Aix-Marseille University SSH data platforms: Skills to support research in social sciences and humanities (SSH) in the Mediterranean (2024)

1The development of computer science significantly transformed scientific practices in the last quarter of the 20th century. Information technology (IT) first made it possible to develop computing power, then information storage capacities. It initially led to an increase in discoveries in the most quantitative sciences as it allowed accelerated calculations, particularly in fields where the data was quantitative and structured (e.g. tables of data in SSH). It then allowed new possibilities to be developed in more qualitative sciences, thanks to the increase in information storage spaces and the development of qualitative or non-explicitly structured data processing (e.g. literary corpora or images in SSH). Interest has gradually shifted from calculation to data.

2Today, there is substantial reflection on what IT produces in science in general (largely a renewed questioning of the role of data and the place of reproducibility), and in the sciences more specifically (a new questioning of existing corpora, for example). At the heart of contemporary thinking, the question of data, defined as an elementary description of a reality, is placed at the centre of science and can be defined as an ensemble of systematized knowledge relating to the same field. Data consists of elements resulting from structured knowledge.

3In this context, the question of accessing data in order to build knowledge, but also to reproduce reasoning, has become an increasingly important issue in scientific research. The recent development of the global Research Data Alliance network (https://www.rd-alliance.org/), whose objective is the “development of infrastructure and community activities to reduce the social and technical barriers to data sharing and re-use”, is testament to this. In France, the creation of the Open Science Committee (https://www.ouvrirlascience.fr) also represents a milestone in the institutional organisation of science and scientific data (ministère de l’Enseignement supérieur et de la recherche 2018).

4In this article, we first return in greater detail to the question of accessing data, that has notably led a number of researchers to propose specific computer interfaces. We will give two representative examples in part two. Finally, in the last part of our article, we will tackle a key element of the transformation of research practices, something too often forgotten in favour of technology: the human factor.

5In order to remain concise and anchored in practice, we will deliberately restrict our reflection to the field of social sciences in their quantitative dimension.

6First of all, we would like to remind that access to data is based on the producer's ability to become a data disseminator (or to entrust it to a third-party disseminator). As well as practical questions, this poses legitimate legal questions, which we will not address here. The task of a disseminator is to make data available, and today this is most often done online. Dissemination must be distinguished from archiving, because the later aims to preserve data with a perspective of long-term storage, while dissemination is driven by the desire to ensure availability.

7In France, for example, INSEE (National Institute of Statistics and Economic Studies) produces public statistical data, some of which is general, recent and made available to the public via a website (http://insee.fr). Since 1987, files considered of “historical interest” have been sent by INSEE to the National Archives as part of the “Constance” system (Conchon 1993). Accessing them is then much more complicated, because the researcher must go through a series of authorisations in order to justify their request for access. It is in this context that the “Archives des Données issues de la statistique publique” [1] were created. They document and archive INSEE (and also data from others official sources like ministries) in order to disseminate it to the research community

8However, making data available online does not guarantee that it will be used properly and in accordance with scientific standards. The acronym FAIR represents the qualities necessary for data to be truly usable: Findable; Accessible; Interoperable; Reusable (Wilkinson et al. 2016).

9These four main principles relate to different dimensions of data archiving and dissemination.

10To be “Findable”, the data must be indexed on the internet, preferably with a unique and persistent identifier (PID). A persistent identifier is characterised as a unique character string which refers to an object (this may be a physical object, but also a web page or any other document). It is said to be persistent since the identifier will not change even if the designated object is moved. The DOI (http://www.doi.org/) is certainly the most well-known and most widespread of these identifiers, though others are also valid (Cousijn, Kendall et al. 2018). Thus, a DOI for a journal article or database will remain the same, even if the article or database is moved to the file tree structure of a website or to another website.

11The notion of accessibility is understood here in a narrow sense and refers to the transparency of data recovery protocol. “Open data” is, of course, accessible. Data that requires special protection (confidential data, for example) can be considered accessible if the authentication and authorisation procedures are explicit and visible.

12Interoperability is based on the principle that data is associated with elements that describe it unequivocally, but also on the fact that this set of data/metadata is structured and conceptualised using shared, documented standards.

13Finally, to be reusable, the data must be organised in databases using common standards and must be well documented, that is to say with precise and relevant metadata. Metadata consists of sets of elements describing the data made available. These are therefore data notices. For example, the creation date of the data is an important information to know. The metadata can be numerous, with details on all sorts of aspects (who retrieved or constructed the data, where, when, when it was made available, etc.).

14While respecting the FAIR principles allows data to be used by the entire scientific community, a growing number of researchers believe this is not enough (see, for example, Boeckhout, Zielhuis and Bredenoord 2018). It is also advisable to provide support for those using this data. The method of facilitating availability is also an issue, whether it is a matter of the ergonomics of the tools offered or even human support for users when it comes to secondary use of this data. A simple file from a spreadsheet on a website, even perfectly documented, is not always enough for a user to be able to retrieve it. To promote this support, a few simple measures can be implemented as a start: promoting the results in descriptive form (graphs, maps) or even referring to earlier analysis (e.g. published results) can help better understand the possible uses. Sometimes this support must be further developed in order to help a user (even an informed one) by means of concrete help in “getting to grips with” the data: advice, training, personalised support.

