An established literature defines inter-disciplinary research (IDR) (e.g., ¹ ), by highlighting the main contexts Where IDR can achieve the largest benefits (e.g., solving problems across naturally complex and socially relevant challenges, by relying on alternative scientific knowledge and management strategies) (e.g., ² ) and the main reasons Why IDR should be implemented (e.g., solving today’s grand challenges, which are too complex to any discipline to tackle alone; generating new research avenues and challenging established beliefs, by enhancing creativity and fostering innovation) (e.g., ³ ). Consequently, IDR has become a catchword in current higher education and science policies across the world. However, interdisciplinary promises in policy discourses are not realised in work practices, since academics confront cognitive-epistemological and institutional-organisational challenges.

In particular, a recent literature discusses the main epistemological obstacles to be dealt with in implementing IDR (e.g., ^{4, 5} ). For example, researchers should renounce prejudices in favour of, and accept simplifications within, their own discipline; researchers should be asking themselves about the origin of their assumptions, in applying the usual procedures to similar topics within IDR; scientists retain incomplete understanding and respect to, or discrimination against, other disciplines; scientists associate IDR with intellectual fashion rather than research substance; IDR is often focused on applied issues and so it is perceived of lower scientific rigor by theoretical scientists; IDR is still risky, poorly defined and variable in terms of interdisciplinary integration.

In addition, a recent literature discusses the main institutional obstacles to be coped with by inter-disciplinary scientists (e.g., ⁶ ). For example, as for IDR supply (e.g., ^{7, 8, 9, 10, 11, 12} ), funding is limited to core sciences; researchers should renounce to funds in favour of other disciplines; academic career is based on publications in few mainstream journals; insufficient training in, early exposure to, opportunities for and encouragement toward IDR; IDR faces difficulty in identifying the best collaborating scientists from different disciplines as well as in managing coordination and integration of distributed knowledge; IDR requires higher initiation effort and time, whereas academic career is based on many latest publications with many overall citations; IDR activities dilutes core disciplinary expertise; IDR focus is more likely to lead to short-term employment rather than tenure jobs in unidisciplinary departments, with possible near-term income risk. Similarly, as for IDR demand (e.g., ^{13, 14, 15, 16, 17, 18} ), difficult assessment of contributions to IDR due to its complex project management and authorship; few brave editors, since journals are ranked in terms of citations and standard articles are more likely to be quoted in the short-run; reviewers are biased against, or they lack understanding of, IDR or scientific methods of other disciplines; reviewers ignore the strength of a papers due to its interdisciplinarity.

However, the focus on either epistemological or (demand and supply) institutional obstacles precludes the evaluation of their relative importance in barring IDR (GAP 1). Next, the use of (unreliable and inadequate) data gathered for a small number of years or a small number of disciplines precludes the consideration of the differences across disciplines in restraining IDR (GAP 2).

The purpose of this paper is to bridge these gaps, by summarising the theoretical literature on inter-disciplinary research (IDR) (THEORETICAL GOAL 1 in Section 2 on issues; THEORETICAL GOAL 2 in Section 4 on solutions) to empirically rank the main obstacles that are preventing its launch (EMPIRICAL GOAL 1 in Section 3) and to empirically suggest the main policies which could favour its launch (EMPIRICAL GOAL 2 in Section 5). To do so, it theoretically characterises IDR in terms of What, How, Where, Why and Who, by linking epistemological obstacles with What (i.e., a set of eight methodologies) and How IDR is implemented (i.e., realism and neutrality) and by linking institutional obstacles with scientists Who implement IDR (i.e., fewer citations of IDR articles and a discriminatory bibliometric indexes to assess IDR scientists). Moreover, it constructs a representative dataset on the interdisciplinary literature (i.e., articles in Scopus with “interdisciplinary” or “interdisciplinarity” in title or abstract) as well as average citations per article and average H-indexes for authors across all disciplines from 2001 to 2020 (i.e., scientific articles are indexed in Scopus). Finally, it empirically ranks the epistemological obstacles, the institutional obstacles due to lack of demand and the institutional obstacles due to lack of supply, by referring to the number of IDR articles, the percentages of IDR articles over the total articles in terms of trends and linkages with citations and H-indexes.

