White Paper — Open Science in a Digital Republic|
Opinion of the Ethics Committee of 7 May 2015: “The ethical issues of scientific data sharing”
“Data Sharing” Group
The ethical issues of scientific data sharing
- 1 See the article “Dix laboratoires mondiaux partageront données et chercheurs” (Ten world laboratori (...)
11- For the past two decades, data have acquired a central role in scientific production, regardless of the discipline. Researchers increasingly need vast data warehouses (big data) – but also datasets of more modest size (small data) – for exploring, viewing and comparing results, validating assumptions or formulating new ones, or even for automatically generating new knowledge (machine learning). Major infrastructures and shared platforms continue to be created for the archiving, storage and processing of information. Rapid advances in digital technologies have greatly improved the way in which data, information and tools can be disseminated, managed, used and reused between researchers, to constitute an ecosystem based around scientific publications. Movements favouring open access therefore become crucial to optimise the exploitation of vast deposits of data. No organisation has enough resources to conduct its work alone.1 The effort required to exploit the digitised knowledge is immense, especially since it requires human intervention at a certain point in the process (text mining is necessary but not sufficient). Facilitating access to and reuse of these data has thus become a crucial issue for sharing and circulating research results more rapidly.
22- However, attitudes with regard to sharing and openness differ greatly depending on the types of data and the disciplines. In certain disciplines (in astrophysics or genomics, for example) the benefits of this data sharing have turned out to be considerable, and the disadvantages small enough to enable a trend for data sharing to develop. For these communities, the matching and comparing of data are clearly sources of new discoveries, and they consider that any obstacle to the circulation of scientific results is not only ineffective but contrary to the fundamental principles of widespread and open pooling of knowledge.
3However, for other disciplines (especially in the human sciences), data are often collected individually: these data, which are related to the subject of the research, may be shared only with the same embargo conditions as those for the publication of results.
- 2 Incidentally, we essentially consider scientific data in this text, which excludes from our analysi (...)
43- Apart from the case of these “self-organised” scientific communities, government policies for open data2 have in recent years aimed for the broad dissemination of data subsidised by public funds. Some of these data may be of interest to scientists and, conversely, the scientists’ data may concern society. The data sharing movement must therefore be adapted to government open data policies that pursue significantly different objectives and are subject to different ethical and legal constraints.
- 3 Which enshrines the principle of “free access to research publications and data”. See www.horizon20 (...)
54- The purpose of this opinion is to examine how the different scientific policies could be coordinated in a much broader field: the ethics of sharing research data. While many researchers support data sharing, most feel powerless or even reluctant in the face of this government obligation to disseminate (the open data), which may seem paradoxical, or even counterproductive: encouraged to disseminate widely, as confirmed by the European Horizon 2020 programme,3 they must at the same time apply legal restrictions to this public dissemination of data, in the name of respecting privacy, copyright, the duty of secrecy, confidentiality and security. Faced with these injunctions, which may seem contradictory, it becomes necessary to inform researchers about their various obligations and about the ethical implications of their choices relating to the data that they collect, share or reuse.
- 4 H. Tjalsma & J. Rombouts, Selection of research data: Guidelines for appraising and selecting resea (...)
6The scientific data (research data) considered here relate to all the data collected in the context of scientific research,4 i.e.:
the primary data (empirical, observed, measured);
the secondary data, derived from the primary data, annotated, enriched and interpreted, adding value to the initial data and possibly involving other actors;
the metadata that structure, manage and facilitate the accessibility of the primary and secondary data.
- 5 P. Uhlir, “Revolution and evolution in scientific communication: Moving from restricted disseminati (...)
7These data may be text documents, graphs, pictures, multimedia or digital representations. The gap between data and publications also tends to be reduced with the concept of the open process, which consists in disseminating the knowledge and data used and created in the process of thinking and writing the scientific publication.5
- 6 http://www.budapestopenaccessinitiative.org/read
- 7 http://openaccess.inist.fr/
- 8 Nature 498/16–17 (5 June 2013), doi: 10.1038/498017a; the initiative http://genomicsandhealth.org/ (...)
