Level of Open Science Practices in the Berlin Research Area
The term open science does not describe a specific action or practice, but is rather an umbrella term for various practices. The Berlin Science Survey took a closer look at the five best-known open science practices: Open access publishing, data sharing, code and material sharing, open peer review and citizen science.
Open access refers to free and open access to scientific publications on the Internet. This means that these publications can be read by other scientists and a broader public who do not have access to licensed publications subject to a fee.
Data sharing similarly refers to the public and free provision of data. Code and material sharing refers to the public and free provision of codes and study-relevant material (questionnaires, blueprints, etc.) used in previous research. Data, code and material sharing only count as open science if both characteristics are fulfilled. If, for example, data is only shared at the discretion and request of colleagues, or if questionnaires, code and material are only available in return for license fees, then these practices do not count as open science.
Less well known is open peer review, which refers to various scientific review procedures in which the review process is made more transparent and comprehensible, e.g., by dispensing with the double-blind procedure or even by making the reviews public so that they can be seen and discussed by the research community. Citizen science refers to the involvement of non-scientific, civil society actors in the scientific research process. This can be done for different purposes and at different stages of the research process.
In order to determine the current degree of dissemination of open science in the Berlin research area, the scientists were asked to what extent they implement these various open science practices in their everyday research.
Open Access Publishing
The scientists participating in the BSS state that on average about 55% of all their publications are freely available on the Internet (see Figure 1). This result is lower than the 63.3% reported by the Berlin Open Access Office for the Berlin research area in 2020 (Kindling et al. 2022).
However, the latter value refers only to journal articles, whereas the self-assessment in the BSS refers to all forms of publication, i.e., also monographs and anthology entries without time restrictions. According to this self-assessment, 17% of respondents state that they publish exclusively in open access, while 9.6% have never published in open access (both not shown).
The breakdown by status group shows that there are no differences between postdocs and predocs (see Figure 1). Both status groups have published a good 56% of their publications in open access format, with the open access shares of professors falling slightly behind at 52.2%. This difference can be interpreted as a cohort effect: the older cohorts also have older publications, which were generally published less frequently in open access formats. Overall, there are no substantial differences in open access publishing behavior between the status groups.
Figure 1 Share of open access publications, by status groups
However, there are clear differences between the subject groups regarding open access publishing. Figure 2 shows that open access is least widespread in the humanities, but even there, 46% of publications are freely accessible. The pioneers in open access publishing are the natural sciences, with an average share of freely accessible publications of 64%. This is not surprising, as the natural sciences already have a long tradition of relevant infrastructures in the form of open science repositories, such as arXiv. In addition, the natural sciences are dominated by an article-based publishing culture that is generally already better adapted to the requirements of open access publishing (Grimme et al. 2019). In contrast, especially in the humanities, monographs and edited volumes are still widespread (Schneijderberg et al. 2022), which are less often made openly accessible than journal articles (Grimme et al. 2019). Differences in open access publishing can thus be attributed, at least in part, to subject-specific publishing cultures.
Figure 2 Share of open access publications, by subject groups
There are various options or open access routes for making scientific publications available to the public free of charge.
Gold Open Access includes all electronic first publications in freely accessible journals (open access journals). Green Open Access includes all electronic second publications (as pre- or post-prints) in freely accessible institutional or specialist online archives or repositories or on the author's own website. In hybrid open access, individual articles in journals that are not otherwise freely accessible are "bought" for the readership in return for a fee.
DFG programs to promote open access publications explicitly only cover the fees for open access publications via the golden path and not for secondary publications (DFG-Ausschuss für Wissenschaftliche Bibliotheken und Informations-systeme 2022; Deutsche Forschungsgemeinschaft 2020). This puts many scientists in a quandary, as many important (prestigious) journals often do not have a gold open access business model (Open Access Monitor 2023).
In the Berlin Science Survey, those who had previously stated that they publish open access (90.4%) were also asked which of the three most common open access paths they predominantly use to publish. This showed that the golden path of publishing predominates with 51%. The green path of pre- or secondary publication is taken by 39% of respondents, and the hybrid path is also taken by a good third of respondents (not shown).
However, these average values conceal some significant differences between subject groups (see Figure 3). While Green Open Access predominates in the engineering sciences, the golden path is dominant in all other subject groups and especially in the life sciences. The latter, on the other hand, use Green Open Access much less frequently (only just under 25%). The format of Hybrid Open Access is used least frequently by the humanities, at 21.5%. The different distribution of the various open access formats may have discipline-cultural, financial and publication-strategic reasons. Since both the federal government's open access Strategy and the DFG's program to promote open access publications usually only cover the costs of gold open access, the costs of the hybrid format have to be carried by the scientists themselves. However, the proportion of publication costs carried by the scientists themselves varies greatly between subject groups (Over et al. 2005). Life scientists in particular have to contribute to publication costs (ibid.).