In other words, the research questions can be summarised as follows: 1) (GOAL 1) What is the relative importance of epistemological and (demand and supply) institutional obstacles faced by interdisciplinary scientists? 2) (GOAL 2) Which (long-run and institutional) changes are suggested for scientists to have lower barriers in implementing interdisciplinary research?

The topical contribution of the present paper can be summarised as follows: •

It finds that epistemological obstacles are irrelevant, while the institutional obstacles due to lack of demand by journals are non-significant and the institutional obstacles due to lack of supply by scientists would be termed the driving variable within the complex system theory (EMPIRICAL GOAL 1).

In addition, the methodological contribution of the present paper can be summarised as follows: •

It constructs the most reliable dataset (i.e., academic and non-academic authors and articles in Scopus) and most comprehensive dataset (i.e., all disciplines over 20 years) used in the literature (GAP 1).

Note that I disregard trans-disciplinary research, ¹⁹ since the focus here is on research- rather than decision-making, although the engagement of stakeholders in decisions is crucial in some contexts (e.g., environmental sustainability in ²⁰ ) and problematic in all contexts (e.g., stakeholder’s representativeness and knowledge in ²¹ ). Moreover, I provide examples from the publication experience of the author to support “positive existence statements”, since the lived experience is a legitimate source of information in research and the author is characterised by a net per-capita per-year H-index (i.e., the only version of H-index being suitable to compare inter-disciplinary scientists) ( ^{22, 23} ) at 2.5, with an inter-disciplinary degree ²⁴ at 40%, with the main focus of his thirty-year experience on the most popular topic in IDR (i.e., environmental ethics, sustainability and decisions). ²⁵ Finally, I disregard IDR education, although it could increase the sensitivity degree of future researchers in the long-run. ²⁶

The structure of the paper is as follows. Section 2 reviews the theoretical literature on obstacles to IDR, by characterising IDR in terms of What & How, Where & Why, and Who (i.e., epistemological and institutional lack of demand and supply) (THEORETICAL GOAL 1). Section 3 empirically ranks those obstacles, by relying on an original dataset (i.e., epistemological < institutional lack of demand < institutional lack of supply) (EMPIRICAL GOAL 1). Section 4 discusses the theoretical solutions from the literature for the main obstacles empirically identified (THEORETICAL GOAL 2), by stressing the strengths and weaknesses of the methodologies adopted in the present paper. Section 5 suggests empirical solutions for the main obstacles ranked in the present paper (EMPIRICAL GOAL 2), by resting on the potential synergies of sequential solutions.

This section summarises the main theoretical obstacles to IDR (THEORETICAL GOAL 1). To do so, it reviews the literature which characterises IDR in terms of What, How, Where, Why and Who, by linking epistemological (i.e., cognitive) obstacles with What and How IDR is implemented (i.e., there is consensus on the social benefits of IDR – epistemological Where and Why) and institutional (i.e., organisational) obstacles with scientists Who implement IDR (i.e., there is no consensus on private costs of IDR - institutional Where and Why).

The previous section highlighted some theoretical obstacles to IDR. This section empirically evaluates the relative importance of these obstacles and the consequences on the future achievements of IDR (EMPIRICAL GOAL 1). To do so, it constructs a representative dataset of IDR for all disciplines and it refers to the observed behaviours by scientists in all disciplines.

As for dependent and independent variables, Table 1 compares the main aspects of the recent empirical studies about the impacts of IDR on scientific actors with the main aspects of the present study. Concisely, the present study combines methodologies used by previous papers, by referring to the averages of citations (CIT) and H-indexes (HIN) across properly represented disciplines to enlarge the sample up to all articles in Scopus about IDR in 20 years and up to a representative random sample of 10,000 authors about CIT and HIN in all countries in all disciplines. In particular, within pros, one should stress that the use of Scopus is more reliable than other bibliometric datasets (e.g., ⁷³ ) and that the focus is on articles as peer-reviewed publications, by neglecting reviews, editorials, and book chapters. However, within cons, one should stress that some specific features of authors (e.g., academic, gender, education background, original discipline) are missed as control variables (e.g., ⁷⁴ ) and that the use of CIT and HIN are proxies for Impact Factors and University Placements or Funds for journals and authors, respectively.

