Two tribes or more? The historical emergence of discourse coalitions of responsible research and innovation (rri) and Responsible Research and Innovation (RRI)

Figures & data

Table 1. Procedural steps followed to gather and analyse the rri corpus.

Step	Justification	Description of the process	Implications
(1) Deciding the search terms: step 1/2	We aimed to gather a corpus of academic references to analyse them with a scientometric approach. The Web of Science and Scopus were the best databases available in terms of data richness and quality. We interrogated both databases with a series of keywords.	We searched for references with the following query: « ‘responsible research’ OR ‘responsible innovation’ OR (‘rri’ AND responsib*) ».	The rri corpus contains scientific references only. It also contains inherent time-lags. It has a cut-off date of March 2014, and therefore captures only the early history of rri as an emergent discursive space.
(2) Selecting the database	We evaluated the two databases to choose which one was the best to work with. We did not want to merge the two to have consistency across the data.	We compared the number of results obtained for both databases. The Web of Science reported 124 references while Scopus had 207. The overlap was 40% and documents specific to Scopus represented 47% of the total references. We chose to keep the Scopus corpus.	Only 29 documents that were specific to the Web of Science were not included in the corpus.
(3) Corpus enlargement	We sought to enlarge this initial corpus to get a picture of the scientific landscape of responsibility in research and innovation. Numerous procedures have been developed to enlarge an initial corpus. They mostly consist either in (1) broadening the list of keywords used to search on the database, or (2) following the cited references and citing references of the initial corpus, (3) mixing both approaches. We chose to increase our initial corpus through a step-by-step introduction of new terms in our initial search expression.	We extracted each term related to the theme of responsibility in research and innovation in the initial Scopus corpus. We gathered the terms in five groups: (1) terms close to the term ‘responsible research and innovation’, (2) ethics, (3) research integrity, (4) (innovation and technology) impact, (5) (innovation and technology) risks.	The process was iterative and therefore influences the analysis.
(4) Deciding the search terms: step 2/2	We selected some of the terms previously found to enlarge our list of search terms and perform a new search on Scopus.	For reading purposes, we chose to include in our final query each and every term that gathered between 100 and 2000 references in step (3). See Table 3 for the details of the final query.	Each of the five previous groups of words identified in step (3) was represented.
(5) Co-citation mapping	This type of analysis is instrumental for positioning ‘responsible innovation’ in the wider more diverse semantic landscape of scientific responsibility and governance of new and emergent technologies. By systematically measuring distances between citations shared within an article, it detects clusters and identifies their historical origin.	We created a map based on the co-citation occurrences of the corpus gathered in step (4)(See Figure 1).	The co-citation analysis provides a visualisation of references cited by others, and captures only the most cited references. It, therefore, favours older references where a citation trail has accumulated. Moreover, a threshold has to be set regarding the number of citations displayed. For readability purposes, the map is limited to the citations that appear at least 7 times in the corpus.
(6) Co-word mapping	This approach is similar to the previous one except that the map produced is not based on the co-citations but on the main terms present in the Title, Keyword and Abstract fields. It is not historical.	We created a map based on the co-word occurrences of the main terms of the corpus built in step (4)(See Figure 2).	A threshold has to be set on the number of terms displayed. For readability purposes, the map is limited to the terms that appear at least 10 times in the corpus.

Table 2. Procedural steps followed to gather and analyse the RRI corpus.

Step	Justification	Description of the process	Implications
(1) Selecting the database	Google Scholar is the best available source for ‘cross-over’ literature: policy briefings, discussion papers, and other ‘grey literature’. Google Scholar is also more likely to reveal new and recent academic texts. Access to new texts was an important criterion for the study of a new and fast-emerging field like RRI.	We chose to work with Google Scholar.
(2) Search on Google Scholar	We searched for literature exclusively dedicated to RRI.	We searched for the exact expression ‘responsible research and innovation’. It brought 548 references.	The RRI corpus contains inherent time-lags. It has a cut-off date of April 2014, and therefore captures only the early history of rri as an emergent discursive space.
(3) Filtering the corpus: step 1/2	A decision was made to exclude the references that were considered off-topic and the duplicated references.	All the references without ‘responsible research and innovation’ in their title, abstract or keywords were excluded. Most of these references were references that were found next to a RRI article on publishers’ website and therefore considered by Google Scholar as related to the topic.	The corpus was narrowed to 107 references (20%).
(4) Searching for full text	The list of references was not enough to perform content analysis so we searched for their full text.	We gathered the PDFs through online research and emails to authors.	The corpus was narrowed to 97 documents. Missing documents were mostly on-going thesis or conference work.
(4) Analysing authors’ profiles	We searched for demographical data on the authors: city of employment, whether they work at the European Commission (EC), are academics, participate in projects funded by the EC, have functions related to national or European science policy.	The research was done online through their personal webpages and social media accounts. We considered authorship in a broad sense since acknowledged contributors were also taken into account.
(5) Graph of citations	We aimed to understand which were the key authors cited within the corpus to define RRI.	We coded the authors quoted to provide a definition of RRI in the documents. The network of authors quoted was then graphically represented (see Figure 3).	79 documents out of 97 provided a quotation when they first use the term RRI.
(6) Filtering the corpus: step 2/2	Most of the remaining texts mentioned the term as an element of context or introduction without engaging with it, e.g. developing or extending the concept. We filtered them so as to keep only the articles that elaborate on RRI.	We classified the documents according to the amount of content dedicated to RRI. We created three categories: discussed, used and mentioned. Discussed means that the reference contains reflections about the concept itself; used means that it is discussed but does not constitute the centre of the article, mentioned means that it is a contextual element or that it is mentioned without further elaboration. We kept only the documents that elaborate on the concept itself (discussed).	The corpus was narrowed to 27 documents.
(7) Coding procedure to identify the discursive space of RRI	We defined three relevant dimensions of the discursive space through an abductive methodology that rested on several iterations between analysis of the corpus and formulation of dimensions stemming from the literature on RRI.	We started the coding process with dimensions extracted from the literature (Owen, Macnaghten, and Stilgoe 2012; Reber and Pellé 2013; de Saille 2013; Randles et al. 2016). We then reworked the categories while coding, through an iterative process. The two first final dimensions are related to the classical dichotomy process vs. product (governance vs. goals): (1) Types of governance of research and innovation (process): self-governance of research and innovators vs. new regulatory State; (2) Major goals of RRI (product): need of a paradigm change to address grand challenges vs. foster acceptance of new technologies. The third dimension is related to the meaning given to ‘responsibility’: (3) Meaning of responsibility: retrospective account (accountability, liability) vs. future oriented responsibility.

