2.1 Procedure of the literature review

The literature review was performed on the platforms ScienceDirect, Web of Science and IEEE Xplore. The review was last updated in November 2023. Only results newer than the original FAIR Principles [10] were considered relevant, causing results to not date back further than 2016. To perform the review in context of the research question, a search string was compiled based on the terms research data management and engineering and survey or synonymous terms, namely analysis, audit, check or inquiry. The resulting search string is explicitly formulated not as strict towards mechanical and industrial engineering as the research question to find results from neighbouring fields. The search string was used in three search engines listed in table 1. Afterwards, the results of the search engines were filtered for their Year (see table 1). Lastly, the resulting papers were exported in the .ris format along with their abstracts.

Table 1: Used search engines, filters and results for the literature review

The .ris files were imported to the PICO Portal to screen the collected papers for their relevance based on their abstracts. For this screening, exclusion criteria were formulated. These are listed in table 2. Any papers matching the exclusion criteria (n=194) as well as any duplicates (n=121) were removed from the review process. It has to be mentioned that the full text of Todorova et al. about ”Comparative Findings from Data Literacy Survey in Three Bulgarian Universities” [11] was not accessible at the time this paper was written and is therefore not included. A complete flow diagram of the process is depicted in 13 in the Appendix in subsection 8.1.1.

The resulting 23 papers were screened a second time, but based on their full texts. Again the exclusion criteria from figure 2 were used, resulting in two excluded records for the first criteria, four for the second, seven for the third and three for the fourth criteria. One of the found papers ([12]) summarised another paper ([13]) which was directly cited instead on relying on the summary. Lastly, six papers have been chosen by the full text review.

In addition to the systematic literature review, other sources of literature have been considered as well. The journals ing.grid and BausteineFDM have also been consulted to identify papers that are relevant but are not listed on the aforementioned platforms. Also, Zenodo as an catch-all repository has been consulted. BausteineFDM contained one more paper relevant in this context while in ing.grid’s preprint server, two additional papers could be found. Zenodo included three additional relevant publications. These six papers are also included in this review.

2.2 Results of the literature review

In table 3, the results of the literature review are shown, sorted by their most common statements on the status quo of RDM. No literature found contains direct information on the status quo of RDM in mechanical and industrial engineering. Yet, insights into current RDM practices and issues are granted, not only for engineering but also neighbouring fields.

Table 3: Results of the literature review

For instance, the most prominent topic in literature are the need for awareness amongst researchers. This seems to be a global and cross-disciplinary problem, as it is mentioned in almost every record found by the literature review. Still, two records state, RDM awareness is not a problem. One of them is based on ”spotlight investigation” [22] based on expert interviews, which might cause a bias on the results. The other is a RDM survey from Slovenia with no specific focus on a research area. [25] Similarly, the need for training and instructions are often mentioned.

While the need for resources is less often mentioned than the aforementioned aspects, the records that empathised on this aspect point out the importance of the effort connected to the application of RDM. This also, to some extend, is mentioned by the papers referring to the need for (specific) RDM tools, as these facilitate the application of RDM. However, this seemingly stands into contrast to the fewer mentions of need for support, which indicates that the effort of RDM can not be outsourced but has to be applied in the very context of a specific project. This is also supported by the many mentions of need for training.

One last interesting investigation has to be made. While the need for awareness is the most mentioned aspect, the need for incentives is the least mentioned one. This leads to the conclusion, that the intrinsic motivation for RDM is more important than external factors enforcing it. Researchers should be aware that and why their RDM is important not only to themselves but to others as well.

All publications presented either include RDM (in engineering) in a broader (e.g. nation wide) survey like [25] and [21] or refer to certain use cases or projects like [17] or [20]. The focus on RDM in Germany can only be found in related fields like IT sciences [14] or physics [19] or are not part of a survey but a case study [20] or a ”spotlight investigation” [22]. While the presented literature does not fully match the scope, it still offers insights on related fields of the research question. All relevant findings are discussed and compared to the results of this article in section 5.

3.1 Interviewees and Approach

The survey took place from October to December 2020. 168 researchers were interviewed, most of which are employed as research assistant seeking a doctoral degree (64%) (see figure 2). The distribution of participants in the survey is slightly (ca. 4%) shifted towards more research assistants and less professors in comparison to the average in the field of German engineering sciences under consideration. [26] Based on the most recent data available for the distribution of scientific staffing in Germany from 2021, this means about 0.3% of German engineering researchers in total (total population: 56,332) and about 0.8% of the specialist areas “General engineering sciences, industrial engineering with an engineering focus and mechanical engineering/process engineering” [26] (total population: 20,355) were reached with the survey.