Note that I disregarded papers based on surveys (e.g., ⁷⁹ ) and multiple case studies (e.g., ⁸⁰ ), since they are based on perceived and subjective rather than observed and objective values of the relevant variables. Moreover, measuring IDR for each author by relying on the citing articles of his/her papers from outside its discipline when his/her articles are produced with many co-authors might overestimate the interdisciplinary degree of each author (e.g., ⁷⁷ ): for example, a statistician could apply the same method to different topics by producing papers with co-authors from different disciplines. Finally, I disregarded papers on societal impacts (e.g., ^{81, 82} ) and technological impacts (e.g., ^{83, 84} ), since they are not focused on private costs and risks in realizing IDR.

In summary, the empirical purpose of the present study (i.e., ranking the main obstacles preventing IDR from taking off: epistemological, institutional on the demand side, institutional on the supply side) suggested to consider both actors (i.e., journals represented by CIT as the demand side; authors represented by HIN as the supply side) and articles, by using the averages of CIT and HIN across disciplines to link the two units of observations.

As for the IDR dataset, I use data on all articles in the Scopus dataset with “interdisciplinary” or “interdisciplinarity” in title or abstract across 27 disciplines (see Table S1 in Extended data at http://osf.io/b3cj6 for the list of disciplines) from 2001 to 2020, whereas I use the average citations per article and the average H-indexes for authors across 25 disciplines from 2006 to 2015 from a sample of 10,000 authors extracted from the Scopus dataset (i.e., Health Professions and Multidisciplinary disciplines are not well represented in the sample of authors). In particular, I applied a stratified sampling procedure to the reference population of authors with at least one publication in the Scopus dataset from 2006 to 2015, by partitioning this population according to the 25 scientific disciplines used by Scopus. By preserving the percentages of authors in each discipline, I then randomly chose 10,000 authors, by attaching each author to a single discipline, according to the attribution suggested by Scopus (i.e., an author is linked to the discipline with the largest percentage of publications).

Note that I used the Scopus dataset, since it was showed to be less easily manipulated than other bibliometric dataset and more reliable as a set of scientific articles ( ^{85, 73} ). Moreover, although there are many examples of bad practices within the peer review process (e.g., acceptance of papers conditional to some quotations of the journal), I trusted it as a crucial community aspect of scientific work. Consequently, I disregarded books, reviews, and proceedings, since these are less likely to be peer refereed. Similarly, I compared articles with “interdisciplinary” or “interdisciplinarity” in abstract from articles with “interdisciplinary” or “interdisciplinarity” in title to distinguish trendy from genuine IDR (i.e., more and more journals mention “interdisciplinary” or “interdisciplinarity” in their “aims and scope”), under the assumption that peer reviewers highlight a possible inconsistency between titles and methodologies, whereas they do not pay attention to the use of the same words in abstracts (see ^{86, 87} and ⁸⁸ for established and recent examples of bibliometric measures of papers’ interdisciplinarity, respectively). Finally, I used the average citations per article as a proxy of the IDR publication incentives for journals (i.e., the opportunity cost of IDR for journals) and the average H-indexes for authors as a proxy of the IDR incentives for scientists (i.e., the opportunity cost of IDR for scientists).

As for the scientists’ behaviours, I link epistemological obstacles to the difference between absolute and relative numbers of published IDR articles. Indeed, epistemological issues in writing an IDR paper do not depend on the proportion of written papers and they are successfully tackled if this paper is published as a scientific article. In other words, epistemological problems are said to be solved, if the increasing dynamics of absolute IDR articles is significantly larger than the increasing dynamics of relative IDR articles. Moreover, I link institutional obstacles on the demand side to the relationship between percentages of IDR articles and the average citations per article for all disciplines. Indeed, journals are ranked in terms of impact factors (i.e., an index based on the number of citations per year) and IDR articles are likely to be quoted to a smaller extent (i.e., only brave and forward-looking editors accept these papers). In other words, institutional obstacles on the demand side are said to be unsolved, if there is a negatively significant relationship between the average citations per article and the proportion of IDR articles across disciplines. Finally, I link institutional obstacles on the supply side to the relationship between percentages of IDR articles and the average H-indexes for authors for all disciplines. Indeed, scientists are evaluated according to their H-indexes and the academic career is likely to be slower or the access to research funds is likely to be harder for IDR scientists. In other words, institutional obstacles on the supply side are said to be unsolved, if there is a negatively significant relationship between the average H-indexes for authors and the proportion of IDR articles across disciplines.

