Process, outcomes, virtues: the normative strategies of responsible research and innovation and the challenge of moral pluralism

ABSTRACT

The paper analyses the normative foundations of responsibility in the recent literature on responsible research and innovation (RRI). Firstly, it finds that RRI approaches invoke three main types of moral reasoning to define responsibility corresponding to key philosophical traditions: proceduralism, consequentialism and virtue ethics. The paper analyses how RRI approaches differ in the kind of elements they borrow from these three traditions. Secondly, the paper shows that this analytical reflection about the normative strategies followed by RRI approaches helps to better understand the tension existing between, on the one hand, the modern requirement of pluralism in democratic societies to allow for various value systems to influence the decision-making process and, on the other, the applicability of normative theories that demands specific and practical norms to be identified. The paper aims to show how each type of moral reasoning strategy used by most RRI approaches deals with this tension and how each attempts to resolve it.

KEYWORDS:

• Responsible research and innovation
• responsibility
• proceduralism
• outcome-oriented theory
• care ethics

Introduction

Responsible innovation and research (RRI) is a newcomer in the landscape of the normative assessment of technology. This interdisciplinary concept, which is at the core of the EU framework program for research and innovation Horizon 2020, has penetrated the network of science and society theories to offer new perspectives on the articulation between scientific progress and social, political and ethical legitimacy.

Among the growing literature on RRI, a central question that has not received sufficient attention has to do with what the theoretical foundations are of “responsibility.” While various definitions, conditions and principles of responsibility have been offered by the literature (e.g. Owen, Macnaghten, and Stilgoe 2012; Owen, Bessant, and Heintz 2013; Lee and Petts 2013; Robinson et al. 2013; Stilgoe, Owen, and Macnaghten 2013; Sykes and Macnaghten 2013; Grunwald 2014; Guston 2014; Nordmann 2014; Pavie, Scholten, and Carthy 2014; Schroeder and Ladikas 2015; Ziegler 2015), only a few works concentrate on the meaning of the concept. For instance, von Schomberg (2013) and Muniesa and Lenglet (2013) reflect on what irresponsible research and innovation is; Pellé and Reber (2015, 2016) identify multiple interpretations of the term (including responsibility as a cause, a role, a capacity, as blameworthiness, liability, accountability or as responsiveness and virtue); and Blok and Lemmens (2015) take issue with the applicability of the term to innovation in the first place.

This paper goes beyond an analysis of the meanings of responsibility and instead seeks to unveil the normative foundations that underlie the various conceptions of responsibility in RRI. Doing so will not only help inform scholarly efforts to define responsibility (and other core concepts) in RRI; it is an important step to articulating practical principles of governance, which could be effectively followed by social actors. If approaches of RRI tend to agree on some crucial “ingredients” of responsibility, such as the need to include relevant stakeholders, they are based on markedly different justifications and more fundamentally, on different types of moral reasoning, which as the reader will see correspond to established normative theories. The purpose of the paper is to investigate which normative strategies ground the idea of responsibility in the current RRI literature and to explore possible resolutions to the tensions that emerge when they are considered together.

Of the three normative theories found to have been mostly invoked so far in the literature on RRI, it is perhaps not surprising that these resonate with longstanding philosophical traditions. Little attempt, however, has been made to explicate these foundations. As we shall see, the first type of normative foundation is procedural and focuses on the conditions that processes of research and innovation (R&I) should try to satisfy. It is inspired by a procedural tradition in political philosophy exemplified by the work of John Rawls, Jürgen Habermas, Amy Guttmann and Dennis Thompson, Joshua Cohen and John Dryzek, among others. If these authors significantly differ on which political procedures best ensure the deliberative ideal, they all agree that at least some fundamental (political) decisions must result from a collective process of decision-making based on deliberation and that the quality of the process must be guaranteed by specific conditions about, for instance, the kind of reasons citizens can advance to justify their decisions.

The second type of moral reasoning underpinning RRI theories is outcome oriented and defines some clear objectives that R&I practices should try to achieve. It belongs to a consequentialist tradition in moral philosophy where utilitarianism has played an important role, and that assess the moral goodness of a principle or an act by assessing the moral goodness of the outcomes it leads to. Finally, the third normative strategy of RRI approaches derives responsibility from the dispositions or virtues of R&I actors, institutions and systems, referring to a tradition going from Aristotle and Confucius to contemporaneous authors who contributed to understand ethics through the idea of virtue, like Alasdair McIntyre or Martha Nussbaum, including care ethicists such as Carol Gilligan or Joan Tronto.

