasion (the potential adopter forms a favourable or unfavourable attitude towards the innovation); 3. Decision (decide to adopt or reject the innovation); 4. Implementation (the innovation is adopted); and 5. Confirmation (the adopter seeks reinforcement of the innovation-decision already made, and may reverse this decision if exposed to conflicting messages about the innovation).

Research questions

This project will answer these research questions:

• RQ1. What consistencies can be identified from the diffusion and adoption of the two innovations among libraries?
• RQ2. What disparities exist in the diffusion and adoptions of the two innovations among libraries?
• RQ3. To what extent findings from this project conform to existing DI theories?

Method

This study builds on two previous projects, each of which was conducted to investigate the diffusion and adoption of research data management service and linked data among libraries respectively. The two projects were conducted using similar methods: collecting and analysing literature about research data management service/linked data adoption in libraries in the United States. Literature was collected through Google Scholar search, citation chasing, and target search for people or libraries that are involved in research data management service or linked data adoption. All the gathered publications were registered in a Microsoft Excel spread sheet. In total, eighty-three publications that cover the research data management service adoption of forty-four libraries were gathered, and seventy-six publications that cover the linked data adoption of fifty-three libraries were gathered. The gathered literature includes journal and magazine articles, conference proceedings, project reports, presentation slides, and blogs. The gathered articles were read through carefully and information relevant to research data management service or linked data adoption was copied to a separate Microsoft Word document. The content of the Word document was then marked up with relevant codes, including name of the library, the year in which particular research data management service/linked data adoption activity was conducted, and emergent themes, such as the decentralised nature of innovation diffusion and the stages in innovation adoption processes. Findings from each of the two previous projects have been reported in Niu (2019) and Niu (2020), respectively. Please refer to these two publications for more details about the methodology, findings, and literature review for each of the two innovations.

This paper compares findings from the two previous projects and identifies consistencies and disparities. In some cases, the consistencies or disparities discovered in the comparison motivated the author to re-analyse the data gathered for the two previous projects. For example, the author identified some stages in the adoption of linked data within libraries. However, adoption stages were not clearly identified during the previous project on the adoption of research data management service. Therefore, the author re-examined the original data collected for research data management service adoption more closely and identified some adoption stages.

Findings

Decentralised diffusion

The diffusion of both innovations among libraries follows a decentralised model which is characterised by two features. The first is multiple lead adopters, which, typically, are top research libraries and some leading vendors in the case of linked data adoption, which adopted either of the two innovations and created numerous related products such as software, linked data vocabularies, and datasets. For example, Purdue University Libraries created the Data Curation Profile Toolkit which helps librarians design interview questions for investigating how researchers create, manage, and use data (Witt et al., 2009). The Library of Congress released the BIBFRAME vocabulary and MARC-BIBFRAME converter. These lead adopters then disseminated their adoption experiences and products to other libraries. They act as innovators and community leaders. The second feature is a high degree of re-invention. This is demonstrated through the variant levels and diverse approaches of the adoption of the two innovations. Libraries vary in the breadth and depth of the research data management service they provide. They may generate linked data datasets only or also develop linked data-based discovery services. They may convert legacy data only or also produce native linked data. They also use different vocabularies for metadata modelling and different technology tools in converting, generating, storing, and providing access to linked data. This decentralised diffusion pattern is likely to be caused by the nature of the two innovations. Neither linked data nor research data management service is a fixed technology product developed by a single vendor and then diffused as a uniform package. Instead, they are a general technology and a service type that could be implemented in different ways by different adopters.

Diffusion paths in the decentralised model

Three paths were identified in the decentralised diffusion of the two innovations: inter-library diffusion, intra-library diffusion, and inter-librarian diffusion. Before deciding how to adopt an innovation, a library typically reviews the adoption experiences of other libraries, usually lead adopters and/or peer institutions. They also disseminate their adoption experiences and products to other libraries. This is the inter-library diffusion path. Within a library, usually a few librarians learn about an innovation from external sources, and then disseminate knowledge about the innovation to other librarians in order to gain administration support or prepare for adoption. This is the intra-library diffusion path. Both inter-library diffusion and intra-library diffusion are carried out through inter-librarian diffusion. It is the librarians who learn about the innovation and diffuse their knowledge about the innovation to other people in and out of their own libraries.

