Multiple authorship of scientific reports—two different evaluations are better than one

In the academic milieu, publishing is an integral part of the current activity of the staff, and promotions in the academic career frequently depend on having a solid publication curriculum. A similar situation exists in research centers. Contemporary scientific reports frequently have more than one author and are frequently signed by multiple authors, a finding interpreted as a marker of research collaboration1. The rise in author number per article in recent decades has been seen in multiple fields and also in biomedical publications2. In the year of 2015, a Physics paper was published with more than 5,000 authors—at the time, presented as setting a new record3. Many other reports exist with a large number of authors. Although the situation seems to be especially important in nuclear and particle Physics, it also exists in other areas of scientific publishing. In Science, we are very far from “The Death of the Author” of Barthes (“Ia naissance du lecteur doit se payer de la mort de I'Auteur”/the birth of the reader must be paid by the death of the Author)4. Not only is the number of authors increasing, but in fact, non-Human “authors” are now entering the stage. In the present text, the problem of knowing if every person presented as an author of a given report should be credited as such will not be discussed. Instead, the problem of the fairness of the impact on the curriculum of each author will be analyzed. It appears clear that if each article is equally and fully credited to each author, a situation is created under which it would be the same thing to publish as a single author and as the 4999th author out of a list of 5000—something that common sense would easily indicate to be absurd. Under this situation, the addition of an extra co-author has no consequence for the other authors, thereby perhaps stimulating this kind of practice. If each article is assigned in its entirety to each author, it will appear counted in full for any single author, for instance in applications. As a major way of evaluating scientific production stands the “h-index”5. Although relatively recent, this tool allows the identification of authors with high productivity (number) of reports able to be highly cited—maybe making industrial production come to mind. This index is independent of the number of co-authors (since each article is equally and fully credited to each author)6 and has been described as unfair in the context of multiple co-authors. When presenting the “h-index,” Hirsch criticized alternative methods, including measuring the total number of citations. However, this author recognized that “a scientist with a high h achieved mostly through papers with many coauthors would be treated overly kindly by his or her h.”5 Several alternatives to the h-index exist, including the i10-index, presented by Google Scholar, counting how many papers have ≥10 citations. Why does multiple authorship exist? We would like to offer the following answer: because it works. In some areas including Physics and Medical Science (especially in large scale clinical trials), the cooperation between many individuals is critical to achieve the goals. Although many “co-authors” may have failed to write a single word or contribute with a single idea, they may not be willing to cooperate without having their name on the authors list. Multiple authorship may act as a stimulus for gratuitous authorship7—including listing as co-authors known scientific figures of the same field. However, it may lead to an increased number of citations—including self-citations and citations by members of each author's network. It is not to be excluded that a large number of authors may transmit an image of high scientific integrity and relevance. A study carried out in German Medical Schools showed higher productivity and higher impact when whole counting, rather than crediting first and last authors with a higher share, took place8. Favorable results may lead to other aspects being overruled. Multiple authorship of scientific reports is unlikely to disappear in the near future, as is “industrial-type” research or the use of the h-index. Given this, and aiming at restoring balance, it may be reasonable to suggest using both a multiple authorship insensitive metric and a multiple authorship sensitive metric and preferably one that may allow the sum of the contributions of one single scientific report to be counted as one9. In the case of Master or PhD (Doctor of Philosophy) theses, the first author is frequently the student and the last author, the supervisor. A similar situation exists with many residents (specialty trainees) in the case of medical doctors. In this type of context, the first author—the “author”, as opposed to “co-authors”—is clearly the most important person in the list. The last author is frequently the second more important person, one that may lead the research group and may have provided the research topic. Sometimes, the last author is just the head of department, who did not contribute significantly, at the intellectual level, to the report. This latter situation is very difficult to separate from the previous one and from other possibilities that