These three key philosophical approaches of moral philosophy have a long histories of analytical development and controversy. By analyzing the RRI literature in their light, it is possible to better understand some contemporaneous challenges and subsequent debates raised by the development of science and technology. One such challenge, which is crucial for democratic societies, relates to the tension existing between the modern requirement of pluralism, on the one hand, and the applicability of normative theories, on the other (Bohman 1998). Attempts to resolve this challenge can in turn raise new sets of conceptual and practical difficulties. For instance, on the one side, to ensure value pluralism and to guarantee that no substantive theory of the good is promoted over another, RRI approaches grounded on proceduralism propose to identify the conditions that R&I processes should seek to satisfy in order to achieve some level of responsibility. Yet, how then to ensure that those conditions will effectively lead to responsible outcomes?

On the other side, the need to have a practical framework for RRI would tend to either (a) identify actions to take or goals to achieve, and thus call for a definition of responsibility that is outcome-oriented (which identifies certain types of results that should be achieved), or (b) determine the virtues and dispositions for responsibility that R&I actors should try to adopt and emulate in order to promote flourishing. But then, how to defend desirable outcomes or virtues in a pluralist context – where what is identified as morally desirable often stems from conflicting values? How to be sure that in defending substantive ends or dispositions, an RRI approach is not imposing some values on people and citizens, with which they fundamentally disagree?

The paper aims to show which kind of normative strategies and justifications RRI approaches provide in attempting to answer these challenging questions. It must be noted at the outset that no RRI approach is purely proceduralist, purely consequentialist or solely based on virtue ethics. Instead, most of RRI frameworks combine two or three of the normative strategies I mentioned. Importantly, however, they differ in the type of arguments they use and in the weight they give to one or the other of these normative strategies. Thus, some have a stronger proceduralist overtone, while also giving a key role to the notion of care (e.g. Owen, Macnaghten, and Stilgoe 2012; Stilgoe, Owen, and Macnaghten 2013). Others combine proceduralism with more substantive outcomes that should be achieved (e.g. von Schomberg 2013). Unveiling the different normative components of RRI approaches will help to clarify (1) what is the nature of responsibility that is mobilized; and (2) how RRI approaches deal with the mitigation of multiple value systems.

Section 1 presents the various conditions that have been proposed by the procedural approach of RRI and how the latter deals with the issue of pluralism. Section 2 scrutinizes how theories that have both a procedural and an outcome-oriented dimension relate these two dimensions. Section 3 focuses on RRI approaches that rely on an interpretation of responsibility based on virtue. Finally, section 4 discusses the relevance of these three general strategies concerning the aforementioned tension that arises between the requirement for value pluralism and the practical constraints raised by the application of norms.

As for the methodology, this account of the emerging literature on RRI does not aim to be exhaustive. Rather it seeks to provide an analytical view of this emerging field, through a selection of prominent articles and book chapters identified through a three-step process: Relevant articles were selected through a broad search across widely used portals for academic research (including Google Scholar and EBSCO Publishing), using the search terms “Responsible Research and Innovation” and “Responsible Innovation”¹; with the help of RRI key actors’ suggestions²; and by consulting the references of already selected articles. Empirical research was not included, as the paper focuses on identifying and characterizing the main theoretical conceptions of RRI.

1. The procedural structure of RRI approaches

To ground the idea of responsibility in R&I, the first type of moral reasoning that has been called on is proceduralist. It focuses on the systemic conditions that R&I processes have to satisfy. Authors such as Owen, Macnaghten, and Stilgoe (2012), Stilgoe, Owen, and Macnaghten (2013), Lee and Petts (2013); Sykes and Macnaghten (2013), Grunwald (2014), Guston (2014), Schroeder and Ladikas (2015), Ziegler (2015) defend RRI approaches that show a strong proceduralist overtone. Most of them³ identify several conditions, dimensions, capacities and the like that both apply to individuals and to R&I or institutional processes and that are supposed to help achieve RRI if adhered to. Together, these conditions are meant to form or guide R&I processes and activities such that they (ideally) guarantee or (practically) increase the likelihood of producing both adequate normative decisions and adequate outcomes.