Innovation adoption is a continuous process that includes multiple stages

The adoption of both innovations within libraries is a continuous process that could last many years and include multiple stages. Some common stages are identified during the adoption of the two innovations:

• Awareness. Libraries hear about the term and concept of the innovation.
• Education. Libraries gain enough understanding of the innovation that they can make informed decisions regarding whether, when, and how to adopt the innovation. For example, George Washington University Libraries trained their cataloguing staff with linked data knowledge and skill through webinars, workshops, and exercises (Shieh and Reese, 2015).
• Preparation. Before adoption, many libraries start with preparatory or exploratory activities. For example, they investigate local researchers’ needs in order to determine what research data management service to provide (Berman, 2017). Some libraries insert uniform resource identifiers in MARC records (Shieh and Reese, 2015), establish authority control in digital libraries (Myntti and Neatrour, 2015), or conduct experimental projects for linked data generation (Shieh, 2013).
• Partial adoption. Many libraries start with some basic research data management service, such as consultation service, and then gradually expand the variety and depth of their services, such as training researchers and acquiring research data into their institutional repositories (Peters and Dryden, 2011). Many libraries start linked data adoption for individual collections or library functions, such as converting cataloguing workflows to linked data environment (Futornick, 2018) or adopting a linked data-enabled digital repository system (Simic and Seymore, 2016).
• Holistic adoption. At this stage, a library is providing a full set of research data management service for researchers in its institution, or linked data is applied in all the relevant library functions and collections.

Not all stages above exist in the adoption of all innovations in all adopters. Libraries that adopt linked data through a service provider or a turn-key software platform might not experience the education stage. Since linked data is a backend technology, a library might have adopted linked data without knowing what linked data is. A library could also start to provide some basic research data management service without initial research into researchers’ needs (Henderson and Knott, 2015).

Adopter categories

The continuous nature of the adoption of the two innovations makes it tricky to decide when an innovation is adopted in a specific library. Thus, it is not possible to decide whether the adoption of the two innovations over time converges to a normal distribution model as described in Rogers (2003). However, the author found adopters of the two innovations can be roughly divided into early adopters and late adopters, each of which demonstrate different diffusion and adoption behaviour. Early adopters face the situation that there are no, or very limited, adoption experiences from other libraries. This determines several types of behaviour or characteristics of early adopters. First, they need to be innovative and conduct original research to decide how to adopt the innovation. Second, this kind of research, due to its innovative feature, is likely to receive external grant support, which in many cases require grantees to develop standards or tools that can be widely used by many other libraries. This makes some early adopters community leaders. Third, early adopters often need to use the expertise and tools created outside of the library community given that no or very little expertise exists within the library community. Late adopters lose the opportunity to become pioneers and community leaders. Yet they enjoy the benefits of being a late adopter: the adoption experience of other libraries allows them to draw upon and many existing tools and services so that they can avoid the high cost and some pitfalls in early adoption.

The higher cost for early adopters was especially evident in linked data adoption. In order to experiment with linked data adoption, some metadata librarians or cataloguers took extra efforts to learn programming skills or conduct some tasks that could be easily done by future software, such as constructing SPARQL queries or writing RDF (resource description frameworks) graphs in Turtle language. In contrast, late linked data adopters can use turnkey software to do linked data related work without knowing the backend linked data technology. It is worth noting that although some research and experimental projects of some libraries are funded by external funding, many research activities related to early adoption of the two innovations were conducted without external funding. This means that only those libraries that have the financial and technical capacity can afford to be early adopters of certain innovations. As stated by Mitchell (2016):

the fact that LAM institutions are still having to select triple stores, SPARQL engines, indexing platforms, and other services means that there is still a relatively high bar for institutions to cross in taking up LD projects (p.30)

Wang and Yang (2018) also said:

So far only big libraries and organizations have the technical expertise and financial resources to devote to the test and development of the Linked Data projects... Most small libraries are watching and waiting rather than participating. (p. 13).

Facilitators in the diffusion and adoption of innovations

Three types of institutions were found to have facilitated the diffusion and adoption of both innovations. The first type of institutions are community leaders. As mentioned earlier, community leaders pioneer the research and practice of innovation adoption and create related standards and tools that are widely used among the library community. The second type of institutions are external funders. They facilitate the diffusion and adoption of innovations through funding original research, workshops and forums, and the development of tools and solutions for innovation diffusion and adoption. Very importantly, many funders require grantees to develop resources and tools that have wide impact and to share project products and findings publicly. External funding is important for all libraries, including top research libraries that are relatively well resourced. The third type of institutions are professional associations. They help build professional networks, organize conferences, publish literature, and provide training on innovation adoption. More details about the functions of professional networks have been discussed in Niu (2020).