Figure 2: Distribution of answers on the question: In which position are you employed? (Key:AM03)

As the research question is focussed on mechanical and industrial engineering in German Engineering Sciences, the target group of survey participants was chosen accordingly. Hence, the surveyed researchers are composed of members of the ”Scientific Society for Production Engineering” (”Wissenschaftliche Gesellschaft für Produktionstechnik”, in short WGP), the ”Scientific Society for Product Development” (”Wissenschaftliche Gesellschaft für Produktentwicklung”, in short WiGeP) and researchers from the RWTH Aachen Cluster of Excellence ”Internet of Production” (IoP) as well as members of the ”Fraunhofer-Verbund Produktion”. These consortia stand for ”Cutting-edge research […] in the area of basic research as well as applied and industrial research” [27] with a ”close collaboration with economy and science” [28] as well as a strong focus on ”application-oriented research” [29]. The IoP states a ”balanced composition of participating researchers from five faculties at RWTH Aachen University and six non-university research institutions” on their website [30]. E-mails were distributed to leading entities of these scientific groups who agreed to further disseminate the survey amongst their employees, asking for participation in the survey. As a result of this dissemination method, the exact number of researchers who received the survey is unclear, meaning the response rate can not be calculated but estimated. The WGP has about 2,000 members, the WiGeP has circa 1,200 while the IoP unites about 600 researchers and the Fraunhofer-Verbund Produktion consists of about 3,000 employees. [27], [28], [30], [31] With this estimation about 6800 researchers were contacted, resulting in an estimated response rate of 2.5%. About 43% of the respondents work at the RWTH, while circa 39% originated from universities all over Germany. Lastly, 18% are employees of Fraunhofer institutes.

All of the listed organisations are focused on engineering, particularly in mechanical engineering and production technology. However mechanical engineering often involves interdisciplinary approaches. Thus, plenty of subject areas are represented within the interviewees. As a result, the survey represents not only Archetype Frank but also gives insights into Community Cluster 41 (CC41). Figure 3 depicts the subject areas of the interviewees. More than half of the surveyed researchers are from the subject area of mechanical engineering. The other half is a wide mix of different subject areas. While some more are in the scope of mechanical engineering and production technology than others, all of them are researching within the context of production technology.

Figure 3: Distribution of answers on the question: What is your field of research? (Key: AM05)

3.2 Survey Structure and Questions

The survey consists of 39 questions with 216 question items on 14 pages. Only fully filled questionnaires were considered within the evaluation of the survey. Participants spend in average 14 minutes to fill the survey. The survey started with a demographic inquiry of the respondents’ data to validate the fit of the respondents. This is followed by an exploratory self-assessment, which contains three introductory questions to the overall usage and knowledge of RDM.

Interviewees were questioned if they are aware of the FAIR principles [10] for research data, if they (or a third party, if applicable) create a data management plan and if they base their research on the data life cycle. The self-assessment is followed by detailed questions of how research projects carried out along the data life cycle as proposed by forschungsdaten.info [32]. One question for RDM-tools and one for file formats used hold the majority of question items, as the usage of many tools and formats were queried. The questionnaire is rounded off by the question about the RDM-specifications and -support available to the respondents. The opportunity to add further comments via free text is given to the respondents throughout the survey, which only counts as a separate question when it is not an opportunity for further explanation or extension of answers to an existing question and is otherwise included in the number of question items rather than the questions number. The structure of the questionnaire with question categories of the survey and the corresponding numbers of questions and question items contained can be found in table 4. Free text answers are included within the numbers of questions stated in the table.

4.1 RDM Knowledge and Perceived Relevance of RDM

The first set of non-demographic questions aims at providing a rough assessment of the respondents knowledge on RDM in general. Regarding research data handling, more than half of the respondents stated that their knowledge was moderate or lower. Only 42% stated that they had a high or very high level of knowledge regarding the handling of research data (see figure 4). At the same time, over 57% of respondents rate RDM as important or very important. Only about 15% perceive RDM as unimportant or completely unimportant (see figure 5).

