Rec. 1: Definitions of FAIR

[sjDCC]

FAIR is not limited to its four constituent elements: it must also comprise appropriate openness, the assessability of data, long-term stewardship, and other relevant features. To make FAIR data a reality, it is necessary to incorporate these concepts into the definition of FAIR.

• The FAIR principles should be consulted on and clarified to ensure they are understood to include appropriate openness, timeliness of sharing, assessability, data appraisal, long-term stewardship and legal interoperability. Stakeholders: Global coordination fora; Research communities; Data services.

• The term FAIR data is widely-used and effective so should not be extended with additional letters. Stakeholders: Research communities; Data services.

• The relationship between FAIR and Open should be clarified and well-articulated. FAIR depends on appropriate Openness which can be expressed as ‘as Open as possible, as closed as necessary’. Stakeholders: Research communities

[ChiaraGabella]

SIB position : We believe that there is an “implementation gap” between the abstract definition of FAIR principles and functional FAIR services. FAIR enumerates good principles, but it appears to us that at this stage it requires an excessive effort to reach formal compliance. This has a cost (in time, knowledge and developer resources), that is often underestimated. We believe that the community needs simple and accessible guidelines. Another problem is that the descriptions are at the same time sufficiently abstract to provide specific useful guidance and at the same time very technocratic and bureaucratically rigid to allow people to find their own practical and simple solutions for concrete cases. The risk in creating a bureaucratic framework with lots of layers of technical requirements or documentation, is to create a system in which the level of “FAIRness” will not be necessarily representative of the level of quality and usage of the resources.

[WietskeD]

The University of Groningen and the UMCG: FAIR data already has the possibility of appropriate openness, the assessability of data, long-term stewardship. A clarification of this could be helpful for the community. FAIR data and Open Data are not the same. In fact, FAIR data goes much further than Open Data and is of more value for the research community. Open Data that is not findable or interoperable is of little use. We support the proposal to clarify the FAIR principles and not to extent the acronym, because of its broad adoption.

[hollydawnmurray]

F1000 position: The third point in this action will require careful handling. As mentioned by @WietskeD above, FAIR data and Open data are not the same – and in fact, this distinction has been made explicit in FAIR workshops, talks, training sessions, etc. Flagging this as it seems in direct contrast to incorporating Open into FAIR, and likely will mean this requires significant effort.

[holubp]

BBMRI-ERIC Position: While generic FAIR principles constitute the common ground, individual communities must be allowed to extend them as they deem necessary (including adding letters if necessary).

FAIR should be subject to critical reviewing as everything else in sciences. BBMRI-ERIC RI community has already published a consensus on extending FAIR to FAIR-Health, with particular focus (a) on extending the FAIR principles to other critical research resources, namely biological material (which is often the source of the data) and software, (b) on reproducibility and meaningful data reuse, on (c) incentive schemes, and on (d) privacy preservation for data sharing [Holub, Petr, Florian Kohlmayer, Fabian Prasser, Michaela Th Mayrhofer, Irene Schlünder, Gillian M. Martin, Sara Casati et al. "Enhancing Reuse of Data and Biological Material in Medical Research: From FAIR to FAIR-Health." Biopreservation and biobanking 16, no. 2 (2018): 97-105.]. Another term complementing FAIR in health-related research is FACT: fairness, accuracy, confidentiality, and transparency [van der Aalst, Wil MP, Martin Bichler, and Armin Heinzl. "Responsible data science." (2017): 311-313].

[Falco-KUB]

I think that "appropriate openness" should be specified more. In our view, FAIR data don't need to be open but the metadata should be.

[katerbow]

DFG position: Consulting with scientific communities with respect to their disciplinary requirements and how FAIR can be aligned with the specific disciplinary working culture is considered as an integral part of putting the FAIR principles into practice (priority on first bullet).

[bertocco]

INAF (astronomy) position: We agree on the issue statements but we would like to underline the difference between open and public data. We consider FAIR principles generally valid, while implementation and possible extensions are in charge of specific research domains.

[ScienceEurope]

Science Europe is currently finalising its Research Data Management (RDM) policies which comprise Core Requirements for Data Management Plans (DMP) and Criteria for Trusted Repositories. These policies cover all aspects of the FAIR principles and even go beyond those. Science Europe fully agrees with the principle that data should be ‘as open as possible, as closed as necessary’.

[ferag]

I would like to remark the importance of the data quality, especially on the "R", making the data Re-usable but ensuring quality.

[RCN2018]

• The RCN supports that the distinction between FAIR and open need to be clarified. The term Open is often misunderstood, especially when it comes to sensitive data. FAIR metadata is important and to emphasised that FAIR meta-data should be a minimum condition when the data itself cannot be shared. Data that cannot be made fully openly accessible may nonetheless be made accessible to specific users accord¬ing to defined access criteria. The need for restrictions may also change over time, allowing the data to be made accessible at a later point. • Where will the information about access rights will be store and how? Will access policy be made into some tagging mechanism to be machine readable? Should this be part of the FAIR data metrics? And similarly, what about preservation time: for how long should data be kept FAIR and Open? Who should make that decision?