15This position, which maintains the need for mediation between data and researchers, is the one we take (Oliveau 2017; Blöss-Widmer 2019). Mediation can be considered in two complementary ways. The first consists of offering data exploration, visualisation and even analysis tools that let a user enter data without having to retrieve it or having specific software tools. This is the case for the many online mapping interfaces (often based on geographic information systems), but also for an increasing number of data visualisation modules (there are numerous examples, but we may cite the Eurostat Data Browser and its tabular, graphic and cartographic interface for thematic explorations).

16The other aspect of mediation is human support. This is based on the involvement of a more experienced user in methodological or technical terms. In the French research system this concerns, for example, the role usually assigned to some of the research assistant found in research laboratories.

17To illustrate our point, we will present two data access tools with which the authors have links. These are good examples to show how the FAIR principles can apply to data dissemination, and how they do or do not include additional elements of accessibility.

18The first example is a national research infrastructure that dates back to a reflection at the end of the 20th century. This infrastructure is based on original digital tools and innovative institutional organisation which aim to make access to data effective. To understand the current situation, we need to look back at the way this infrastructure has developed over the last 20 years.

19In 1999, the “Social Sciences and their Data” report submitted to the Minister of National Education and Technology proposed the creation of an institute for disseminating data in the social sciences (Silberman 1999). In 2001, the Consultative Committee for Data in Human and Social Sciences (CCDSHS) was created by decree [2]. The mission of the CDDSHS is to define a data policy for the social sciences. On 1 July 2001, the CNRS (National Scientific Research Centre) in partnership with the EHESS (School of Advanced Studies in Social Sciences), the INED (National Institute for Demographic Studies) and the University of Caen, created the “Quetelet Centre” [3]. In 2005, the centre participated in the creation of the Quetelet network (in the form of a Scientific Interest Group). The network includes the INED Survey service, the Sciences Po Sociopolitical Data Centre (CDSP [4]) and the ADISP team [5] (Caporali; Morisset, Legleye et al. 2015).

20In the 2010s, the tools were gradually simplified. The roadmap for French research infrastructures (ministère de l’Enseignement supérieur et de la recherche, 2008) [6] provided for the existence of a Research Infrastructure called PROGEDO [7]. It was materialised in 2012 with the creation of the “Quetelet PROGEDO” Mixed Unit of Service which took over the activity of the Quetelet network. In 2017, the CNRS decided to assign the ADISP team to Quetelet-PROGEDO. In 2018, Quetelet-PROGEDO was transformed into a Service and Research Unit, called simply “PROGEDO”. It fulfils the role formerly carried out by the CCDSHS, namely that of defining a data policy for the humanities and social sciences, as well as the mission of the Quetelet network (data dissemination) through the “Quetelet-PROGEDO-Diffusion” internet portal. It is also responsible for archiving quantitative data via the ADISP team [8].

21Today, the “Quetelet-PROGEDO-Diffusion” portal (http://quetelet.progedo.fr/) is firmly positioned in the FAIR data perspective, providing access to data according to its level of sensitivity (see below) and documenting it up to the level of questionnaire variables when possible. The portal makes it easy to find data even from a large variety of sources (INED, Sciences Po Paris, INSEE, Ministries, Universities, etc.) [9]. These data are made accessible thanks to protection which varies according to the type of data: open access for data that does not concern individuals; data with reserved access for so-called “production-research” files that do not directly identify individuals but are considered sensitive; data with secure access for individual data. The DDI documentation protocol is used to make this data reusable and also allows interoperability.

22Respect for the FAIR principles has undoubtedly contributed to the success of this ambitious data dissemination enterprise, as shown by the results of the survey conducted in 2019 [10], with 90% of users declaring themselves satisfied. As we will see, despite positive evaluation, experience shows that it is essential to provide additional support for the least skilled users and to continue efforts to publicise the services offered. This explains why in France, in particular, this tool is not only digital, but is simultaneously based on human resources which can answer to various requests from users. In order to make the data accessible, in the broad sense of the term [11], efforts must be continued in supporting visitors to websites which provide data in order to make the best use of it in their analysis. Two tools we are responsible for have inspired us to illustrate this necessary articulation of digital enhancement tools with personalised support systems using support staff. Namely, the specific interface of the demographic observatory of the Mediterranean, DemoMed, and its tools, designed to simplify the reuse of spatial demographic data for the Mediterranean region, and the Aix-Marseille University data platform.