Figure 4: Distribution of answers on the question: How high do you rate your own knowledge of handling research data? (Key: EF01)

Figure 5: Distribution of answers on the question: How important is/was research data management to you in your personal dissertation project? (Key:EF02)

When comparing those two statements above, there seems to be a gap between the group of researchers with (very) high RDM knowledge and a (very) high perceived importance of RDM. There are 14% less researchers who have a RDM-knowledge specified as high or above than there are researchers who perceive RDM as at least important. This leads to the first hypothesis of this paper, that there is a gap in knowledge of researchers. Additionally, missing knowledge may also lead researchers into perceiving RDM less important, potentially widening the gap.

• 1.

  There is a need for RDM knowledge among researchers in the engineering sciences, specifically for researchers of the Archetype Frank respectively amongst researchers in the field of mechanical engineering and production technology (CC41).

To better understand the relevance and reliability of the self-accessed RDM knowledge, the following question was asked: ”Have you ever heard of the FAIR principles (Findable, Accessible, Interoperable, Reusable) [10] for research data?”. The responses are mapped on the answers from figure 5 and shown below in figure 6. It becomes apparent, that the relevance of RDM in ones own dissertation and knowledge about the FAIR principles are somewhat correlated, yet it is unclear which is caused by what.

Figure 6: Importance of RDM in one’s own dissertation in dependency of the share of respondents who have heard about the FAIR principles (Findable, Accessible, Interoperable, Reusable) [10] for research data (Keys: EF02 over DL01_01)

The survey also asked for the usage of the Code of Conduct of the ”Guidelines for Safeguarding Good Research Practice” published by the DFG [33]. These have already been applied several times by almost three quarters of all respondents (see figure 7), however this does not lead to a consistently high level of knowledge regarding research data management. The correlation coefficient between these factors is 29%, which does indicate a mild correlation. Generally speaking, the correlation coefficient measures how close two values are linearly dependant [34]. As the correlation coefficient is positive, this indicates an increase in RDM-related knowledge when a person regularly uses the DFG guidelines. This effect can also be seen in figure 7.

Figure 7: Perceived relevance of RDM among the participants in dependency of the usage of the Code of Conduct of the ”Guidelines for Safeguarding Good Research Practice” by DFG (Keys: EF01 over EF03)

A similar effect, can be seen between the perceived relevance of RDM in the interviewees own dissertations and the knowledge about RDM (see figure 8). Here, the correlation coefficient amounts to 33%, indicating a mild positive correlation, meaning that the more important RDM is perceived in context of the one’s own dissertation, the more one knows about RDM [34].

Figure 8: Perceived relevance of RDM among the participants in dependency of the perceived relevance of RDM in the researchers own dissertation (Keys: EF01 over EF02)

4.2 Application of RDM Related Tasks

While 58% (see figure 8) claim to find RDM important in their own dissertation, the self-assessed knowledge amongst the interviewees is mostly moderate to very low. Moreover, the claim of regular use of the ”Guidelines for Safeguarding Good Research Practice” is questioned by the answers of the interviewees in the later questions of the survey. For example: The Guidelines state that “Researchers decide autonomously […] whether, how and where to disseminate their results.” This includes the process of determining copyrights and the control of access, which is especially important when handling data that is not shared due to reasons such as secrecy or of patent applications. In that case, a decision has to be made to control the access to only those who are allowed to access such data. However, less than 10% of the interviewees regularly determine copyrights, control access or share their data (see figure 9).

Figure 9: Activities from the sharing phase - Distribution of answers on the question: Please indicate whether and to what extent you use the individual steps of the data life cycle. (Keys: DL02_15 to DL02_18)

Even less make their data publicly available (<5%). To set this into perspective, 44% of the surveyed researchers claimed to regularly use the DFG’s ”Guidelines for Safeguarding Good Research Practice” [33]. In other words, only about one in nine researchers who regularly use this guideline ”make all results available as part of scientific/academic discourse”, although research data should be included ”where possible and reasonable” [33] as proposed by the DFG.

Similar low rates of regular application of research data management tasks can be observed throughout various steps of the data life cycle. This leads the following hypothesis:

• 2.

  While the use of Guidelines like the ”Guidelines for Safeguarding Good Research Practice” tend to improve the self assessed RDM knowledge among the interviewees (see figure 7), it does not necessarily imply the application of RDM connected tasks.