8Successive agreements and charters have marked the history of the data-sharing movement. In 1996, for the first time, researchers involved in the sequencing of the human genome signed a series of agreements laying the foundations for the open sharing of pre-published data. Then the first definition of open data was given by the International Declaration on Free Access, in Budapest, on 14 February 2002, known under the acronym BOAI6 (Budapest Open Access Initiative). Since then, many other initiatives have seen the light of day, with for example the Berlin Declaration of 2003 on open access to knowledge in the sciences and humanities,7 followed in March 2005 by a new Declaration called Berlin III aimed at strengthening the measures adopted within the framework of Berlin I. Most scientific organisations, including the CNRS, have signed these declarations, thus legitimising this culture of open access. Several general recommendations are currently available (scientific consortia, OECD, etc.) and funding agencies rely on such principles to secure their requirements in this area, which constitute conditions for the granting of subsidies. Thus, in 2013 the Horizon 2020 initiative defined a European policy of openness and sharing of scientific data to which national bodies must now comply when European funds are involved. Similarly, one of the most recent initiatives again comes from the field of human biology, with the launch in 2013 of the “Global Alliance for Genomics and Health”.8
9The need for deliberation has already been included in the framework of the CNRS strategic plan “A better sharing of knowledge, an open policy for scientific and technical information of the future”.9 This mobilising strategy must be translated into ethical obligations for researchers, the main producers and users of scientific data. Part of this programme also provides for the “establishment of a charter of ethics transcending instrumental considerations” and reaffirming the goals of public research.
10Scientific activity relies increasingly on the shared creation and use of multi-source and multi-use data infrastructures. These recent transformations in the scientific process are related to three types of change:
the technological development of measuring instruments and sensors of raw data for producing masses of data;
computing capacity in terms of storage, archiving and analysis (birth of bio-informatics, for example);
collaborative Internet and networking, enabling databases and platforms to be populated directly online by numerous stakeholders and enabling economies of scale.
11This change is leading to a major upheaval in principles and practices, from hypothesis-driven research to the generalisation of data-driven research, i.e. to a process of construction based on data that are already formed. In this context, the data that are now being annotated, mined and analysed have become the essential components of the research activity. Multiple uses for data become the rule and the masses of data generated require infrastructures for multiple-use data to be created. The formation of data infrastructures used to underpin research operations (research processes), and no longer simply to archive results, becomes an important step in science. The capacities for data enrichment and annotation generate a need firstly for the corresponding databases to be monitored and developed, and secondly for metadata to be organised, to enable the effective use of the processed, aggregated and correlated data. Lastly, the data, which have been produced in vast quantities, independently of any particular hypothesis, often requiring large budgets that support a wide range of different research projects, become what are called community resources, and for which maintenance and access are essential for the collective action of the group.
- 10 In early 2014, five articles were published in The Lancet on the theme “Research: Increasing value, (...)
12However, while 10% of the available data from experiments are provided via publications, 90% remain on the hard drives of computers. Data do not circulate quickly enough in the scientific world. Regardless of the disciplines, too many results remain unpublished and much of the data are under-used or lost.10 The data from negative results are forgotten. How can researchers be encouraged to participate in the opening and dissemination of their data?
13The open sharing of scientific data intersects with another global movement that extends beyond the scientific field: the opening of public data, i.e. data subsidised with public funds (open data). Open data policies, developed at state level, require authorities and public institutions to make their data accessible for sharing. However, although these two trends reinforce each other, they are not based on the same rationale.
- 11 “The Academy of Science owes its origins both to circles of scholars, who at the beginning of the 1 (...)
- 12 Evelyne Barbin (ed.), Arts et sciences à la Renaissance (Art and science in the Renaissance), Ellip (...)
14The dissemination of knowledge first took place through circles of scholars and then via exchanges between academies of science.11 Scientists could debate directly via the accounts of their experiments and they thus became a community12. The raw data were shared and replicated because they were outside any exclusive appropriation. It was customary to say that the mousetrap was patented, but that the data from the experiment were not. The market for scientific data, raw or not, did not yet exist. But in the last few years, following the explosion of Big Data and the emergence of data-based science, the recognition of their value or even their monetisation, and the resulting legal guarantees, have prompted the scientific community to become organised in order to reaffirm the principles of openness and availability of data.