It must be noted that these RRI frameworks and approaches are proceduralist rather than formalist since they do not present purely formal rules (Ceva 2012); instead, they ascribe substantive content to the conditions that the envisioned procedure should follow.⁴ For instance, as we shall see, the condition of responsiveness requires that actors have the ability to adapt to their context; while reflexivity implies that they should be able to engage in the equivalent of a meta-analysis of the presuppositions they hold when they invoke a normative position. Both qualities involve a substantive dimension.

Generally, five conditions have been put forward, although there are sometimes only four (e.g. Stilgoe, Owen, and Macnaghten 2013 and Stahl, Eden, and Jirotka 2013⁵): anticipation, reflexivity, responsiveness, inclusion and transparency.

1.1. Transparency

An oft-cited condition of a responsible innovation and research concerns access to information and echoes the condition of publicity that is often invoked to foster the vitality of democracy. René von Schomberg (2013, 39) advances a vision of RRI as a transparent process by which social actors become “mutually responsive” to each other regarding, among others, the ethical acceptability of R&I. Thus, transparency would imply not only that available information about technology, including its anticipated consequences and uses, but also information that stems from technology assessment, deliberation or any other tools used to engage in ethical and critical reflection on technology, should be made accessible by R&I actors and distributed among stakeholders.⁶

1.2. Anticipation

Furthering a trend initiated by Beck ([1986] 1992), Callon, Lascoumes, and Barthes (2001), Funtowicz and Ravetz (1992), Jasanoff (2003), Gibbons et al. (1994), Nowotny, Scott, and Gibbons (2001), among others, most RRI authors argue in favor of anticipatory activities to manage science and technology development (e.g. Hellström 2003; Grunwald 2010; Sutcliffe 2011; Owen, Macnaghten, and Stilgoe 2012; EC report 2013; Lee and Petts 2013; Owen, Macnaghten, et al. 2013; Stilgoe, Owen, and Macnaghten 2013; Guston 2014). These authors relate anticipation to knowledge under a broad array of contextual conditions. For instance, from the standpoint of risk assessment, RRI authors emphasize anticipation in cases when the specific context of R&I activities generates possible risks (probabilities and effects are both known), uncertainties (only probabilities are known) and also ignorance (neither the nature of the event nor its probability of occurrence are known). In addition to knowledge, anticipation in RRI also applies to actors’ actions and strategies, for instance in “dealing with risks beforehand” (Setiawan and Singh 2015). As such, anticipation in RRI is related to both aspects of the well-known Collingridge dilemma, which is both a dilemma not only of knowledge but also of control.⁷

RRI authors who invoke anticipation differ, however, regarding their confidence in rationalistic approaches. Approaches that are both proceduralist and outcome-oriented often have a strong rationalistic overtone and rely on more traditional tools of risk assessment (Sutcliffe 2011; von Schomberg 2013). In these cases, anticipatory research has to be conducted so as “to design socially robust agendas for risk research and risk management.” (EC report 2013, 59). On the other hand, approaches that are more purely proceduralist tend to treat anticipation as a normative process in itself. For instance, such approaches may consider anticipatory activities such as the building of scenarios as forms of stakeholder engagement and in the process link it to the dimensions of inclusion, reflexivity and responsiveness (see below).

In these cases, anticipatory activities are conceived of less as instruments of objective knowledge (through cost/benefit analysis, for instance) than as tools of normative reflection, whether through scenarios (Rip and te Kulve 2008), narratives (Macnaghten 2010) or dialogue (Blok 2014). Against a purely consequentialist and rationalistic view of knowledge (Jasanoff 2003), tapping the power of imagination through narratives and other means helps different normative systems to express and unfold their complexities (Nordmann and Macnaghten 2010; Grinbaum and Groves 2013; Robinson et al. 2013; Guston 2014; Richter et al. 2016). Furthermore, as it has been claimed in the context of nanotechnology for instance (Selin 2007; Grunwald 2010), anticipatory activities – when they relate to normative conceptions of a “better world” rather than alleged objective anticipations of a future as it will be (Nordmann 2014) – contribute to the elaboration of a common normative horizon. By debating about (often conflicting) ethical interpretations of possible outcomes, rather than merely anticipating those possible outcomes, social actors have an opportunity to adjust their visions to that of others and to achieve a form of co-construction of norms that could apply to certain possible pathways of R&I development. Such a stabilization of plausible scenarios into expectations and a given rhetoric of science helps co-constructing norms but can also have a direct influence on the development of technology (Selin 2007).