The only step of the data life cycle that has a high rate of regularly performed tasks is the ”prepare and analyse data” phase, as shown in figure 10. The highest rated task is ”Interpret data”, which scores a 38% regular application rate. An additional 36% occasional application rate is adding up to 74% of the researchers who at least occasionally interpret their data on their own. Taking into consideration that 16% of the interviewees are professors or academic councillors, this initially rather low rate of data interpretation among researchers becomes clearer.

Figure 10: Activities from the prepare and analyse data phase - Distribution of answers on the question: Please indicate whether and to what extent you use the individual steps of the data life cycle. (Keys: DL02_07 to DL02_14)

This leads to the next hypothesis put forward in this article:

• 3.

  RDM-related tasks that are not directly part of the everyday research activity (like determining copyrights) are much less likely to be carried out than those who are mandatory to receive results from data, such as transcribing, preparing, interpreting or validating data.

4.3 Data Sharing with Third Parties

Another set of questions asked about the willingness to share research data with third parties and the reuse of third party research data. This set of questions however seems to be inappropriately specified, as the results are inconsistent. One participant gave feedback on this topic:

”The questions [regarding sharing research data with third parties] are flawed, as the attitude towards any third party is different than within the institute or a network.”

Anticipating focus group interviews that took place months after the survey with different participants, it can be said that this definition of ”third parties” harshly varies in the understanding of researchers. The questions in this survey aimed towards the interpretation of third parties as ”not related to the research project in any way”. This however seems to be misinterpreted by some of the participants. The questions that asked for the data life cycle, specifically the ones for the sharing data phase, show that 57% shared data at least once, which was shown in figure 9.

When asked for the actual possibility for third parties to access one’s own research data, this value raises to 65%. This can be explained in two ways:

1. The additional 8% of interviewees did not specify an answer in the corresponding question set at the data life cycle section of the survey.

2. The surveyed researchers interpreted the expression ”third party” as ”involved in the actual research project, but not part of the own institute”.

It is unclear which of the two applies in this case. It has to be noted that, although the expression ”third parties” is used in the ”Guidelines for Safeguarding Good Research Practice”, it is never specified in the document itself [35].

4.4 Usage of RDM Tools and Services

The next part focused on tools and services. A distinction is made between usage and awareness of tools. The term usage refers to the following options: ”regular use” and ”occasional use”. Awareness means the tool is either ”known by name” or has at least a ”one-time use”. Respectively, unawareness refers to the option ”unknown”. A ”not specified” option was given as well.

More than 70% of all responses are ”unknown”. A further 19% are assigned to the answer option ”not specified”. It has to be noted that this distribution also applies if only the answers of those are taken into consideration, who have stated to have a high or very high self accessed RDM knowledge. In this case, 69% answered ”unknown” and 20% answered ”not specified” or did not answer the question at all. In general, the answers of the respondents are strongly polarised. A few tools stand out due to regular use, while others are almost completely unknown.

Literally the most prominent example is Git, with 72% awareness among respondents. Almost 30% use the tool regularly and 25% occasionally. 7% have used Git at least once and 10% are familiar with it by name. No other tool has a similar level of awareness and use among researchers. Although mySQL is better known than Git (78%), it is used much less frequently (regularly 12% and occasionally 22%) and is limited to one-time use (28%).

An overview of awareness (”known by name” and all mentions of useage) and usage (sum of the mentions of ”occasional” and ”regular use”) is given in table 5, sorted by the proportion of respondents who state multiple uses. Due to the large number of tools surveyed, only those used more than once by at least 5% of the respondents are mentioned below for the sake of clarity.

As shown in table 5, of the 90 tools and services surveyed, only 13 have been used more than once by at least 5% of the respondents. Seven of those 13 come from the field of software development, i.e., they are directly or indirectly related to programming. Those can be recognised by the categories ”Data organisation” and ”Databases and repositories”. The remaining six tools/services are two tools for citation (DOI Citation Formatter and STD-DOI), one for persistent identifiers (TIB PID Competence Centre), one for finding research data (Google Dataset Search), a password organiser (KeePass) and a tool for collaborative working (Microsoft Project).

As shown in table 6, a similar distribution can be observed when only reviewing the answers of researchers who have stated to have a high or very high self accessed RDM knowledge. Here, 14 have been used more than once by at least 5% of the respondents. The same focus on software development becomes apparent with eight of the 14 listed tools related to this area.