Thus, a significant difference between the dynamics of absolute vs. relative numbers of published IDR articles means that the epistemological obstacles are irrelevant. In particular, I compare exponential and linear dynamics. Moreover, a significantly negative relationship between percentages of IDR articles (in abstracts ABS or titles TIT) and the average citations per article (CIT) means that the institutional obstacles on the demand side are relevant (i.e., lack of demand). In particular, I refer to the following equations: CIT i , t = η 0 + η 1 ABS i , t + η 2 YEA + θ + ε i , t (1)

Or CIT i , t = η 0 + η 1 TIT i , t + η 2 YEA + θ + ε i , t (2)

Where YEA depicts the yearly trend and θ depicts the discipline fixed effects. Finally, a significantly negative relationship between percentages of IDR articles (in abstracts ABS or titles TIT) and H-indexes for authors (HIN) means that the institutional obstacles on the supply side are relevant (i.e., lack of supply). In particular, I refer to the following equations: HIN i , t = η 0 + η 1 ABS i , t + η 2 YEA + θ + ε i , t (3)

Or HIN i , t = η 0 + η 1 TIT i , t + η 2 YEA + θ + ε i , t (4)

Where YEA depicts the yearly trend and θ depicts the discipline fixed effects. Note that I did not analyse papers with “interdisciplinary” or “interdisciplinarity” as a keyword, since they do not provide additional insights (see Figure S1 to Figure S4 in Extended data at http://osf.io/b3cj6) and they could include contrasting interests between editors, referees and authors. Indeed, editors often modify keywords suggested by authors to increase citations and referees rarely argue on keywords suggested by authors. Moreover, the yearly trend YEA combined with the discipline fixed effects θ enable to register the effect of a change in interdisciplinarity on CIT and HIN, while sorting out potential confounding factors and the influence of a change in other dimensions. In other words, YEA and θ allow to compare only articles with similar characteristics, by considering different sources of unobserved heterogeneity which may interfere with the effect that interdisciplinarity has on the scientific impact of an article (e.g., different citation patterns across disciplines, different co-authorship practices across disciplines). Finally, CIT is different from HIN, since HIN includes the number of articles, although journals can affect citations (e.g., management of peer review processes, policy of open access) and authors can affect H-index (e.g., reciprocal co-author citations to increase the number of citations, large teams of co-authors to increase the number of articles).

The previous section ranked the main obstacles to IDR. This section reviews the theoretical literature to suggest some solutions (THEORETICAL GOAL 2), by referring to the empirical literature to highlight the main methodological advances of the present paper. Note that the most complicated issue in theory (i.e., the institutional obstacles on the supply side, which require a revolution) turned out to be the most urgent issue in practice (i.e., a structural break is required).

The main theoretical solutions to the epistemological obstacles suggested in the literature are as follows: •

An improved division of cognitive labour and scheduling of joint epistemic work and evaluation could favour an adequate organisational arrangement, since IDR can be represented as a cognitive coordination problem to be characterised in terms of heuristics ^{89, 90}

The main theoretical solutions to the institutional obstacles on the demand side suggested in the literature are as follows: •

Handle editors should be identified with an interdisciplinary culture to appreciate the novelty across disciplines.

