4 way comparison of Data Citation Principles: Amsterdam Manifesto, CoData, Data Cite, Digital Curation Center FORCE11 Data Citation Synthesis Group Should define data!

DataCite: DataCite was founded in 2009 to “increase acceptance of data as legitimate, citable contributions to the scientific record,” among other objectives. A DC: 1 and C-I are similar as to intent, but C-I goes a bit further, and emphasizes the need to elevate the status of data citations to that already accorded citations to other types of objects that comprise the scholarly record Recommendation: General agreement; modify to capture sentiments of all.

F-2 and C-III both mention persistence, but differ in that F-2 specifies the mechanism of “public” repositories (which is partly about access), while C-III is agnostic about the nature of the repository and mode of access (open or not). There is a legitimate debate to be had over the value of open access, but citation practices need to be applicable to data stored in repositories that are either open or subscription, public or privately owned (and even to data not stored in a repository.) DataCite: Research data for which identifiers will be assigned must be located in data centres or repositories committed to persistence and maintenance.” Recommendation: Do not get into the specifics of persistence of data and what should be kept for how long; do make it clear that all data may not be fully public; do recommend that data be stored in places committed to maintenance and persistence rather than on an individual’s machine. Make it clear that the citation may outlive the data

We can infer that the reasons for using the mechanism specified in F-are those referred to in C-I, C-II, C-V, and perhaps C-VI. DataCite Metadata Schema supports the ability to describe a rich set of relationships between the article and data, including IsCitedBy IsSupplementTo IsReferencedBy IsDocumentedBy suggesting that a registered dataset could “provide” metadata to assist with its placement in relationship to the article. A Recommendation: Merge: data should be cited…so as to facilitate credit, discovery, provenance and attribution. Perhaps need to be a couple of levels of detail here

DataCite recommends a bibliographic citation style, but makes no statement about its location in the publication. A F-4 is about the means rather than the purpose. We can infer that this is a means to enforce the purpose mentioned in C-I. The degree to which a data citation should resemble a bibliographic citation is debatable. It might be better to specify the purposes and functions that the citation should fulfill or facilitate, and leave the details of implementation to the communities who will need to implement them. Recommendation: Use CoData but perhaps give example that a bibliography might be an appropriate place For example, in our current mode of publishing, it should be handled at the same level as bibliographic materials Recommendation: Use CoData but perhaps give example that a bibliography might be an appropriate place For example, in our current mode of publishing, it should be handled at the same level as bibliographic materials Major discussion point: Data need to be distinguishable as data

DataCite: a persistent approach to access, identification, sharing, and re-use of datasets” is offered by DataCite, which uses DOIs to achieve this aim. A DCC 1: The citation itself must be able to identify uniquely the object cited, though different citations might use different methods or schemes to do so F-5 goes beyond C-IX in recommending a DOI as the specific type of persistent identifier, which is a means rather than a purpose. The purpose of using registries of persistent identifiers such DOIs, ARKs, or other handles is to provide persistence of findability if the location of a digital object changes. (Purpose stated in C-III.) Many communities of practice have already developed systems of persistent identifiers, some of which pre-date the existence of DOIs. Some people argue that proper use of URIs (and the redirect mechanism that is already a part of HTTP) could accomplish persistence without the intermediate step of a PID registry. While use of widely-accepted metadata standards (See C-IX) helps 5 to ensure interoperability, it seems presumptuous to specify DOIs over other PID systems already in use. Recommended: Persistent method of identification, machine readable but agnostic to the exact system, globally unique; widely used by the community In the example section: Examples of several identifier systems would probably be helpful Example of what is meant by globally unique In the example section: Examples of several identifier systems would probably be helpful Example of what is meant by globally unique

F-6 specifies a both a purpose (actionability by both humans and machines) and a particular means (a landing page). C-IV specifies only the purpose. The best means for accomplishing this purpose may evolve over time. DCC 3 a: it must provide the reader with enough information to access the dataset; b. indeed, when expressed digitally it should provide a mechanism for accessing the dataset through the Web infrastructure Data Cite: “Clients will ensure that the URL assigned to the identifier provides users with the necessary information for making meaningful use of the data. Often this will be in the form of a landing page…It is best practice to have a landing page for all registered data…” B Recommendation: Perhaps remove specific example, e.g., landing page and say it should be Resolvable to information regarding data accessibility and use. If you are authorized to access the data, you should be able to gain use of the data** Recommendation: Perhaps remove specific example, e.g., landing page and say it should be Resolvable to information regarding data accessibility and use. If you are authorized to access the data, you should be able to gain use of the data** Example: Could be a landing page What does it mean to make data actionable? Does that mean you must supply code? Metadata?

F-7 refers generally to the need to identify the specific version of the data being referenced. C-VI, C-VII, and C-VIII refer to distinct aspects of version: Provenance, Granularity, and Verifiability. The distinction among these aspects is useful. DataCite supports content negotiation: dex.html DCC 1: The citation itself must be able to identify uniquely the object cited, though different citations might use different methods or schemes to do so. 2. It must be able to identify subsets of the data as well as the whole dataset. Explicitly include versioning just because of pragmatic aspects. But also need the issues of granularity/subsets for dynamic databases -Provenance -Granularity -Verifiability, e.g., versioning Explicitly include versioning just because of pragmatic aspects. But also need the issues of granularity/subsets for dynamic databases -Provenance -Granularity -Verifiability, e.g., versioning Example: Look at DCC recommendations and opine them “Snapshots and time slices” -attribution time stamp Example: Look at DCC recommendations and opine them “Snapshots and time slices” -attribution time stamp

F-8 and C-II both address the function of attribution. C-II draws the subtle distinction between legal attribution and scholarly norm of giving credit to others for work they have performed. These are similar concepts, but there are some important differences between them, and what is necessary to accomplish them. The DataCite Metadata Schema supports the ability to supply metadata for unlimited contributors, including name, role (or type) and identifier information. A DCC 4B: In particular, there need to be services that use the citations in metrics to support the academic reward system, and services that can generate complete citations. Summary: Leave principle in; perhaps leave legal attribution in but provide example. Perhaps add “Normative” in addition to “legal”. Is citation of data exactly the same as citation of papers? Are additional mechanisms going to be necessary? Different types of data that will be cited in different ways. All of the previous principles allow this to happen.

Associate it with several examples but leave in.

Summary: Next steps Create a document that includes the results of this discussion – Maryann and Dan will take first pass – We should pull together examples when at all possible – Create versions of a consensus principle where we can – Note areas that need to be discussed Two products: – Pithy declaration: can refer people to appropriate examples – Article that has embellishments

Meeting Agenda: – Not one group or even a couple of group’s meeting; synthesis group meeting Want all groups to be included; should be able to provide comments both at the meeting and subsequent to it Leaders of discussion: people on these calls and are most familiar with the crosswalk and drafting the dissemination plan – Consensus of participants on principles – Quite a few people attending (30-40 people); need to manage meeting very tightly