[pkdoorn]

The FAIR guiding principles are fine at the level of the general idea of the four letters. The specification of the principles at the level of F1, F2, etc. needs to be improved. First of all, many of the statements say “(meta)data …” -- and once even, under R1, the parentheses are incorrect “meta(data)”. It is often unclear what applies to data (actual content) and what to metadata (documentation/description of data). The FAIR apostles explicitly state that FAIR is not about “Open”; yet, there are FAIR principles about licenses, and clearly data with an “open” license are more accessible and reusable than data with restricted or closed access. Another aspect not covered by FAIR, neither in this recommendation, is the accuracy and validity of the data. These two points are essential for data to be reusable.

[mromanie]

ESO position Full support of FAIR in its four constituent elements, of course. We also agree that FAIR in its four constituent elements is not broad enough, but:

• What is the process to come to an agreed extension of the definition?

• Using the same term for two different meanings, as suggested in the second bullet, is bound to create confusion

[gtoneill]

The FAIR acronym is already culturally embedded and should not be adjusted or extended. That said, there is room for improvement on adjusting and adding to principles within the four FAIR components. This should only be done through careful and inclusive consultation with all major stakeholders and should not be a frequent occurrence: otherwise the principles become less standard and attractive. At the same time, the FAIR principles should be left general enough to be cross-disciplinary and cross-sectoral as well as to be further interpreted and filled in within each discipline and sector. The distinction and choice between FAIR versus Open Data should be made clear for all stakeholders.

[npch]

SSI position:

The FAIR Guiding Principles state that “Importantly, it is our intent that the principles apply not only to ‘data’ in the conventional sense, but also to the algorithms, tools, and workflows that led to that data.” However the discussion around this recommendation is focussed on the limitations of FAIR solely when concerned with data. When considering differing implementations and extensions of the definition, particularly around “interoperability” and “reusability”, other outputs such as algorithms, tools, workflows and other forms of software should be addressed.

This is also true around the discussion of FAIR vs Open, and the implications for closed source software or occasions where there is a hybrid such as a closed source plugin on an open source platform.

[aidaturrini]

I agree with all the comments above, but I found particularly interesting the BBMRI position. They talk about the problem of re-use. An appropriate data-driven culture must be promoted. Reading data documentation, other than metadata, is necessary, like the knowledge of ISO, CEN, and all helpful standards. A specific library would help in this regards. Some rules should be also be included like a mandatory feedback by the data deliverer on the interpretation of the results derived from the data use. This will avoid unintended mistakes due to some elements the user not have because related to a specific research field. Not always the knowledge of a phenomenon is complete through data only. There are rules that it is not yet possible to transfer completely via metadata, documentation, and data itself.

[fniccolucci]

Within the next CIDOC2018 conference (CIDOC stands for the Documentation Committee of ICOM, the Intl. Council of Museums) we organise a workshop - probably the first of a series - for the Cultural Heritage & Archaeology community, to discuss the implications of FAIRness, titled "How fair are FAIR data?". It is more or less on the line indicated by BBMRI above. The stress is here on the "R", which is anticipated to split into an Rcube: Re-usable, Reliable and Relevant - not in the acronym of course, but in practice. In this community, "F" is being addressed by large-scale projects (e.g. ARIADNE) as well as by national initiatives. "A" and "I" are well-advanced, although as everywhere there are issues of openness - but the solution of "open as possible, closed as necessary" seems acceptable by all, even by conservative ministry officers! "R" is actually a different dimension than "F-A-I", and should be properly addressed from the community's perspective, to show researchers that Re-usable means also Useful to re-use.

[rachbruce]

[@jisc] we support the use of and reference to FAIR as a widely used and understood acronym, with a helpful message. We support the recommendation set out above. From our perspective, as few aspects should feature more prominently in the definition: 1) Disciplinary differences in the adoption and application of FAIR should be recognised more strongly; e.g. different disciplines might have different requirements for how and when data can be make available (or open) or be at different stages with regards to disciplinary practice. 2) The relationship between FAIR and open is a crucial point for the adoption of FAIR. While FAIR should always imply “as open as possibly, as closed as necessary”, ethical, security, and data vulnerability considerations are important factors which have differing impacts on how (and to what degree) FAIRness can be achieved in different disciplines. 3) Some further attention to the scope of how research is ruely FAIR - for example we read "data" as beingn inclusive of other elements required as part of a research object so it can be truly interoperable. and reuseable - @ssi comments should be taken into account in this regard. 4) Data quality and accuracy is important as raised by @pkdoorn.