exist. The second author may also have played a significant role if the number of authors is small, but the case may be different with a long list of authors. It may be useful to credit a given value (a contribution share) to each author in the list of authors for a given report, accepting that the sum of the total values assigned should be one10. The method presented by Abramo et al. to assign a given weight as co-author of a publication11 is probably a good one, as is the method used in Germany8. We would like to suggest a simple version, with a change in the relative weight assigned to each author: The first author should be credited with a value of at least 0.5 (50%), for the reasons already discussed, the last author at least 0.25 (25%), and the remaining 0.25 should be divided among the remaining co-authors. In the case of one author, the assigned value is one. In the case of two authors, the first author is credited with 0.5 + 0.125 = 0.625, and the second and last author is credited with 0.25 + 0.125 = 0.375. In the case of three authors, the first author is credited with 0.5 and the other two authors are credited with 0.25 each. In the remaining cases, the first author is credited with 0.5, the last author with 0.25, and the remaining authors with 0.25/(n − 2), n being the total number of authors. This contribution share should not apply to reports in which authors are presented in alphabetical order—in this case, the assigned value would be 1/n (n=number of authors). If and when the authors specify in the report an ad hoc contribution of each author in numeric terms, these values may be used instead of the values obtained with the method presented above—with the specification that the total sum of the credits assigned for all authors should be one. If a list of publications of a given author is under evaluation, the sum of the credited values for each report (the corresponding contribution shares) can be calculated. Each of these values can be multiplied by a numerical value assigned for the journal in question (for instance, the Impact Factor of the journal, as defined in the Journal Citations Report, by Clarivate Plc, the h5-index of the journal, from Google Scholar Metrics, or one of other alternatives). If citations are under analysis, a similar calculation can be made, by multiplying the number of citations for a given report by the corresponding contribution share, and adding the values thus obtained. The scientific curriculum of a given person may therefore be calculated, both in what concerns published reports and citations received, by adding the numbers obtained, for each report, by the method presented above. The calculation for citations, we could term the C1-index, since it relies on the basis that each paper is computed as a total contribution value of one. The calculation of this index depends on the database under use. For the list of publications, the P1-index could be calculated. This index can be adjusted according to the credited importance for each journal (impact factor, h5-index, etc.)—which we could term the wP1-index (w standing for weighted). This approach differs from previous models12 in which the total sum of credits for a single report is not necessarily one. The evaluation of a list of publications and citations of a given author can be made both by the h-index (or by the i10-index) and by the method now presented (Table 1; Fig. 1)—perhaps evaluating different components of the scientific publication phenomenon. Table 1 - Examples of multiple authorship insensitive and suggested multiple authorship sensitive metrics. Multiple authorship insensitive metrics Total number of citations Sum of citations for each authored report h-index The maximum value for which the author has published h or more papers that have each been cited at least h times i10-index The number of publications of the author that have received at least 10 citations each Multiple authorship sensitive metrics Calculate author contribution share for each report (sum of all contribution shares = 1) First author—at least 0.5; last author at least 0.25; remaining authors—0.25/n − 2; alphabetical order—1/n for each author Publications P1-index Sum of contribution shares for each authored report Publications weighted by journal evaluation wP1-index Calculate P1-index after weighting (multiplying) each report by adjustment factor—impact factor, h5-index, or alternative metric Citations C1-index Add the list of contribution shares for each report multiplied, in each case, by the citation count N, number of authors. Figure 1.: Flow chart showing suggested evaluation of list of scientific reports. See text and table for explanations.After evaluating millions of papers, Wu et al found that smaller teams tend to disrupt science, whereas larger teams tend to develop existing ideas13. These authors conclude that “science policies should aim to support a diversity of team sizes”13. This conclusion is a useful companion to the present suggestion—that a balanced method to evaluate authors or institutions, encompassing both a multiple authorship insensitive metric and a multiple authorship sensitive metric, may add to the credibility, depth and reach of the evaluation, and therefore may help the development of Science.