- 13 Now Science at Creative Commons. See also: http://sciencecommons.org/about/
15There are many legal or technical obstacles to the dissemination of data: either the major databases are subject to rights of access, or the data are available in formats that are closed or require proprietary software. For this reason, in 2005 a community of researchers, aware of the resistance encountered during the implementation of open data policies, launched a global initiative to effectively create Science Commons, with tools and methods (access platforms, standard author contracts, etc.) to accelerate the circulation of results and reuse of the data on which they are based.13
- 14 This was the basis of the Science Commons project. See D. Bourcier, “Science et Communication : l’e (...)
16Similarly, what has become known as the “Open Data Protocol” (which rather meant “data sharing”) emerged in the research arena: this also effectively encourages the members of the global scientific community to pool their resources, regardless of their legal status. These shareable research platforms14 facilitate the development of new services for:
reusing research through policies and tools that help individuals and organisations make their work accessible and indicate, on their results and data, this option of reuse;
immediate access to tools (online calculations) through standard contracts that offer economies of scale to researchers, enabling them to duplicate, check and extend their research, and also to take part in the entire value chain through to peer review;
the integration of fragmented sources of information by giving researchers the means to find, analyse and use data from disparate sources by marking and integrating the information through a common, standardised language that can be translated into machine-compatible form.
- 15 See the report by Serge Bauin, L’Open accès à moyen terme : une feuille de route pour HAL (Open acc (...)
17This movement to open and share data has been facilitated by open archive policies developed within scientific institutions (ArXiv, 1991). In France, HAL (Hyper Article onLine), was created in 2000 based on “the model of direct communication between researchers”15 of their pre-prints: its management and missions, currently being overhauled, are still to be defined with regard to the archiving of scientific data, in order to integrate in the design the relationship with the embargo period and the open licence, which must be decided by the researcher alone at the time of filing.
- 16 Nature 498/16–17 (5 June 2013), doi: 10.1038/498017a; the initiative http://genomicsandhealth.org/ (...)
18In 2013 an initiative from the field of human biology launched the “Global Alliance for Genomics and Health”.16 This is a unified movement of member federations from 170 countries which decided to provide a platform for engagement for non-governmental organisations and create a powerful network designed to describe non-communicable diseases. Other disciplines such as physics and Earth and space sciences insist on other imperatives. Thus, the long-term observation of natural phenomena brings into play processes that may have long time constants compared to human life (the variations in the Earth’s magnetic field, tectonics, climate change, the seismic cycle, etc.). They are in essence non-reproducible and underpin our knowledge of the world around us, its changes and the risks it poses to our societies. As nature is a common good, the archiving and free dissemination of these data is a public duty.
- 17 H. D. Evers, Knowledge hubs and knowledge clusters: Designing a knowledge architecture for developm (...)
- 18 The motto of the British Royal Society is nullius in verba (take no man’s word for it).
19The digitisation of observations and the exchange of digital files offered new opportunities to those who wished to reinterpret or compare the data. These various movements have had a definite heuristic effect on the traditional scientific process that used to be described sequentially from design to the written results, whether digital or printed. Now, scientific discourse can no longer be described in a linear fashion but resembles a process where partial results evolve and are interdependent. These cognitive interactions are manifested in what are called knowledge hubs,17 where several layers of more or less developed knowledge coexist. In this framework, the access to the primary data becomes the determining factor, by making it possible to check their quality and gauge the methodology and resulting interpretation. In addition, the open access structure of knowledge has an influence on research itself, which is no longer an “independent variable”18 of the development process, but the dynamic result of continuous brainstorming between researchers (email science).
- 19 Toronto International Data Release Workshop Authors, “Prepublication data sharing”, Nature 461, 200 (...)