The majority of the best-known or most-used tools have in common that they offer solutions to researchers’ everyday problems (compare finding 3). For example, the versioning tool Git offers a possibility to version source code, which can hardly be kept manageable without versioning. The added value of Git is known and is also passed on to other researchers, at least in groups that regularly deal with source code. The immediate applicability is what separates those best-known and most-used tools from especially the less-used RDM tools.

Such RDM tools that should mainly accompany the research process, are virtually unknown and unused. The majority of respondents thus lacks knowledge about suitable programs, supporting tools or services in the context of RDM. Therefore, such programs, tools or services are not used by the majority of respondents, which is another hypothesis in this article:

• 4.

  Researchers lack awareness about existing solutions for RDM specific problems and therefore the knowledge and ability to use those solutions.

4.5 Specifications and Support Structures

The last question set is directed at the requirements and support structures for RDM that are specified or offered by the respondents’ respective institution. Those include, but are not limited to, RDM-Teams at universities, available tools for RDM or specific support at institutions. The exact question was ”Is there support within your organisation in the area of RDM?”.

Shown in figure 11 are the responses of researchers asked if they use offered support structures at their organisation. Only about one tenth of the surveyed researchers have used offered support structures while almost a quarter states there was no support available at their institution. The survey did not include any questions regarding why support structures are not used by researchers.

Figure 11: Distribution of answers on the question: Do you use the existing support services? (Key:FO06)

However, there might be two reasons for this. Firstly, support structures are available but not known, which is relevant only for the 23% of researchers who claim that there are none. Secondly, the benefit of such structures is not perceived as important enough to be worth the expense. One third of researchers who know about support structures do not use them despite having the opportunity to do so. This, in turn, might be a result of either insufficient support structures (may it be in terms of offered service, format or content) or lack of knowledge about how and why such structures could improve the interviewees RDM. The survey also asked for an evaluation of the offered support structures with the results being shown in figure 12.

Figure 12: Distribution of answers on the question: How good do you rate the support? (Key:FO06)

Combining the data basis from figure 12 with figure 11, there are several groups of researchers to be identified, clustered by their access to and their usage of RDM support, shown in table 7.

4.6 Further Open Questions

In further open questions, respondents were given the opportunity to mention possible reasons that might prevent researchers from RDM in the form of free text answers. Most interesting are the answers on the question ”What reasons could prevent researchers from sustainable RDM?”, which 39 of the 168 interviewees (23%) answered. A detailed list of quotes of the respondents can be found in the Appendix. The effort or workload for the establishment and operation of RDM is with 16 mentions the most recognisable reason against proper RDM usage. Likewise, the lack of clear standards or guidelines for RDM is cited twelve times, closely followed by the lack of awareness of RDM among researchers (nine mentions). This last statement is specified: RDM is primarily perceived as an additional expense, there is no incentive to use it and no necessity for RDM is seen. The lack of necessity is justified by the time-limited nature of projects and their isolation in research environments. Other reasons against RDM application are a lack of knowledge (seven mentions), the concern of data misuse or data usage without permission or citation (six mentions) and problems with missing or complicated support structures, which five interviewees mentioned.

The feeling that the own data can only be used for the own projects prevails for many. Contrarily, others who consider their data to be usable, fear data misuse. In this case the protection of the own research is seen as more important than a provision of data within the framework of RDM. This is expressed, for example, in the following quote from one of the respondents:

”Real data, e.g. from production, is not easy to obtain. Those who have such data sets have an advantage. Therefore, data is not shared, although it would be useful to promote scientific progress and test results for reproducibility.”

Many of the interviewees’ statements can be condensed into the following statement (adapted in wording for the purpose of anonymisation), which was formulated by an interviewee:

“ Besides the most obvious reason - lack of knowledge - I think [RDM] just meets [ignorance] by and large. One Example: For [research] I have collected publicly available data. Of course I maintain and cherish my data and go through large parts of the data life cycle, but for that I don’t need thousands of tools that nobody else [in my organisation] uses. It is also likely that others will not (be able to) continue to use this data - which is why it makes sense to maintain it sustainably. It is similar with research projects. The more isolated and smaller the project is, the less sense there really is in elaborate management […]. This is not only true for the data. Furthermore, it is unfortunately inherent in the research system that I could suffer great professional damage if I give out my data beyond a certain level. In applied research projects the situation is certainly different, but even here I need (at least initially) a more or less exclusive use of data so that I can firstly secure my livelihood. Furthermore, there are often confidentiality clauses that do not allow me to pass on the data.”