1.3. Inclusion

Closely related to the other conditions, one of the most important dimensions that is outlined by RRI approaches focuses on the participatory process by which R&I activities are carried on. Whether it should be interactive as in von Schomberg’s definition; “inclusive” (or “collective”) as in Mitcham (2003), Grunwald (2012, 2014) and Guston (2014); framed as “inclusive deliberation” (Owen, Macnaghten, et al. 2013), “distributive and participative” innovation (Callon and Lacoste 2011) or “participative” (Sykes and Macnaghten 2013; van den Hoven 2014); those terms, if not equivalent, outline the need for involving various actors affected or concerned by innovation, typically at an early stage: end-users, researchers (including social scientists), civil society, citizens, consumers, NGOs, industry, policy makers, capitalists, etc.

Promoted in participative technology assessment and other approaches that have informed RRI, for instance, the ideas of participation and inclusion have histories that pre-date responsible innovation (Jasanoff 2011). Furthermore, by affirming the need to widen the type of relevant social actors taking part in the discussion, RRI theories reenact an old issue, common to most of the theoretical and practical propositions inspired by the participative turn of the 1960s concerning the epistemic role of the people’s voice: does a more democratic management of emerging technologies or innovation achieve a higher normative quality of innovation, science and technology production? Following Daniel Fiorino’s (1990) and Andy Stirling’s (2005, 2008) taxonomy,⁸ three main reasons are advanced in favor of an affirmative answer in the RRI literature.

First, the inclusion of stakeholders into the making of norms regulating scientific practices and innovation processes has a normative value: dialogue would be the right process to enhance democracy, equity or justice since it helps closely linking ethical issues to the motives, interests and the desirable ends of individuals (Ferry 1991; Nordmann and Macnaghten 2010).

Secondly, deliberation and participation have an instrumental value and are claimed to increase the political and social legitimacy of innovation and research, its economic and scientific accuracy and finally, its ethical desirability. By contributing to the elaboration of norms, stakeholders are made co-responsible for the development of R&I in a pluralist way (Owen, Macnaghten, et al. 2013; Sykes and Macnaghten 2013). A collective construction of norms is supposed to prevent from selecting any conception of the good as a priori and that conflicting values and value systems will be dealt with by favoring a process of identification, interpretation, argumentation and reconstruction of norms (following Ferry 1991) so as to identify possible areas of agreement.

Finally – and this would be the third argument – public and stakeholder dialogue would also have a substantive benefit related with the power of transformation opened by participative or deliberative processes that lead to the embodiment of “social knowledge, values and meanings in a substantive manner” (Stirling 2005; Sykes and Macnaghten 2013).

1.4. Responsiveness

Connected with anticipation and inclusion, one of the most interesting conditions put forward as an enabler of RRI is responsiveness (e.g. Pellizzoni 2004; Groves 2009; von Schomberg 2011; Owen, Macnaghten, and Stilgoe 2012; EC report 2013; Owen, Macnaghten, et al. 2013; Fisher and Miracle 2014; Nielsen 2016). It is argued that innovators and other social actors should become mutually responsive so as to increase the flexibility of R&I processes and strengthen the ability to shape research and innovation according to societal values and changing parameters (new scientific knowledge, political contexts, etc.). Not only downstream once the technology is already widely distributed (as it was the case for genetically modified organisms (GMOs), for instance), but continuously during the life cycle of technology from the early first concepts to their wide application and commercialization. The ideas of “improvisation” (Groves 2009), “midstream modulation” (Fisher, Mahajan and Mitcham 2006), “safety by design” (Kelty 2009) and “value sensitive design” (van den Hoven 2013), for instance, are all illustrations of this focus on “an iterative, inclusive and open process of adaptive learning, with dynamic capability” (Owen, Macnaghten, et al. 2013, 9). In the context of sustainable development, the concept of systemic “resilience” – that is, the ability of the system to adapt to unforeseen changes (for instance, Pisano 2012) – also evokes a need for a capacity to quickly adapt and move forward.

