Introduction

To be effective, scientific research has to be communicated. Books, book chapters, and articles in journals are some of the research outputs produced by scientists' in their communication activities. Scientists need to use an apparatus of citations and references in view of establishing their priority in the discovery of a new finding or of demonstrating the originality of their contribution [Merton, 1957]. Citations and references have many functions by allowing to acknowledge predecessors, to trace back the origins of a new idea, to differentiate the new findings by the received ones, and so on [Bornmann and Daniel, 2008b; Baccini, 2010]. Scientific outputs, references, and citations represent the observable features or the raw facts on which bibliometric indicators are built.

Bibliometric indicators were originally developed for the quantitative study of science [De Bellis, 2014]. They represented the frontier for scholars analyzing science from the point of view of sociology by permitting to quantify Robert K. Merton's theoretical construction. The pioneering work of Eugene Garfield was aimed to build efficient tools for transforming raw bibliographic data in useful information for scientists and librarians. The computer revolution, the possibility to realize complex bibliographic and citation databases, fostered the development of the new discipline called scientometrics.

During the last 20 years of the past century, the main traits of public policies were the so-called audit explosion, the governance by indicators, and the consequent diffusion of policy evaluation [Dahler-Larsen, 2012]. In this context, research and education policy also needed sets of indicators to inform governments' decisions about research funding. The growing number of scholars working on the new scientometric machinery supplied many instruments able to satisfy governments demanding tools for monitoring science policies. Universities were induced to follow this trend by improving their accountability, and as a consequence they requested their professors and researchers to document the results they achieved and to prove that they produce value for money. In these years, the seed was planted of the publish or perish age. Bibliometric and citation databases ceased to be managed by scholars at the boundary between academic and market and became truly for profit products maintained and diffused by multinational publishers. Many firms started up in the fledging market of research evaluation and consultancy, and newspapers and consultancy firms started to release the now ubiquitous university rankings.

The current explosion of bibliometric indicators is the last step of a dynamic started many years ago. It may be useful to try to disentangle three different perspectives in which bibliometric indicators are currently developed and used. A first perspective can be labeled as positive bibliometrics. Its main aim is to describe and explain phenomena in science and scientific communication. In this perspective bibliometric indicators represented phenomena or proxies of phenomena. For example, the diffusion of a new idea in the scientific community may be proxied, under suitable assumptions, by the number of citations received by the article in which the idea was exposed for the first time, which is by the impact of the article. Since bibliometric indicators are only tools functional to the study of science, this perspective is the nearest in spirit to the pioneers of bibliometrics. Citation analysis, cocitation analysis, citation networks, informetric laws, and productivity analysis are examples of this stream of literature.

A second perspective, probably the one growing at a faster pace, is the so-called evaluative bibliometrics [Narin, 1971; Cronin, 2000] aimed to define quantitative instruments for evaluating articles, scientists, journals, institutions, and so on. In the majority of applications, an ordered list, that is, a ranking, is produced as a result of the application of the chosen tools [Glänzel, 2009; Waltma, van Eck et al., 2013]. In this perspective bibliometric indicators implicitly or explicitly depend always on a value judgement about observed features. This stream of literature includes articles which proposed indicators for ranking scholars, such as h-index and its many variants, or for ranking journals - such as the impact factor or SNIP.

A third perspective, not yet well developed, can be labeled as normative bibliometrics, and it consists in the definition of indicators and of rules for their use in research evaluation and in research policy [Elkana, Lederberg et al., 1978; Leydesdorff and van den Besselaar, 1997; Moed, Glänzel et al., 2004; Baccini, 2010; Herranz and Ruiz-Castillo, 2013]. In this perspective the focus is primarily on the goals of research evaluation; the problem is to define an evaluative strategies and to design evaluative tools which are coherent with the chosen goals. In particular, bibliometric indicators should be defined and used in view of designing a system of incentives for researchers and institutions able to push them to adopt behaviors pointing to the achievement of the predefined goals. For example, if the goal of an institution is to climb a university ranking, a system of internal evaluation and rewards should be defined for inducing scholar to behave coherently with the chosen goal.

The growth of evaluative bibliometrics and the universal diffusion of citation data and bibliometric indicators about researchers, papers, journals, and institutions are shaping the contemporary landscape of science [Leydesdorff, 2001]. "Policymakers, scientists, scholars, publishers, research administrators, and funding bodies now have at their fingerprints an unprecedented trove of data" [Cronin, 2014]. Whether indicators are formally inserted in research assessment or not, since they involve money or reputation, they tend to affect researchers' behavior in ways that are not yet sufficiently explored [Lawrence, 2003; Bornmann, 2011a; Wouters, 2014]. The strategic responses adopted by researchers may change research priorities [Butler, 2003], publication activities [Csada, James et al., 1996; Bissell, 2013], or research organization [Taylor, Perakakis et al., 2013] in ways that policymakers may judge positively. Strategic behaviors may give rise also to unintended and undesirable consequences for the long run development of science [Gillies, 2008] and for research integrity by favoring misconduct ranging from plagiarism to fabrication of results [Brembs, Button et al., 2013] or to impolite self-citational conducts.

It is natural that dangers emerge when indicators are poorly constructed and misused, but it is not sufficient to have well-constructed indicators to be safe [Lewison, Thornicroft et al., 2007]. For this reason many efforts were devoted to discern best practices and metrics [Rehn, Gornitzki et al., 2014]. Of these efforts at least three need a short consideration here. The first two have a scholarly nature, while the third is more policy oriented. The first document is the San Francisco Declaration On Research Assessment (DORA) originated in December 2012 in a meeting of the American Society for Cell Biology [San Francisco DORA, 2013] and signed up to now by about 600 organizations worldwide and by more than 12000 individuals. The main aim of the declaration is to support the adoption of good practices in research assessment. Dora recommendations suggest the adoption of good practices to funding agencies, institutions, publishers, organization supplying metrics, and finally to researchers. The main aim of DORA is about the need to eliminate the use of journal-based metrics such as journal impact factors, in funding, appointment, and promotion consideration. Indeed the improper use of journal impact factors as the master key for every kind of assessment is probably the most diffuse abuse in everyday practices. According to DORA, a research should be evaluated according to "its own merits rather than on the basis of the journal" in which it is published; indeed the general recommendation sounds as "Do not use journal-based metrics . as a surrogate measure of the quality of individual articles, to assess individual scientist's contributions, or in hiring, promotion, or funding decisions." It is worthwhile to note that DORA encourages publishers to make available article-level metrics so to allow a shift toward assessment based on the scientific impact of individual published articles or research outputs. It requires also to organizations that supply metrics to be clear about what constitutes an inappropriate manipulation of metrics. DORA commits signatory researchers to adopt the aforementioned general recommendation when they are called to make decision for assessments and to adopt deontological rules for citing: a researcher should give credit where credit is due, by avoiding to cite, when possible, review articles instead of primary literature.

The second document [IEEE, 2013], entitled "Appropriate use of bibliometric indicators for the assessment of journals, research proposals, and individuals," was issued in September 2013 by the Board of Directors of the Institute of Electrical and Electronics Engineers (IEEE), the world's largest professional association for engineering. The document stigmatizes as "misguided and biased" decisions taken by funding agencies and tenure committees in which only bibliometric data are used for the assessment of "scientific quality" of individual scientists and for funding evaluation of research...