The free-text answers were used to formulate more hypothesis, as they allowed for deeper insights, especially when considering reasons against RDM. However, it needs to be mentioned again that those only originate from 39 of the 168 interviewees (23%) of the interviewees, which further diminishes the sample size.

• 5.

  The interviewees see the effort of RDM in terms of initialisation, familiarisation with it and everyday work as a reason that prevents researchers from sustainable RDM.

• 6.

  The interviewees name the lack of clear guidance through the RDM process like guidelines, standards or processes as a reason that prevents researchers from sustainable RDM.

• 7.

  The interviewees perceive that RDM as a topic does not receive enough awareness yet, which is a reason that prevents researchers from sustainable RDM.

• 8.

  The interviewees see a lack of knowledge among themselves and other researchers, which is a reason that prevents researchers from sustainable RDM.

• 9.

  The interviewees consider the risk of data misuse and data usage without citation or permission as a reason that prevents researchers from sustainable RDM.

• 10.

  The interviewees see the lack or quality of support structures as a reason that prevents researchers from sustainable RDM.

The acceptance of the reuse of data among the respondents is limited. Thus, the ”not-invented-here syndrome” [36] is cited by the respondents. This effect describes the rejection of ideas and inventions not founded in one’s own institution for reasons other than monetary ones. For example, openly available data might not be reused because it is not trusted as it is of other origin as the own institution. As a result, the subsequent use of existing data is omitted and additional work is done, since data must be collected by the institution itself [36].

5.1 Findings

The survey indicates that the knowledge, awareness and usage of RDM has to be fostered to enhance the management and therefore FAIRness [10] of research data. To achieve this, researchers firstly need to know what to do when starting managing research data (see hypotheses 4., 5. & 8.). An appropriate approach needs to be handed to them with a clear entry point and a structured and adaptable process needs to be defined (see hypothesis 6.). When questions occur, those have to be answered right away (see hypotheses 5. & 10.). Also, training materials to the very topic of the question have to be provided and suitable tools have to be introduced (see hypotheses 1. & 4.). Those materials should be light-weight and focused on applicability. Light-weight in this context means that provided information should only focus on the very specific problem of the researcher. A huge amount of additional and inapplicable instructions will compromise the will of researchers to use RDM and cause frustration. The process of RDM has to be embedded within everyday research (see hypothesis 3).

Incentivation for RDM usage needs to be provided as the requirements of, for example the DFG, are not sufficient to enhance the application of RDM (see hypothesis 2.). Also, the awareness for RDM has to be broadened (see hypothesis 7.). Suitable measures could be the requirements of RDM in connection with dissertations or bachelor/master theses.

Opposing to the incentivation is the fear of data misuse or missing citations of the own work (see hypothesis 9.). This could be addressed by the possibility of storing data in closed repositories and clear instructions of how data can be made publicly available in a way that it is unambiguously recognisable who the author is and to whom the data belongs. Access management and licensing has therefore to be taken into consideration, granting the possibility of a controlled reuse of data.

5.2 Comparison of hypotheses and related work

To conclude this discussion, a comparison of hypotheses to findings of the literature review shall be given, ordered by the number of hypotheses listed above. This comparison is drawn to different disciplines and countries than the scope of this survey. Yet there are some similarities and common challenges that form a reoccurring pattern in the nature of RDM. In the following, each hypothesis is referenced by its number and a short hand at the corresponding paragraph.

1: Researchers are missing RDM knowledge For instance, hypothesis 1 is supported by several papers. The ”lack of trainers in RDM practices” [15], ”lack of knowledge/training” [18], a lack of ”data sharing skills” [23], or the need of training as stated by Elsayed and Saleh [13] is represented in many papers. The only contradiction found in literature by Costanzo et al. states that ”Lack of RDM Knowledge [is a] low barrier” [16].

2: Guidelines do not equal good RDM Costanzo and Cooper support hypothesis 2, describing the ”lack of institutional understanding and awareness of […] expectations” [16]. Wilms et al. state, a ”requirement to comply with possible guidelines” [14] is not enough incentive for researchers to adhere to good RDM practices.