20As a result of this movement, publishers have become used to asking researchers, in addition to their scientific results, to place their data online19 in order to verify the reproducibility of the experiment or the process. This has enabled them to first check the published results by comparing them with the data and therefore try to avoid plagiarism and fraud, which obliges them to retract articles. In the longer term, the possibility of these accumulated data eventually creating a “data market” for the benefit of publishers cannot be ruled out. This very disturbing phenomenon should be denounced and refused by researchers.
21Scientific data sharing refers to a traditional community practice in science: making the data used for scientific research available to other researchers and creating pools of resources managed by the scientific community. Many agencies, institutions and journals have supported data-sharing policies because openness and transparency were regarded as ethical principles inherent in scientific work.
22However, making this practice of sharing universal raises questions such as: To whom must these data be open? To which scientific community? Should they be open to citizens, and to the public?
- 20 www.creativecommons.fr, which is a project for sharing content and a platform for open licences, wh (...)
- 21 The status of data banks in Europe, defined as sui generis, cannot be transposed in most other coun (...)
23Sharing policies require researchers to be informed about the limits of this sharing. The data concerned may be unavailable because of their nature as non-anonymised personal data, or may be subject to special regimes such as that of national security and professional secrecy, or to restrictive contractual clauses or various commercial interests. Then, if researchers own the rights to these data and wish to share them, they are advised to place their protectable data under a free licence such as Creative Commons20 in order to at least inform future users that most of the works are protected but have been “freed” under the author’s conditions. Researchers need to be vigilant when they transfer their rights of exclusivity on their data or data banks to third parties,21 or vice versa when they use data generated by other researchers or by open government platforms.
- 22 Directive 2003/98/EC of 17 November 2003 on the reuse of public sector information.
- 23 Directive 2007/2/EC of 14 March 2007 establishing an Infrastructure for Spatial Information in the (...)
- 24 Established by the Circular of 17 September 2013.
24Unlike the data sharing movement developed by the researchers themselves, public open data policies emerged outside the scientific community. Every day, a growing number of data are produced or collected by different actors operating in different business sectors, which differ according to their objectives and purposes. The state, first of all, is a major producer of data. Increasing quantities of statistical data are being produced, reproduced, collected, disseminated or re-disseminated by the public authorities in the framework of their institutional missions. These are mainly demographic, geographical, weather-related, economic, financial, cultural, tourist-related data, etc., which are designed to ensure the quality and continuity of public service – but which can also constitute new resources for researchers. Thus in Europe, following the Directive on the re-use of public sector information22 and then the Directive establishing an infrastructure for geographical information,23 most countries adopted policies to promote the opening of public data. In the public sector, therefore, open data could be defined as the open and (almost) free provision of data, implying the option of reusing them with as few constraints as possible. In France, the mission of Etalab,24 the service that manages public open data under the authority of the Prime Minister, is to communicate research data. Ideally, besides its network of experts it should include the ethical skills needed to define data of interest to research.
25However, the policies that promote the opening of public data, i.e. promoting sharing and reuse, do not have the same objectives and the same targets as data sharing. One of the objectives of public sector open data is to enable the exploitation or even the monetisation of these data by creating wealth for the companies that exploit them. In addition, the targets of this opening concern all the actors in the public sector, authorities and communities receiving public money. However, these policies all share the desire to promote the transparency of knowledge production methods and create deposits of data that are accessible and shareable.
26The fundamental difference between data sharing and open data is that in the scientific world the movement emerged from the community itself and its purpose is ethical because it concerns values, i.e. defining the limits of what is good or bad for the community that applies it. In contrast, for government open data, the incentive was originally normative and applies legally to all public officials, including those working in public research.