In all these approaches, responsiveness ensures the epistemic quality of R&I processes by integrating new parameters (conflicting voices, new scientific knowledge, etc.) into the decision-making process. Yet, it is not always clear in the literature whether this capacity applies to the whole process, to individuals or to both.

In an individualist interpretation, responsiveness is conceived of as a set of individual virtues: social actors (innovators, researchers, policy makers, Civil Society Organizations, hereafter CSOs) should be able to listen to the others, understand the issues at stake and answer them in the best way, as suggested by the Latin root respondere. Individuals should be reactive and show the cognitive capacity to find relevant solutions in given situations (Fisher, Mahajan and Mitcham 2006). Moreover, as claimed by René von Schomberg (2013, 21): “technical innovators [should] become responsive to societal needs and societal actors [should] become co-responsible for the innovation process.” In the end, such an adaptive reciprocity should contribute to elaborate a common normative horizon that would ensure the co-construction of societal desirable products. As also suggested by Groves (2009), this process recalls Aristotelian idea of phronesis, the practical wisdom of prudence and sagacity that informs our choices and actions in a scientific, political, legal, social and ethical environment constantly moving.

In parallel to this, responsiveness is given a more systemic interpretation (e.g. Owen, Macnaghten, et al. 2013), where the capacity to adapt refers to the system as a whole rather than to individuals. This interpretation furthers a reflection in terms of governance and institutional design, which investigates the way in which critical voices can be listened to and carry weight in the decision-making process (e.g. Wynne 1993; Jasanoff 2003; Stirling 2008). In this interpretation, the focus is geared towards the institutional settings that allow to integrate conflicting voices: funding criteria, risk/benefit assessment practices, the relationships of innovators and scientists with ethical issues or the dialogue with interest groups and civil society.

1.5. Reflexivity

Proceduralist approaches to the responsible practice and management of research and innovation also tend to invoke the condition of reflexivity, which entails that the framings of social actors through which they identify issues, assess situations and propose possible answers can be called into question by those actors themselves. As with responsiveness, reflexivity both involves an individual ability to reflect about one’s own perspective and a systemic quality of R&I processes. This condition bears on a tradition inspired by Beck ([1986] 1992), and furthered by Jasanoff (2003), Stirling 2008, Rip (2006), Voss and Kemp (2006), Wynne (1993), van Oudheusden (2014), among others, which outlines how the ways in which social actors (policy makers, researchers, engineers, end-users, the public at large, etc.) frame the development of science and technology influences the quality of solutions to perceived social problems and leads to the idea of reflexive governance.

In the RRI literature, a first level of reflection focuses on the cognizance of R&I actors concerning the various issues raised by their activities (Sutcliffe 2011; EC Report 2013; von Schomberg 2013). In the tradition of ELSI and ELSA approaches, this first level of reflexivity corresponds to an awareness about the ethical, societal and political issues created by scientific activity and technological development. It is a matter of engaging natural scientists, to only mention them, “in the wider impact of their work, while doing research in the laboratories” (von Schomberg 2013, 89). In this conception, embedded social scientists play a crucial role by challenging R&I actors and enhancing their capacities for responsibility (Fisher et al. 2015).

However, reflexivity is not limited to the identification of a list of ethical issues. In addition, it cannot only depend on the virtuous capacity of designers (scientists and innovators) to identify those issues. Such an approach would presuppose a deterministic relationship between “ethical issues” and technology that would neglect the constructivist power of social actors to define what is considered as an ethical issue. The construction of the problem would be ontologically bound to the properties of technology regardless of its social fabric. To escape this individualistic overtone and its epistemic consequences, a second level of reflexivity implies the unpacking of the different presuppositions held by R&I actors.

Reflexivity is then understood as “a mirror up to one’s own activities, commitments and assumptions.” (Stilgoe, Owen, and Macnaghten 2013, 1571). Individuals should enhance their ability to acknowledge the ways in which they seize and answer problems (Fisher and Rip 2013), including a questioning about the purposes and motivations underlying technology (Owen, Macnaghten, and Stilgoe 2012). Such a reflexive stance opens an area of reflection about the framing of the actors; it challenges the implicit presuppositions that guide and shape the norms and normative theories individuals defend.