27However, in practice, our team, comprised of teachers and researchers, found that computer mediation is not always sufficient to stimulate the reuse of data. In fact, while digital technology makes it possible to reach a large audience of potential users, it has greater difficulty engaging new audiences and supporting less experienced users. With this in mind, many local initiatives were carried out at a number of university sites before being coordinated and deployed on a national scale by the Large Research Infrastructure PROGEDO, through its University Data Platform (UDP) [12]. The Large Research Infrastructure PROGEDO effectively coordinates a network of research assistants and scientific advisor, located at the heart of universities (in Social Sciences and Humanities faculties). These skill platforms all constitute real skill supports for developing a community of quantitative data users. The UDP of Aix-Marseille University, which we belong to, was officially created in 2018 (http://pud.mmsh.univ-aix.fr/) and, like other UDPs, offers training activities as well as specific support for researching quantitative data in SSH in the Mediterranean.

28The history of UDPs dates back to 2001. The idea of locally developing a support unit for researchers first appeared at the University of Lille [13]. This came about in 2003 (Duprez, Cros 2010). Supported by the University of Lille 1 since its creation, the Lille University Data Platform joined the Maison Européenne des Sciences de l’Homme et de la Société (European House for Social and Human Sciences) in 2015. It remained the only initiative of its kind until 2009, when a second platform was created in Lyon [14] on the initiative of the Lyon-Sainte-Etienne House for Human Sciences. Then came the UDPs in Nantes (named PROGEDO Loire) and Caen in 2015. Since 2016 in particular, and under the leadership of the Ministry from 2018 onwards [15], the network of UDPs has expanded to cover all major university sites. At the end of 2020 there were 14 UDPs in France: 12 hosted in Houses for Human Sciences outside of Paris and 2 in Paris University and Sciences Po.

29These UDPs were created due to a strong need felt by local research teams, primarily to support researchers, but also to support students as they begin research work, i.e. in discovering and using the breadth of data available. More generally, there is a significant need for advice and training on the use of survey data or data from large databases. Faced with this situation, TGIR PROGEDO therefore took the initiative to stimulate the development of these UDPs, which form the necessary link between digital platforms and users. The initiative promotes the use of data from French, European and international research and public statistics (survey and administrative data, etc.) (Chenu, Lesnard 2011). This data is mostly quantitative. For example, the research assistant attached to Aix-Marseille University’s UDP provides assistance in searching for data, accessing it, processing and exploiting it. An academic fellow is designated as scientific advisor for the UDP exercises support functions for all researchers, research professors, administrative and technical staff, doctoral students and master's students at Aix-Marseille University.

30The additional aim of the Aix-Marseille UDP is to develop expertise on human and social science data for the Mediterranean region. In addition to their primary mission of personalised research support, the research assistant organises training, seminars, study days and summer schools. He visits various teaching and research institutions to publicise the platform and its missions. He also actively participates in exchanges and associated activities in the national network formed by all UDP. Research assistants recruited by the UDPs have a common profile: they have strong statistical skills and teaching motivations. They also must all have a very good knowledge of quantitative disciplines like demographics, statistics, geography, economics or quantitative sociology. This is to ensure that these new research support staff can share their experience in processing quantitative databases and statistical data processing methods with less experienced applicants. They must also demonstrate disciplinary openness to accommodate somewhat diverse requests for help from various backgrounds.

31The extremely positive feedback from those using the services offered by the UDPs once again demonstrates that while IT has managed to replace humans in many tasks, there is still a need to maintain close human interaction in many situations. Information technology first supported researchers in their calculations, then in managing their data. Free of these constraints to a large extent, SSH researchers are now asking for support in the choice of available data and the methods for making the best use of it. Today, there is too much data and too many methods for a researcher to master them all. We can rejoice over this, since it allows greater possibilities. Nevertheless, it is all the more necessary to get support in this work, and the UDPs fulfil this mission. “The challenge of making statistical data digitally available would therefore be to combine training in data sciences and digital sciences” (Blöss-Widmer 2019: 64).

32In France, the social science research system is reaching maturity and is structured in the form of collective tools, which notably allow better management of the data produced. We are moving away from the individual observations and interpretations that characterised the system and heading towards a more shared organisation, comparable to that found in other sciences. This method of organisation has a cost, but it is essential for developing and maintaining databases.

33One question remains: that of this work's sustainability. The increase in skills due to training by human means is sustainable. It allows those involved in SSH research to be better trained and better supported. This fully justifies investing in university data platforms, as well as in any other human support tool. However, we also know that these investments should not be reduced, at the risk of seeing these gains disappear. With incentives disappearing, researchers could turn away from these objects; technical tools must be maintained at a high level of service so as not to become obsolete; individual trajectories also lead to regular staff turnover.

34On the other hand, data management, such as that of IT sites, has revealed itself to be less sustainable than expected. IT tools do indeed become obsolete, and the methods used require regular adaptation. Data formats are evolving, as are website technologies. It is estimated that the long-term lifespan of a website is around 5 years. Some data created ten years ago can no longer be read by current tools.. In addition, the development of documentation and metadata appears to be an absolute necessity to reuse this data (even for the researcher themselves). All of this is time-consuming and expensive. This raises the issue faced by libraries in the 20^th century: that of selecting elements to preserve and deleting others. The stakes are high because statistical data “is central to research in the humanities and social sciences and for political decision-making” (Blöss-Widmer 2019: 64).