27One way to clarify the applicable regimes would be to differentiate scientific data and public data. But scientific data, most of which are produced with public funds, are all intended to become public, with just a few exceptions. And yet we find that not all researchers, depending on the disciplines, are favourable to the opening up of their data. In the human and social sciences, the embargo period from six months to one year, depending on time to publication, may be necessary for the primary data to be made available. This requirement varies in the exact sciences and in fact depends greatly on the fields: in many cases there is no embargo on the data. For example, in biology, data are generally provided at the same time as the publication. Regarding the exploitation of data from major instruments in physics or astronomy, there is a delay before the whole community can benefit, because the raw data must generally be processed before they can be exploited; furthermore, in astronomy, very specific rules, laid down in advance, give a preference for a limited time (generally one year) for the exploitation of data by researchers having built an instrument involving large equipment (satellite, telescope). For researchers in these disciplines, this means that general policies on opening public data are not compatible with the customs of the community to which they belong, and they can contribute to them only by defining the limits of their practices.
- 25 Overview of the health insurance information system. See the proposal for opening and sharing these (...)
- 26 This implies that the researcher must also inform the patient.
28On the other hand, researchers must benefit from public open data promoted by the state. Public data, especially in health, are destined to become scientific data that the researchers can use. Thus the SNIIRAM25 is defined as the world’s largest database on health: for decades it has been populated by the information generated from the delivery of all healthcare and hospitalisations in France. These data are by definition sensitive: they have therefore undergone anonymisation procedures. However everyone – and particularly any researcher – knows that these procedures are not 100% reliable.26 Responding to this request to open this deposit of (sensitive) data, which the report’s authors call “common assets from research in public health”, the French National Health Insurance Fund (CNAM) therefore provided a randomised sample of beneficiaries (one file out of 100) and not all the data, for the researchers and public bodies responsible for public health. Among the sometimes contradictory principles it wishes to apply, the CNAM decided to open access to batches of anonymous data by distinguishing the publication (free) and customised extractions or trend charts (payable). It thus intends to develop its policy of openness, whose criteria will be public interest (mainly research), the quality of the protocol, the need to access the data, the security of the procedures and the status of the applicant. It is therefore possible to understand how public data, even sensitive, can be useful for researchers and be the subject of specific negotiations with this community, while at the same time respecting the rights of the subjects concerned.
- 27 M. J. Khoury & J. P. A. Ioannidis, “Big Data meets Public Health”, Science, 26 November 2014, 346/6 (...)
29Ensuring the quality of the data and validating its processing in light of evidence-based scientific methods also represent another major challenge. Big Data – which refers to massive volumes of information that are complex and likely to be connected – can improve our understanding and prediction (by machine learning) of behaviours likely to affect health, and accelerate the cycle of knowledge dissemination. However “Big Error” can threaten “Big Data”.27 In this article, the authors ask for the systematic replication of epidemiological results and collaborative studies on a broad scale to test predictive tools and move from correlations obtained to real causalities.
- 28 Around 20 hearings were organised at the CNRS.
30Researchers who use personal data – whether the person has been identified or is identifiable indirectly by profiling or targeting – are first confronted with heavy legal constraints. Yet in some cases the data-processing model from which the privacy protection policy had been developed in the French Data Protection Act of 6 January 1978 may appear to be too restrictive in the new contexts of big data in relation to research objectives, or even obsolete as reported by the researchers questioned.28 Indeed, the rapid and open circulation of data between researchers disrupts the order of procedures and makes data flows relatively autonomous in relation to their sources or authors. It often becomes impossible to adhere to the principle of purpose (assumptions are not developed a priori), the principle of proportionality (it is not possible to know which data will be necessary before they are actually used) and the principle of non-conservation (the data are not destroyed at the conclusion of the research because of their open access and reuse). It should be noted, however, that all archived data are subject to an exception for research purposes, once the original research deadlines have passed.