3: RDM is only done when necessary for results The third hypothesis is not supported by any findings in the literature. Therefore, this hypothesis could benefit from a revision in the future. However, Palsdottir states that RDM ”is not a normal practice” in the researchers work [21]. Still, the reasons for the usage of tools should be clarified. The hypothesis can not be supported by literature but is still a finding of this paper.

4: Specific solutions for RDM are unknown While Björnmalm et al. see the problem in too many generic and yet too few specific RDM tools [15], Israel et al. state that ”respondents continue to rely on […] paper laboratory notebooks” [19] instead of electronic laboratory notebooks. While there are many tools available for RDM activities both generic and specific [19], the ”lack of knowledge” [18] about these tools can be seen as the actual challenge RDM is facing in this context. This also supports hypothesis 4.

5: Effort is a hindrance for RDM Hypothesis 5 is also represented within literature. RDM is seen as ”a significant burden” [17] as ”the amount of time it takes” [18] is a ”perceived increased workload” [14], opposing a ”lack of resources (time, budget, personnel etc.)” [16].

6: Uncertainty is a hindrance for RDM Connected to the effort required for RDM, the lack of guidance (hypothesis 6) is found both in the answers of this survey as well as the literature. Björnmalm et al. found a lack of ”specific instructions (or links to relevant guidelines)” [15], which is supported by Costanzo et al. regarding the ”lack of institutional understanding and awareness of […] expectations” [16] as well as the findings of Borghi and Van Gulick that there is missing guidance through ”lack of best practices” [18]. The ”large number of tools and methods” [19] and ”complexity in data structures [,] formats [and] documentation” [19] is a challenge yet to be faced. As ”processes are not yet clearly defined, let alone standardised” [20] ”researchers needed assistance” [20] in RDM, which is also supported by [21]. Additionally, ”establishing […] guidelines” can improve RDM [22].

7: Awareness for RDM is low Many papers also address hypothesis 7, however some support it while others oppose it. While Björnmalm et al. see ”too few incentives for researchers that reward and incentivise implementation of RDM practices into everyday workflow” [15], Wilms et al. see that the ”overall acceptance of RDM policies is low” [14]. According to Austin et al. there is a ”need to demonstrate to researchers the value of data management” [17]. Simmilarly, Borghi and Van Gulick point out that the importance of RDM is not commonly known [18]. These four statements support hypothesis 7. Israel et al. point out that ”making data FAIR needs to start most importantly, awareness” [19], also supporting hypothesis 4 to some extend. However, Vilar and Zabukovec oppose these theories, stating that researchers are rather convinced by RDM [25]. Ortloff et al. also argue in their spotlight investigation that ”most of the partners are strongly aware of the benefits provided” [22] by RDM. The incentivation of RDM, as for example brought up by Borghi and Van Gulick, has to be addressed by funding organisations, universities and institutions. However, it is not part of this paper, as the focus lies on the researchers perspective on RDM. Still, the topic of incentives has to be considered from all sides, from making funding dependent on concrete RDM practices to the demanded RDM in the context of a dissertation.

8: Missing knowledge hinders RDM’s application While hypothesis 8 is not directly supported or opposed by literature, it is to some extend a consequence from hypotheses 1 and 4. Palsdottir states the ”limited knowledge” and that RDM ”is not a normal practice” as well as an ”urgent need to increase the researcher’s knowledge and understanding of the importance of data managenent” [21]. However, it can neither be contradicted nor be proven that the lack of knowledge hinders the application of RDM. The lack of knowledge has been stated several times, both in this survey and the literature. A plausible outcome might be the hindering of (sustainable) RDM.

9: Researchers fear data sharing The ninth hypothesis is addressed by five papers. Austin et al. state that more than half of the involved partners in the projects rejected data sharing [17]. This is mostly based on the ”concerns regarding IP protection” [22] respectively ”intellectual property rights” [24] and the ”fear of losing control” [14]. The ”partner’s consent for publication was the biggest hurdle” [20].

10: RDM support is insufficient Lastly, hypothesis 10 is supported by some papers. Elsayed and Saleh see a need for support [13] as well as [21], while Björnmalm et al. see a lack of ”support at a faculty level” [15], similar to the ”lack of availability of support materials” [16] stated by Costanzo et al. Wuchner et al. also see a need for support, but on a more immediate level. While the aforementioned papers focus on generic support, Wuchner et al. see a direct assistance needed for ”data publications – especially FAIR ones [because they are] are a major challenge for researchers” [20]. This last statement excluded, all papers revolve around the lack of support, which is partially true, but might also be a consequence of the lack of knowledge and awareness, as stated in hypotheses 1, 4 and 8.