Communication channels

Communication happens during every step of the diffusion and adoption of the two innovations: from other communities to the library profession, within the library profession (inter-library, intra-library, and inter-librarian diffusion), and during the research and implementation activities conducted to generate solutions and products for innovation adoption. The following communication channels were identified from the diffusion and adoption of the two innovations. These communication channels are not mutually exclusive. For example, education and training involves reading literature and forming professional ties.

Literature

Librarians learn knowledge about the innovations, adoption experiences of other institutions, and related tools through numerous kinds of literature: books, journal articles, conference proceedings, policy documents, project reports, online blogs, video tutorials, etc. Other than scholarly literature that requires high-level intellectual effort to consume, there are also news articles and other non-technical documents that are used to share general information about a research project or an innovation. For example, the linked data study group at University of Nevada, Las Vegas libraries used a comic strip mounted on a large poster to educate other library staff and administration about the benefits of linked data adoption (Lampert and Southwick, 2013). This kind of literature is more like the traditional mass communication channels, such as newspapers and radio, which usually provide general and high-level information for the public.

Conferences and meetings

Conferences provide venues for interpersonal communication and help build professional and social networks among librarians. Many conferences are organized by professional associations.

Education and training

There are various kinds of education and training opportunities offered by library and information science schools, professional organizations, vendors, community leaders, and other types of organizations, such as the Library Juice Academy, Zepheira’s LibHub early adopters training, and Library of Congress. Some research projects aim to develop teaching and learning tools for linked data adoption. For example, the Linked Data for Professional Education (LD4PE) Project produced a competency index for linked data. The Educational Curriculum for the Usage of Linked Data (EUCLID) project published a comprehensive textbook focused on linked data creation and use.

Professional network

Librarians belong to various professional associations, communities, and interest groups, such as the Library Linked Data Interest Group which is a joint group between Library Information Technology Association and the Association for Library Collections and Technical Services, and the DataCure group of data curation librarians (Barbrow et al., 2017). These professional networks spawn inter-personal communication and collaboration among librarians.

Collaboration

Multi-institutional or multi-disciplinary collaboration is very common in the adoption of the two innovations, especially during the research and exploratory stages. For example, the Library of Congress collaborated with Zepheria in developing BIBFRAME, which was tested by many other libraries (Steele, 2019). The Linked Data for Production (LD4P) project is a collaboration between six leading libraries (Schreur, 2018). Purdue University Libraries collaborated with the Graduate School of Library and Information Science at University of Illinois at Urbana-Champaign (UIUC) in the Data Curation Profiles project (Witt et al., 2009). Multidisciplinary collaboration happens also within the same institution. Libraries often collaborate with information technology, research administration, and other data service units (data centres and archives and data-intensive departments) in creating and managing a research data management service. In a pilot project for BIBFRAME conducted by the Library of Congress, other than cataloguers and technicians who catalogue materials, there are also people who train these cataloguers and technicians with knowledge on semantic web, linked data, and the use of the BIBFRAME editor, and software developers who would modify and improve the BIBFRAME editor based on suggestions from cataloguers and technicians (Library of Congress, 2016).

The above communication channels involve the use of various communication media or technologies. For example, users can attend conferences virtually through Skype, Google Hangout, Webinars, or other teleconference tools. E-mail is used widely for communication between individuals, within small groups and large communities. Cloud storage and digital repositories are often used to share project documentation within a project team or the public. For example, the Library of Congress, the LD4P Ontology Group, and National Library of Medicine all used GitHub for sharing project documentation and software tools (McCallum, 2016; LD4P Ontology Group, n.d.; Fallgren, 2015). Some libraries also allow the general public to post comments, issues, and questions in their GitHub repository. The University of Oregon and Oregon State University libraries also used Google Drive for sharing administrative documentation with project team members (Simic and Seymore, 2016).