Figure 3 Use of various open access formats, by subject groups
Overall, it is clear that open access publishing is already widespread in the everyday research practices of the vast majority of scientists in all status groups. Nine out of ten scientists have open access publications; only one tenth have not published any at all, and these are predominantly predocs. So far, 17% of all respondents have published exclusively in open access. However, there are still differences between the subject groups in the proportion of publications made publicly and freely accessible and in the format of their publications. While the natural sciences are trailblazers in open access publishing, the life sciences are pioneers in Gold Open Access, which is the path most strongly demanded and promoted by research policy.
Other Open Science Practices
In addition to open access publishing, the Berlin Science Survey also surveyed the prevalence of other open science practices (data sharing, code and material sharing, open peer review, and citizen science), which aim to increase transparency and accessibility. These four practices have been found to be less widespread compared to open access (see Figure 4). On the one hand, this may be because they have not (yet) arrived in certain parts of the scientific community. On the other hand, this may also be because they are not relevant in some research contexts. For example, if no data is collected or produced in the research process, data sharing is not a relevant practice here. This difference was considered in the survey. In fact, 18.2% of respondents in the BSS confirm that data sharing has no relevance to them. For code and material sharing, this is 24.2% of respondents. In the case of open peer review, 14.3% state this. And with regard to citizen science, as many as 24.7% state that this practice does not apply to their research practice.
Figure 4 Distribution of different Open Science practices
The respective open science practices can be said to have a certain routine in scientists’ research practices when they are carried out regularly, i.e. "often", "very often" or "always". In this sense, data sharing has already become routine for 26.8% of the respondents. Code and material sharing is part of the daily research routine for 26.7% of the scientists surveyed. About one fifth (21.2%) of the respondents regularly participate in open peer review processes. Far behind the other practices is the involvement of non-scientific actors, which is part of everyday research for only 10.6% of the respondents (see Figure 4).
This insight can be further differentiated when status groups and subject groups are compared. The difference in prevalence of open science practices between status groups (see Figure 5) points to the different roles and responsibilities that professors, postdocs and predocs perform in academia. The greatest differences can be seen in the sharing of research data, and especially in the involvement in open peer review processes. In both cases, professors are far ahead of junior researchers. In the case of open peer review, the difference between professors and predocs of around 34 percentage points and between professors and postdocs of almost 18 percentage points is particularly striking (see Figure 5). This is hardly surprising, however, since scientific quality control in review processes in general - i.e., also in blind or closed peer review - tends to be taken over by more established scientists.
Figure 5 Distribution of different Open Science practices, by status groups
Similarly, the prominent role played by professors in sharing research data may be, among other things, a result of institutionalized and often legally regulated accountability structures for such data and thus for their sharing. For code and material sharing, status group differences are somewhat less pronounced. As many as 26.4% of the predocs surveyed regularly share code or material relevant to their studies. This can be interpreted as an indication that this practice offers a lower-threshold access also for junior scientists. The proportions are also much more balanced when it comes to the regular practice of citizen science. The status group hardly plays a role here. However, more than one third of the predocs surveyed state that citizen science is not relevant to their research, compared to around 16% and 19% for professors and postdocs, respectively.
If we look at the implementation of open science practices by subject groups, we see that the life sciences, natural sciences and engineering sciences, each with a share of regular implementation of around 30%, are well ahead of the humanities and social sciences, which have shares of 20% and 23% respectively (see Figure 6). This indicates that the types of data processed in different research contexts are associated with different difficulties in processing and provision. In particular, the sensitivity of the data collected and the associated data protection constraints sometimes differ greatly between research contexts.
Compared to data sharing, the picture is reversed for citizen science: While only about 6 to 8% of the respondents from the life, engineering and natural sciences stated that they 'often', 'very often' or 'always' collaborate with non-scientific actors in their research, more than 23% of the social scientists and 11.5% of the humanities scholars stated this (see Figure 6).
Figure 6 Distribution of different Open Science practices, by subject groups
Sharing research-relevant materials and codes is most common in engineering. For 41.3% of respondents from this discipline, this practice has become routine. In the other subject groups, the degree of prevalence is significantly lower. While the social sciences, life sciences and natural sciences are relatively close together here with 25.6 to 31.3%, only 12.3 % of the humanities researchers regularly share code or material (see Figure 6). This comparatively low value is mainly explained by the fact that more than half of the humanities scientists state that this practice is not relevant in their research context (not shown). If one excludes those in whose research practice code and material sharing is not relevant, there are smaller but still noticeable differences between subject groups (Figure 7).
To date, open peer review has been practiced particularly in the life sciences and engineering (26.5% and 23.6%, respectively). There are hardly any differences between the humanities, social sciences and natural sciences for this practice - in these three subject groups, between 17 and 18% of researchers regularly participate in alternative peer review procedures.
Figure 7 Distribution of different Open Science practices, by subject groups, only with relevant research context
Overall, there are clear discipline-dependent differences in the implementation of the open science practices considered here. This indicates that some research contexts are more suitable for a particular open science practice than others. Consequently, science policy governance should not treat all subject groups in the same way and measure all researchers by the same yardstick. Rather, a differentiated view of the implementation of individual open science practices that does account for the diversity of different research realities is needed.