31There is the same need for resources in many disciplines. Take for example computer vision systems, whose purpose is to automatically recognise visual scenes. Facial recognition is one area of computer vision, which can be used in biometrics applications. The problem is then made more complex by rights of personal portrayal, as the identity of a person can be determined from their face. The case was raised recently in the framework of the organisation of a campaign to assess facial-recognition systems. After lengthy negotiations, the French Data Protection Authority (CNIL) gave its agreement on condition that these data were not kept beyond the duration of the project, unless an application for extension was submitted requiring a new case to be made, and the agreement of the CNIL. This therefore results in the paradox of prohibiting experiments by other systems on the same data in order to compare their performance, even though this is a normal scientific approach. The same paradox exists in the request made to Google by the G29, combining the CNIL’s various European counterparts, to not keep for longer than three months the Google Street View images used in developing the algorithm for automatically “blurring” faces. It seems paradoxical to limit the use of these images, when the aim is to enable the development of the most effective face-blurring algorithms possible in operational mode. It therefore seems that this case is confusing the needs of research and the constraints of operational use, which are of a different nature. It would be useful to consider the introduction in French law of the concept of fair use found in common law in the field of copyright. It corresponds to reasonable or acceptable use. Transposed to the area of research, Parliament or the courts should then be asked to define a set of legal rules, which would try to take into account the concerns of both research and public interest, in authorising certain uses that would otherwise be considered illegal. Such a law enshrining fair research use would facilitate the development of research requiring the use of protected data under certain conditions.
32More specifically, it is sometimes difficult to apply the basic principles of personal data processing, such as informing people about the fate and use of the data, or obtaining their consent. Thus, for example, the researcher’s approach may require obtaining information without the knowledge of the person being investigated. It would be necessary to stipulate the principles to be observed in the absence of consent, such as a commitment to inform this person a posteriori. In other disciplines, the data belong to non-identifying datasets, but if they are combined this can lead to re-identifications that require procedures for the possible change of data “status”: identifying or non-identifying. Similarly, the protection of “anonymity” imposes other types of guarantee (commitment by institutions and researchers not to use the identifying characteristic of the data, in the event of re-identification). In any case, the researcher must inform the subjects of the impossibility of guaranteeing the strict anonymity of the data, and give them an assurance that all efforts will be made to ensure that measures are taken to protect their rights.
33Lastly, the issue of the contribution of each party requires the unique, unambiguous and persistent identification of the researchers, which, by giving them credit for these contributions, will thereby also indicate their responsibilities (see the ORCID initiative, for example).
34In the context of the rapidly accelerating circulation of data encouraged by their supervisory authorities and by their community, researchers must:
abide by the international undertakings of the institutions to which they report;
participate in the definition of ethical principles specific to their discipline in data sharing and Big Data in general.
- 30 This PSI Directive is transposed by an Order supplemented by a Decree of 30/12/2005 pursuant to the (...)
35Data sharing had been launched by communities of researchers and fell within the scope of soft law, i.e. non-binding rules of conduct. Today, the institutional commitments to which the public researcher is subject have become binding since the application of the aforementioned open data policies.30 The implications of these policies regarding the ethical dimensions of research need to be assessed on a case-by-case basis and the opening of the data should be applied reasonably.
36As a general rule, public researchers must pursue an ideal of sharing and exchange between peers and take part in the dissemination of data obtained with public funds while respecting any exceptions of a contractual nature to which they may be committed. Conversely, standard consortium agreements involving public and private partners (in particular in competitiveness clusters) are often too restrictive with respect to the opening of data: they should now be negotiated in advance by public researchers in a way that does not lead to the confiscation of unexploited data by private partners at the end of the project.
37Although subject to the principles of sharing and openness, the data are not free: whether or not they are structured in the data bank, they possess a market (economic) or non-market (ethical) value. Like published material, they are increasingly concerned by copyright. It is therefore necessary for the producer to explicitly define the restrictions or exceptions to researcher reusers. In addition, when they are sensitive, the data must strictly adhere to personal data protection policies throughout their processing, to avoid causing problems for subsequent uses.
38Researchers have discovered that the opening of data – but also the software, ontologies and metadata that enable them to be exploited – implied a new responsibility: to take particular care over the quality of the information and data offered as well as the clarity of the accompanying documentation. To enable others to replicate or to reuse data, it is necessary to check the integrity and interoperability of the data, the identification of their sources, the dates they were collected or processed, and to conduct a detailed examination of the different steps leading to the constitution of the data deposits: collection, classification, standardisation, provision, reuse, storage, destruction. The issues relating to image rights, confidentiality and security also raise legal and ethical issues which, although they existed before data sharing, have become more difficult to interpret at a time of generalised international sharing of research results.