The main theoretical solutions to the institutional obstacles on the supply side suggested in the literature are as follows: •

Larger trainings on IDR for undergraduate and graduate students could favour interpersonal communication and collaborative skills ^{94, 95}

The main empirical solutions suggested by the present paper are as follows: •

Scholars should be ranked also in terms of the IDR proportion in their publication history (e.g., ^{100 , 101, 102} ). This could favour the recruitments of interdisciplinary scientists at universities in addition to monodisciplinary scientists. However, this policy is likely to be aborted in the short-run, since it is likely to be opposed by unidisciplinary scientists in power thanks to financial systems based on monodisciplinary structures (e.g., ^{103 , 104} ). As an example from my experience, ScholarGPS, apart from showing me within 0.5% of all scientists, 0.5% of all social scientists and 0.5% of all economists, it could detail my first speciality ranking (i.e., 99 in Sustainability, which is an interdisciplinary issue) in terms of my IDR proportion to show my ranking as an interdisciplinary scientist.

Note that the use of bibliometric indexes to suggest an empirical solution to epistemological and institutional obstacles to IDR leads the present paper to disregard educational issues (e.g., ^{115 , 116 , 117} ) and administrative issues (e.g., ^{118 , 119 , 120} ).

The methodological strength of the present study is threefold: 1.

It referred to all disciplines within IDR rather than to few disciplines (e.g., ²⁹ ). This is theoretically supported by a broad definition of science, where all disciplines accepting its methodologies have the same scientific dignity.

The methodological weakness of the present study is the reference to different observation units. Indeed, it did not link the interdisciplinary degree of each single author with his/her H-index and the average citations of his/her articles, but it used interdisciplinary indexes (i.e., dynamics of trending and genuine IDR) which refer to all articles, it used the H-indexes (i.e., the costs of writing IDR papers by authors) which refers to some scientists, and it used the citations (i.e., the costs of accepting IDR papers by editors) which refer to articles published by the same scientists. However, it linked these variables by calculating their averages across the 25 disciplines. In particular, the different sizes of the different disciplines are represented by the proportions of sampled scientists and by the percentages of all articles, whereas the additional differences between disciplines are caught by applying the fixed effects method.

The purpose of this paper was to empirically rank the main obstacles that prevent the IDR launch. By relying on indexes from informetrics, Section 3 empirically showed that institutional obstacles due to lack of supply > institutional obstacles due to lack of demand > epistemological obstacles (EMPIRICAL GOAL 1), whereas Section 4 reviewed some theoretical solutions (THEORETICAL GOAL 2). However, the relative empirical importance of obstacles leads to a sequence of suggested practical solutions (EMPIRICAL GOAL 2), by referring to the short-side theory in economics. First, the adoption of a net per-capita per-year H-index to evaluate scientific activity, with a bonus for IDR, would foster academic careers of interdisciplinary scientists and favour their access to research funds (i.e., supply side of institutional obstacles). Second, a larger group of (old) interdisciplinary reviewers would increase the awareness of the IDR potentials and the publication of IDR articles (i.e., demand side of institutional obstacles). Third, a greater probability of publication would increase the demand for training on IDR by (young) undergraduate and graduate students (i.e., epistemological obstacles).

Note that this sequence would avoid the IDR take off to be based on a few heroic scientists (i.e., around 0.03% of scientists, if each IDR scientist publishes around one IDR article per-capita per year) who bear a private opportunity cost to produce a public benefit (i.e., IDR is a collective action problem). Moreover, it would reduce the time to move from trending to genuine IDR, by keeping the characterisation of IDR. Finally, it would increase the likelihood of an institutional change which is needed to break the vested interests (i.e., IDR scientists are not behind the veil of ignorance by Rawls). In other words, since university leaders and administrators continue to allocate significant resources to monodisciplinary research (i.e., financial systems are essential socio-institutional dimensions of science as a human activity), obstacles to IDR will never be overcome, since the lack of interdisciplinary scientists makes it difficult to identify and group interdisciplinary reviewers and the lack of interdisciplinary funds makes it difficult to set up and preserve interdisciplinary centres.

The present study could be developed by considering transdisciplinary science instead of IDR (i.e., a different topic). However, the theoretical insights should include additional frameworks such as engagement of stakeholders and communication of information, while empirical results should rely on a larger dataset to include variables such as average or top-down decisions and majority or bottom-up decisions by representative and informed stakeholders. In addition, the present study could be developed by using semantic analysis to measure IDR (i.e., a different metric). However, the dataset should include all disciplines, while empirical results should consider different hierarchical relationships between the same keywords in different disciplines.