The possibility of stepping back “from the substantive debates among theoreticians and practitioners of responsive or reflexive regulation in order to examine the presuppositions and commitments that are shared by those who engage in the debate” (Perez 2011, 760, my emphasis), can been framed as a “second order reflexivity” (Lenoble and Maesschalck 2003; Schuurbiers 2011; Wynne 2011).⁹ This would be a crucial step towards a co-construction of innovation and research since such a constructivist view allows social actors to reflect and eventually agree on both the construction of the problem and possible solutions.

2. Responsibility and substantive ends

In addition of the procedural structure detailed above, RRI frameworks and approaches also rely on specific sets of outcome-oriented norms that the process of RRI should try to achieve. This approach can be seen at the policy level, such as in the case of the pillars or “keys” defended by the European Commission, or the “broader Impacts” of the US National Science Foundation.

Taking the European context for instance,¹⁰ RRI approaches such as Sutcliffe (2011), the EC report (2013) or von Schomberg (2013) do not only argue for the fulfillment of conditions that are recognized as necessary to achieving responsible innovation and research. In these views, outputs of science and engineering should also adhere to general, often democratic (Wong 2016) principles and norms. For many European RRI authors, such norms coincide with those promoted by the Treaty of Lisbon, for instance, which have been democratically settled (approved by the European Parliament)¹¹ and which defend sustainable development, the enhancement of quality of life and a “highly competitive social market economy, aiming at full employment and social progress”.¹² This set of heterogeneous, general and substantive norms defines a normative horizon for RRI approaches and provides a foundation to build a normative theory of science and innovation management. In these perspectives, R&I practices, beyond being inclusive, reflexive, etc. (procedural dimension), should also achieve market competitiveness, through a sustainable way of production that meets EU citizens’ democratically legitimate norms about management and employment (outcome-oriented dimension). Responsibility would then also follow from the achievement of these norms.

Compared to procedural approaches, theories which also rely on outcome-oriented norms could seem to provide a more practical framework to implement RRI. Definite and possibly measurable objectives can even be assigned to R&I processes (in terms of CO2 emissions, levels of employments, for instance). However, in a pluralistic context, this strategy also increases the likelihood of value conflicts between opposed interests, such as when market competitiveness and sustainability inform divergent decisions, or when European citizens disagree on their assessment of science and technology.

A similar issue looms with the recent framework for RRI developed by the European Commission, which identifies six pillars of RRI:

• Engagement of societal actors

• Science education;

• Gender equality;

• Open access (to scientific knowledge, research results and data);

• Research and innovation governance;

• Ethics.

Owen, Macnaghten, and Stilgoe (2012) and de Saille (2015) have argued that this conception of RRI stemmed from a (successful) attempt to translate previously established Science in Society (SIS) categories into a RRI framework so as to position the SIS program within the new European funding program known as Horizon 2020, to maintain budgets, and to domesticate innovation into the program. In this sense, political drivers might have overshadowed issues of conceptual coherency in the elaboration of the various conditions for RRI (see Pellé and Reber 2015, 2016).

Concerning its normative strategy, this approach illustrates how procedural and outcome-oriented conceptions of responsibility can coexist in a same framework. Indeed, pillar one directly refers to what has been labeled as the condition of inclusion. Pillar two and four suggest a logic of better publicity and informed dialogue and thus echo the condition of transparency. These principles correspond to conditions that belong to the procedural structure highlighted in section 1.

Pillar three introduces an epistemological disruption since the promotion of gender equality can be both principle-oriented (by ensuring equality of opportunities for men and women, for instance) and/or outcome-oriented (e.g. guaranteeing parity in R&I practices). The question then emerges of why there is a focus only on equality between men and women instead of including a broader objective of social justice such as defended by Schroeder and Ladikas (2015), who propose to apply the Rawlsian difference principle to funding decisions in R&I, or Ziegler (2015), who reflects on the principles of a fair space for innovation.