Table 3. Characteristics of the two corpuses.

Corpus	rri corpus	RRI corpus
Source	Scopus	Google Scholar
Query: Key Terms (cf. Appendix A)	‘responsible research’ OR ‘responsible innovation’ OR (‘RRI’ AND responsib*) OR ‘responsible development’ OR ‘ethics in research’ OR ((‘ELSA’ OR ‘ELSI’) and ‘ethic*’) OR ‘responsible conduct of research’ OR (RCR AND responsib*) OR ‘research integrity’ OR ‘scientific integrity’ OR ‘scientific misconduct’ OR ‘research misconduct’ OR ‘broad* impact*’ OR ‘technolog* risk	‘responsible research and innovation’
Time span	up to March 13, 2014	up to April 15, 2014
Retrieved references	4,585	548
Relevant references	4,585	27
Analysis	Description of socio-semantic networks	Coding of discursive dimensions
Aim	Identifying the intellectual roots & discursive space of rri	Identifying the discursive space of RRI
Software	Cortext manager	Nvivo

Figure 1. Three main strands of research deal with the notions of research and innovation responsibilities. Distinct discourses deal with (i) scientific (mal)practice, responsible conduct of research and ethics (ii) management of risk and risk perception and (iii) and the sociology of science and technology studies, governance of emerging technologies and responsible innovation.

Notes: Co-citation network of the references that are used at least seven times in the corpus (n = 179). Measure: distributional. Edge filtering: threshold = 0.24. Clustering: Louvain. Chi2 measure of top three associated author’s keywords. Node size set according to the node's number of links. Citations displayed according to a timeline (older citations on the left).

Figure 2. Two main topics and seven subtopics describe the academic discourse on scientific responsibility: innovation governance and ethics/deontology.

Notes: Co-occurrence of Abstract and Title terms used at least ten times in the corpus (n = 276). Measure distributional. Edge filtering: threshold = 0.31. Louvain clustering. Node size set according to the node's number of links. Color circles size set according to the number of links.

Figure 3. A graph showing which authors are cited by whom when it comes to defining RRI (data from the RRI corpus). The size of the nodes is proportional to the number of citing authors.

Owen, Richard, Phil Macnaghten, and Jack Stilgoe. 2012. “Responsible Research and Innovation: From Science in Society to Science for Society, with Society.” Science and Public Policy 39 (6): 751–760. doi:10.1093/scipol/scs093.

 Web of Science ®Google Scholar

Reber, Bernard, and Sophie Pellé. 2013. “Innovation Responsable.” In Dictionnaire Critique et Interdisciplinaire de La Participation, edited by I. Casillo, R. Barbier, L. Blondiaux, F. Chateauraynaud, J.-M. Fourniau, R. Lefebvre, C. Neveu, and D. Salles. Paris: GIS Démocratie et Participation. http://www.dicopart.fr/it/dico/innovation-responsable.

 Google Scholar

Saille, Stevienna de. 2013. “Innovating Innovation: Rri as a Guiding Principle in the ERA.” SSRN Electronic Journal, doi:10.2139/ssrn.2401076.

 Google Scholar

Randles, Sally, Philippe Laredo, Allison Loconto, Bart Walhout, and Ralf Lindner. 2016. “Chapter 3. Framings and Frameworks : Six Grand Narratives of de-Facto Rri.” In Navigating Towards Shared Responsibility in Research and Innovation, edited by Ralf Lindner, Stefan Kuhlmann, Sally Randles, Bjorn Bedstead, Guido Gorgoni, Erich Greisler, Allison Loconto, Bart Walhout, and Niels Mejlgaard, 31–38. Karlsruhe: Fraunhofer Institute for Systems and Innovation Research. http://res-agora.eu/publications/res-agora-book-now-available/.

 Google Scholar