Disparities between the two innovations

Apart from the consistencies mentioned above, there are some noticeable differences between the diffusion and adoption of the two different innovations. Community leaders in the diffusion and adoption of research data management service are exclusively university libraries. In the diffusion of linked data, in addition to top university libraries, national libraries (the Library of Congress and the National Library of Medicine), and some vendors also play leading roles. This disparity exists because linked data are relevant to all types of libraries, not only college and university libraries, as in the case of a research data management service. This broad scope of application makes linked data adoption among libraries profitable for vendors, who are then motivated to provide services and software, publish datasets and vocabularies, host knowledge sharing forums, and provide training. Many libraries rely on software vendors and service providers to adopt linked data, especially on a production scale. In contrast, most well-known software tools that were created to support a research data management service are mostly open source and created by and for research libraries. It is worth noting that although some vendors and service providers are also community leaders in the diffusion and adoption of linked data, top research libraries play more important roles. As pointed out in Niu (2020):

leading libraries are more innovative than many vendors... In fact, except some leading vendors like OCLC and Zepheira, most vendors have been watching and waiting for the library community to develop relevant standards, articulate functional requirements and nurture a market for their software products and services. (p.9)

Due to the involvement of vendors and service providers, commercialisation is a boosting factor in the diffusion and adoption of linked data. Standardisation is another boosting factor for the diffusion and adoption of linked data. The maturity of BIBFRAME paves the way for vendors to upgrade their systems and incorporate linked data functionalities. Neither of these two factors exists for research data management service adoption; instead, data management policies of government and private funders are a strong boosting factor. They play a comparable role as standardisation and commercialisation in the diffusion and adoption of linked data.

Research data management service is a new service, whereas linked data adoption in many cases means doing the same things differently using newer technologies. In the adoption of a research data management service, librarians need a new set of skills that they did not have. In the adoption of linked data, librarians, especially cataloguers and metadata librarians, need to be re-trained do the same work in a different way. The old way of doing things and the previous knowledge about cataloguing may get in the way of learning new knowledge and skills. For example, in the Library of Congress pilot for BIBFRAME:

The participants’ mastery of MARC may in some ways have hindered their initial ability to use the Editor effectively. They are accustomed to identifying cataloging elements by their MARC tags and subfields; but the Editor was configured … to use RDA terminology rather than MARC terminology; i.e., they need to think of the title as the “title”, not “the 245”. (Library of Congress, 2016, Teaching experiences).

Although research is involved in the early stages of the adoption of both innovations, different research methods are used. As a service innovation, research data management service adoption involves investigating heavily in the needs of researchers through surveys, interviews, etc., which can be categorised as social research. For linked data adoption, much research is conducted through experimental projects that involve technical trials. User testing of technology tools for linked data conversion or generation are also conducted. For example, Cornell University Libraries conducted usability testing of VitroLib in order to understand how the software tool can be customised to conform to catalogers’ needs and expectations (Kovari et al., 2017). UIUC libraries assessed how users perceive and use linked data-based new features through user interviews, focus groups, and server log analysis (Jett et al., 2017).

A steep learning curve is identified as a major barrier in the early adoption of linked data (Smith-Yoshimura, 2018). Some libraries form study groups specifically for learning and understanding linked data technology (Lampert and Southwick, 2013). In addition, hands-on training stands out as a communication channel for the diffusion of linked data (Lampert and Southwick, 2013). These unique features were not found in the diffusion and adoption of research data management services.

Discussion

Practical implications of findings

Findings from this study reveal the challenges that libraries face in innovation adoption. First, a single library might not have all the necessary infrastructure and expertise to develop widely useful adoption solutions. Thus, it is important to collaborate with other parties, such as other libraries, technology partners, and vendors. In addition, forming multi-disciplinary groups within the library is often necessary to bring together various kinds of expertise needed for innovation adoption. Secondly, libraries cannot always rely on vendors to provide solutions for innovation adoption. Vendors are involved only when the market is large enough to be profitable. In addition, many vendors rely on the library community to create standards and specify functional requirements. Thus, during the early stages of the diffusion and adoption of an innovation, top research libraries or library consortia should take on the responsibilities of leading the research and development of standards and technologies needed for innovation adoption. Thirdly, early adoption involves much learning of new knowledge and original research. It is costly and even well-resourced libraries need external funding support. Thus, libraries that strive to be early adopters of innovations should hire employees who are willing to learn, able to apply for external grants, and to conduct original research. These libraries should also provide learning opportunities for their employees to acquire the skills needed for innovation adoption, such as attending conferences, workshops, or taking courses offered by information schools and other training institutes. They might also consider including a research assignment in some employees’ workload to give them time for the learning and research needed during the exploratory stage of innovation adoption. Given the high cost of early adoption, it might be a wise decision for less well-resourced libraries to be later adopters and enjoy the benefits of doing so. There will be plenty of adoption experiences and tools shared by early adopters. They might be able to achieve faster and easier adoption through using mature technologies, skipping some stages, and avoiding some pitfalls in the adoption process. Finally, libraries should also be aware that different types of innovations might bring different challenges. A service innovation might require research on user needs and technology innovation might need hands-on training and technical trials. In adopting an innovation that enables librarians to do the same things differently, previous knowledge and skills might get in the way of acquiring new knowledge needed for innovation adoption.