- 31 “Ensuring the integrity, accessibility and stewardship of research data in the digital age”, Report (...)
- 32 M. Vito, “Partageons nos données” (Let’s share our data), Le Monde, 28 May 2014.
39Therefore, the organisation, maintenance and accessibility of high-quality interoperable data become fundamental for ensuring the integrity of scientific data in the digital age and creating new legal and ethical responsibilities between researchers.31 Who owns the data? The laboratory? The researchers? The agencies? Are all the necessary means (software, algorithms) available to use and reproduce them? The change of scale also necessitates real international data infrastructures, which further complicates their governance. Thus the rights concerning data and data banks are not homogenous, or even harmonised at European level. There is therefore a shift in the centres of gravity of scientific activity, which calls for continued deliberations, not only in terms of strategy but also in terms of ethics. As was rightly noted in an article on this subject in 2014, “the current trend towards the commercial exploitation of scientific results, with the emphasis on intellectual property, goes in the opposite direction to that of data sharing”.32 The intention of this opinion is to sound the alert against these practical contradictions facing the world of public research.
The COMETS recalls that the CNRS is a signatory to the Berlin Declaration (2003), like most major international research organisations. This commits researchers to the global movement of open data sharing. The COMETS invites all CNRS researchers to join this movement in accordance with the practices specific to each discipline.
The contribution to the work of data sharing must be recognised in assessments and decisions concerning the promotion of researchers. To facilitate this recognition, the COMETS recommends that appropriate indicators be created and that a section on these activities be added in the activity report and the annual activity sheet of researchers.
Researchers and personnel from the world of research must be trained in the ethical management of data (what is known as “privacy, accuracy, property, accessibility”) and informed about the rules of good practice, as well as the legal rules concerning responsible sharing of data, including the fair and proportionate collection of personal data or data likely to re-identify individuals.
Data-sharing practices should be encouraged in the publication policies of scientific journals and in the organisation of symposia, with regard to both authors and evaluators. The COMETS recommends that authors refuse to enable their data to be subject to special pricing by scientific publishers and/or separate subsequent exploitation by the latter (resale or paywall).
The COMETS advocates that the HAL open archive be preferred for depositing the data on which publications of research results rely and that the researcher be able to choose, by open licences such as Creative Commons, the conditions of their reuse.
It recommends that the CNRS should ensure the existence of sustainable infrastructures enabling management of the data platforms in the long term at the team, laboratory or network level. It suggests that the CNRS encourage its researchers to participate in the establishment and activity of international bodies to process metadata using unique, lasting identifiers for these data.
The costs of sharing data, assistance with the creation and maintenance of data warehouses or databases, and the construction and maintenance of multi-use platforms or open archives must be taken into account when the organisation is allocating the appropriate resources (grants, subsidies, etc.) to teams, without prejudice to any pricing of on-demand and customised data processing.
The COMETS recommends that a discussion be held with the French Data Protection Authority (CNIL) and the data protection representative at the CNRS, as well as with Etalab, in order to take account of the specificity of the data and their processing in the world of research. It suggests the creation of an advisory committee for administration of scientific data, involving various disciplines in this debate.
Finally, it stresses the importance for the scientific communities of identifying in an open, collaborative way the legal obstacles to the ethical sharing of data (intellectual property data and sui generis status of data banks), in order to promote real scientific commons, to integrate the concept of fair research use and participate in the adaptation of data rights to the legitimate interests of research.
1 See the article “Dix laboratoires mondiaux partageront données et chercheurs” (Ten world laboratories will share data and researchers) (Le Monde, 4 February 2014). This project orchestrated by the NIH in particular asks public and private laboratories to “not develop their own drugs from discoveries obtained before they have been made public”.
2 Incidentally, we essentially consider scientific data in this text, which excludes from our analysis all types of data tracing individual activities that pose ethical problems of a different nature.