8.1 Additional info on the review of related work

In the following, firstly the review process is displayed before an explanation on every included record and their contents.

8.1.1 Review process

A complete diagram of the review process with references to exclusion reasons is depicted in figure 13.

Figure 13: PRISMA 2020 Flow Diagram, c.f. [37]

8.2 Further information on results

Below, additional results of the survey are presented that do not directly contribute to the answering of the research question but may be beneficial for further research on different aspects of RDM.

8.2.1 Usage of File Formats

The survey also asked about the frequently used file formats. 31 file formats as well as opportunities for free text answers were given. The interviewees could choose whether or not they use that file format. File formats cover the MS Office family, PDF and common image and video formats as well as formats for quantitative data and text-based formats. The later ones also contain file formats for source code such as .py or .cpp.

When reviewing the results for file formats in text-based applications, a strong distinction between commonly used and not commonly used formats is possible (see figure 14). MS Word files (.doc or .docx), just like PDF documents, are frequently used by 87% of the respondents. With 78%, .txt is the most frequently used format for unformatted text. Other file formats are commonly used by a minority of the interviewees as shown in figure 14.

Figure 14: Common usage of text-based file formats among interviewees - Distribution of answers on the question: In the following, we ask you to mark the file formats you use frequently or to add further formats. (Keys: D101_12 to D101_18 and D101_23)

MS Excel files (.xls or .xlsx) are used by 87% of the respondents (see figure 15). Close behind (86%) is .csv, another file format usable in Excel. Again, other file formats are much less commonly used than the aforementioned, making the distinction between commonly used file formats and not commonly used file formats very unambiguous.

Figure 15: Common usage of file formats for quantitative data among interviewees - Distribution of answers on the question: In the following, we ask you to mark the file formats you use frequently or to add further formats. (Keys: D101_02 to D101_09)

For media files (image, audio and video files), the spread in the answers given is not nearly as pronounced as for example in quantitative data. However the aforementioned formats .jpg/.jpeg, .png, .mp3 and .mp4 are predominant for their respective category (see figure 16).

Figure 16: Common usage of media file formats among interviewees - Distribution of answers on the question: In the following, we ask you to mark the file formats you use frequently or to add further formats. (Keys: D201_25 to D201_30, D201_33, D201_34 and D201_37 to D201_40)

The commonality of the aforementioned formats is their general widespread use, familiarity and the resulting usability. All these can be used on a standard Windows PC with MS Office installed, without the need for further installations. The latter is a factor not to be neglected. On the one hand an installation of further programmes may have to be carried out by corresponding IT departments, which is associated with personnel and time expenditure. On the other hand, depending on the file format, there are licence fees for associated programmes. The latter becomes more important if there are free or already available alternatives in the work environment.

This relation is expressed most strongly in the processing of quantitative data, e.g. table-based evaluation of data through Excel. MS Office, including Excel, is one of the standard installations on Windows PCs, as already mentioned above. Therefore, the use of .csv, .xls and .xlsx files is possible on the majority of Windows PCs; these formats are used by 87% of the respondents. In contrast, the use of the .por format, which was developed by IBM for the statistical programme SPSS and is only used by 6% of respondents, is only possible in this very programme [38]. For other formats in the field of quantitative data, the usage rates are hardly higher and formats usable with Excel seem to be the only option. In contrast, only 15% of respondents use the .odt format, although this can also be opened and edited in licence-free and openly available programmes.

The usage of file formats is primarily based on programmes and tools available and the usability of the formats. The usability is partly dependent on the availability of programmes or their corresponding licences. It is unclear why specific programming languages and file formats (see figure 17) are used in software development. The reasons for or against an approach are not part of the survey, as researchers should be supported in everyday research and not forced into new directions. The collected knowledge about the used file formats used does not provide any direct recommendations for action to advance RDM. It rather shows the heterogeneous file formats that need to be taken into account when working with research data.

Figure 17: Common usage of file formats used in programming among interviewees - Distribution of answers on the question: In the following, we ask you to mark the file formats you use frequently or to add further formats. (Keys: D201_19 to D201_21)