Findings of this study are also informative for other relevant parties. In order to become champions of innovation adoption in a library, librarians should develop relevant capacities. In addition to being life-long learners and capable researchers, librarians need to be able to collaborate effectively with professionals in and out of the library community, and to master communication skills for inter- and intra-library diffusion of innovation. They also need to be proactive in gaining administration support for innovation adoption rather than waiting passively for administration decisions. Findings of this study suggest that funders should pay special attention to the early stages of the diffusion and adoption of innovations. They also prove the validities of some existing funding strategies of many funders, such as requiring grantees to develop tools, standards, and solutions that are widely useable, requiring project deliverables to be shared publicly and freely, encouraging multi-institutional and multi-disciplinary collaboration, and supporting forums or conferences that facilitate the diffusion of knowledge about innovations. Professional associations should provide or help the formation of free or low-cost communication channels, such as online communities, virtual conferences, free Webinars, open-access journals or conference proceedings, and local communities of practices that facilitate face-to-face meetings.

Theoretical contributions of findings

In Rogers, innovators and opinion leaders are two different groups. As stated in Rogers (2003):

The most innovative member of a system is very often perceived as a deviant from the social system and is accorded a status of low credibility by the average members of the system. This individual’s role in diffusion… is therefore very limited. Certain other members of the system function as opinion leaders. They provide information and advice about innovations to many other individuals in the system. (p. 26).

While the author believes this scenario does exist in the diffusion and adoption of other innovations, this study reveals the existence of alternatives. In this study, community leaders are both innovators and opinion leaders. They produce widely useful findings and tools, share information and advice regarding innovations and their products and adoption experiences widely in the community.

Niu (2019) discovered that scholarly communication plays an important role in the diffusion and adoption of research data management service. She also pointed out that, although scholarly communication can be considered a sub-type of mass communication, it is particularly important for the diffusion of sophisticated innovations that require professional expertise to adopt. This study confirms the importance of scholarly communication (literature, conferences, etc.). In addition, it discovered the importance of hands-on training in the diffusion of linked data, which is a sophisticated technology innovation. Hands-on training is different from scholarly communication, which is more effective in diffusing concepts and ideas than technical skills. Using the categories defined in diffusion of innovation theories, hands-on training can be either mass communication or interpersonal communication depending on the number of information recipients.

Existing diffusion theories focus on one-way communication from innovators, early adopters, and opinion leaders to later adopters. Niu (2019) pointed out that collaboration is a two-way communication channel that allows knowledge related to innovations to flow back and forth among collaborators. This study has found that two-way communication between the senders and receivers of information is prevalent and easy to do due to new communication technologies. For example, people can easily leave comments on a blog, ask a question in a Webinar, and email the author of a paper. As mentioned earlier, some libraries allow the general public to post comments, issues, and questions in their GitHub repository.

In Rogers (2003), mass communication is mediated through communication technologies such as radio, television, and newspapers. This remains true in this study except that the communication technologies are updated. For example, mass communication technologies can be a website, a digital repository, or a YouTube video. However, Rogers defines interpersonal communication as the face-to-face exchange between individuals. This study has found that inter-personal communication is often mediated through communication technologies, such as e-mail, instant messaging, and tele-conference software (which did not exist when Rogers first published his theory). Thus, the definition of interpersonal communication needs to be updated in the current technology environment. Mediated inter-personal communication not only happens between people that are geographically distant, it also happens among co-located people who can easily meet face-to-face, such as people in the same library.

The authority innovation-decision model described in Rogers (2003) applies to library innovation adoption. However, this study found that although library administration makes decisions regarding innovation adoption, practicing librarians are not merely passive followers. In some cases, they actually play an important role in administration’s decision to adopt an innovation. For example, at University of Nevada, Las Vegas Libraries (Lampert and Southwick, 2013), a few librarians became aware of the innovation and took the effort to learn more advanced knowledge about the innovation, then they educated the administration in order to gain support for innovation adoption and various related activities, such as purchasing software and hiring new employees. This study also found that for linked data adoption, the authority innovation-decision model not only applies to individual libraries, it also applies to the library community at large. For example, community leaders decide to implement BIBFRAME, which is then incorporated in many integrated library systems. In this case, many other libraries have no other options but to adopt BIBFRAME, although they have the freedom to decide when to adopt and which software product to adopt.