5 P. Uhlir, “Revolution and evolution in scientific communication: Moving from restricted dissemination of publicly funded knowledge to open knowledge environments”, 2008, http://www.communia-project.eu/node/278
8 Nature 498/16–17 (5 June 2013), doi: 10.1038/498017a; the initiative http://genomicsandhealth.org/ is in the process of drawing up an International Code of Conduct for Genomic and Health-Related Data Sharing, currently available for comment on its website (http://genomicsandhealth.org/our-work/work-products/international-code-conduct-genomic-and-health-related-data-sharing-draft-6).
10 In early 2014, five articles were published in The Lancet on the theme “Research: Increasing value, reducing waste”. See in particular An-Wen Chan, Fujian Song, Andrew Vickers, Tom Jefferson, Kay Dickersin, Peter C Gøtzsche, Harlan M. Krumholz, Davina Ghersi, H. Bart van der Worp, “Increasing value and reducing waste: Addressing inaccessible research”, in The Lancet 383, pp. 257–266.
11 “The Academy of Science owes its origins both to circles of scholars, who at the beginning of the 17th century would gather around a patron or scholarly figure, and permanent scientific societies that formed at the same time, such as the Academia dei Lincei in Rome (1603), the Royal Society in London (1645), etc. Through its work and publications, the Academy makes an essential contribution to expanding scientific activity.”
12 Evelyne Barbin (ed.), Arts et sciences à la Renaissance (Art and science in the Renaissance), Ellipses, 2007.
14 This was the basis of the Science Commons project. See D. Bourcier, “Science et Communication : l’exemple de Science Commons” (Science and communication: The example of Science Commons), Hermès 57, 2010, pp. 53–160.
15 See the report by Serge Bauin, L’Open accès à moyen terme : une feuille de route pour HAL (Open access in the medium term: A roadmap for HAL), DIST, CNRS, September 2014.
16 Nature 498/16–17 (5 June 2013), doi: 10.1038/498017a; the initiative http://genomicsandhealth.org/ is in the process of drawing up an International Code of Conduct for Genomic and Health-Related Data Sharing, currently available for comment on its website (http://genomicsandhealth.org/our-work/work-products/international-code-conduct-genomic-and-health-related-data-sharing-draft-6).
18 The motto of the British Royal Society is nullius in verba (take no man’s word for it).
19 Toronto International Data Release Workshop Authors, “Prepublication data sharing”, Nature 461, 2009, pp. 168–170.
20 www.creativecommons.fr, which is a project for sharing content and a platform for open licences, whose options extend from the most open licence (mention of the granting of rights) to the most “commercial”.
21 The status of data banks in Europe, defined as sui generis, cannot be transposed in most other countries. The United States, for example, does not recognise copyright over data banks.
22 Directive 2003/98/EC of 17 November 2003 on the reuse of public sector information.
23 Directive 2007/2/EC of 14 March 2007 establishing an Infrastructure for Spatial Information in the European Community (the INSPIRE Directive).
24 Established by the Circular of 17 September 2013.
25 Overview of the health insurance information system. See the proposal for opening and sharing these “public data”: Rapport sur la gouvernance et l’utilisation des données de santé (Report on the governance and use of health data) by Louis Bras and André Loth, September 2013.
26 This implies that the researcher must also inform the patient.
27 M. J. Khoury & J. P. A. Ioannidis, “Big Data meets Public Health”, Science, 26 November 2014, 346/6213, pp. 1054–1055.
28 Around 20 hearings were organised at the CNRS.
29 “Sharing Publication-Related Data and Materials: Responsibilities of Authorship in the Life Sciences”, Committee on Responsibilities of Authorship in the Biological Sciences, National Research Council, National Academy of Sciences.
30 This PSI Directive is transposed by an Order supplemented by a Decree of 30/12/2005 pursuant to the CADA Act (Commission on Access to Administrative Data) of 1978.
31 “Ensuring the integrity, accessibility and stewardship of research data in the digital age”, Report of the Committee of Science, Engineering and Public Policy, Washington, the National Academies Press, 2009.
32 M. Vito, “Partageons nos données” (Let’s share our data), Le Monde, 28 May 2014.
© OpenEdition Press, 2016