Consistent with Rogers (2003), this study found that innovation adoption processes can be complicated and include multiple stages. The five adoption stages identified from this study provides an alternative view of innovation adoption processes and have equivalences in the innovation-decision and adoption processes reported in Rogers. See Table 1 for the correspondence between adoption stages identified in this study and those of Rogers. The awareness and education stages identified from this study are equivalent to the knowledge stage in the innovation-decision process in Rogers. They are broken down into two separate stages because some adopting libraries might not experience one or both stages. This author believes that a library might use the linked data functionalities of a newly adopted software platform without knowing the technical details of linked data, or even not knowing what linked data are. Persuasion and decision from Rogers might happen during both the education and the preparation stages identified from this study through learning about the innovation, what other libraries are doing with the innovation, and researching and experimenting with the innovation. Implementation, confirmation, and redefining and restructuring happen during both the partial and holistic adoption stages identified from this study. Clarifying and routinising happen during the holistic adoption stage and thereafter. During the holistic adoption stage, certain partial adoption of the innovation should have been routinised. Routinising continues after holistic adoption. The routinising stage is not identified from this study probably because of two reasons. First, many libraries have not reached that stage yet; secondly, libraries that had reached the routinisation stage did not report it in publications.

The five-stage adoption process (agenda-setting, matching, redefining and restructuring, clarifying, and routinising) in Rogers (2003) describes a common innovation adoption scenario in many organizations. In this scenario, innovation adoption in an organization starts with identifying needs and problems, followed by searching for innovations that could possibly meet the needs, and deciding how well the innovation fits the needs. This author believes that this adoption process applies to the library community at large and some leading libraries. Library community leaders have long recognised the limitations of MARC and have been searching for new technologies that could replace it. linked data was found as a match. Rogers acknowledges an exceptional scenario where it is the discovery of an innovation, rather than organizational need, which triggers innovation adoption. As stated in Rogers (2003, p. 423), ‘Sometimes knowledge of an innovation, rather than the recognition of a problem or need by an organization leading to search for a solution, launches the innovation process’. The author believes this exceptional scenario applies to the innovation adoption process of many late adopting libraries. These libraries learn about an innovation and then decide whether, when, and how to adopt it.

Limitations

This study is based on analyses of literature reporting the adoption process. However, not all adopters report their adoption experiences through literature and published literature does not reflect all the adoption activities of adopters. In addition, non-adopters are completely excluded. Thus, it is unknown what keeps a library from adopting an innovation. In order to solve this problem, the author plans to survey and interview librarians to gather additional data about the diffusion and adoption of the two innovations among libraries. It is possible that findings from surveys/interviews may temper some of the conclusions from this study. For example, standardisation might also contribute to the diffusion and adoption of research data management service.

The two innovations studied in this project, although very different, share some common features. They are not technology products that can be easily adopted. For example, there is not much intellectual burden to create a Twitter account and start to tweet. Both innovations require library professionals to spend a considerable amount of time to learn and research for early adoption. Findings from this study are likely to apply only to the diffusion and adoption of similar innovations. They may not apply to dramatically different types of innovations, such as technology products that do not require much learning and training to adopt.

Conclusions

This study has attempted to identify consistent patterns in the diffusion and adoption of two recent innovations among libraries. The findings are likely to be useful for the decision-making of libraries and relevant parties facing similar future innovations, namely service innovations and backend technology innovations that require significant intellectual effort to adopt. They also contribute to existing DI theories through revealing new communication channels and alternative adoption processes, as well as redefining existing concepts. The author hopes this study will inspire further research on other types of innovations, so that findings from this study could be confirmed, augmented, or refuted, and a portfolio of consistent patterns in the diffusion and adoption of innovations among libraries could be formulated over time, and become robust guidance for dealing with future innovations.

About the author

Dr. Jinfang Niu is an associate professor at the School of Information, University of South Florida. She received her Ph.D. degree from University of Michigan, Ann Arbor. Before that, she worked for the Tsinghua University Library in China.

How to cite this paper