3. RRI approaches and virtue ethics

A third strategy observed in the literature to confer a normative dimension to responsibility is inspired by the literature on virtue ethics and more precisely, by one strand of this literature focused on the idea of “care” (Gilligan 1982; Noddings 1984; Tronto [1993] 2009; MacIntyre 1999; Engster 2007; among others) and relies on individual and systemic virtues (Groves 2009, 2014; Pellé and Nurock 2012; Grinbaum and Groves 2013; Owen, Macnaghten, et al. 2013; Pavie 2014; Pellé and Reber 2016). In this perspective, relying on social actor’s virtue of responsibility – or as framed by Muniesa and Lenglet (2013), internalizing moral values – is a prominent condition of an ethical management of science and technology. This concerns not only researchers, who hold a responsibility linked with the artifacts they contribute to create (Douglas 2003; Mitcham 2003), but other social actors such as members of NGOs, CSOs or industrial firms, who also have to take their own responsibilities when taking decisions and actions. As mentioned by Grinbaum and Groves (2013, 240) “responsibility is only one of the virtues. Other, relevant to the scientific endeavor, perhaps include integrity, impartiality, honesty, lucidity, or openness.” This list might also apply to other concerned actors, from innovators to activists.

In this perspective, responsibility does not stem from conditions that should be satisfied or from the achievement of substantive norms. It relates with the capacity of individuals and organizations to act virtuously, and more precisely to care about other human and non-human entities. In the tradition of Hans Jonas, Groves (2009, 2014) and Adam and Groves (2011), for instance defend a future-oriented interpretation of responsibility: “Care is thus a way of conceiving of human self-concern as inherently futural, as bound up with the potential things have for making my projects turn out well or badly” (Groves 2009, 24). This ontological feature of future-oriented human beings underpins the relationship we weave with others: our interrelations lead us to care about our family, our friends and more generally about other human and non-human beings, future generation, the environment, animals (i.e. what is of constitutive value to us, Groves 2009), but also institutions, ideas and cultural objects (Groves 2014). This creates a form of responsibility through which we develop capacities for perceiving, acting and judging, so as to answer the needs of others.

More precisely, the idea of responsibility as care has been grasped through an analogy with the parent/child relation (Groves 2009, 2014; Grinbaum and Groves 2013). In this approach, with a sensitivity on the vulnerability of others (and future generations), scientists and innovators have the responsibility to “‘teach’ or ‘encode’ the virtues” in their created artifacts like parents try to do with their children. This analogy will encounter some limits insofar as the autonomy of technological artefacts, once they evade their creator’s control, cannot be compared to the one of a child becoming adult. But the idea underlying the metaphor is that “during early elaboration stages, technologies are dependents with sensitive and malleable potentialities, in which it is hoped the ingenuity of innovators will produce […] particular virtues”, that is, “the efficient production of ‘right impacts’” (Grinbaum and Groves 2013, 131).

As in the value-sensitive design approach to technology (Kelty 2009; van den Hoven 2013) and other forms of socio-technical integration (Fisher et al. 2015), researchers and innovators are supposed to shape the design of technology and to accompany its development in a “good” way without being tied to the obligation of forecasting all possible consequences. Because it focuses on needs, care ethics includes a form of anticipation but in a much less rationalistic way than traditional technology assessment activities (Koehn 2011; Grinbaum and Groves 2013).

Yet, a crucial question that is raised when applying care ethics to R&I concerns the way in which “good” or “responsible” or “virtuous” practices are defined and discussed. As we saw with the case of selecting one set of principles or measures over another in outcome-oriented approaches, so also the selection of one or more virtues in RRI approaches is itself normative and may also occasion value conflicts. Moreover, is it possible – and even desirable – to identify a set of virtues that would ensure or at least promote RRI? Finally, going back to the tension between the democratic requirement for pluralism and the practical need for applicability, how can such a set be identified so as to be compatible with multiple value systems?

4. Moral pluralism and the application of norms in RRI approaches

These questions directly stem from the tension outlined in introduction between, on the one hand, the need for a pluralist assessment of responsibility and, on the other, the constraints imposed by the practical implementation of norms. The three normative strategies (procedural, consequentialist and virtue-based) used so far by the RRI literature to ground and justify the idea of responsibility in R&I deal differently with this tension and offer various types of resolutions. The purpose of this section is to highlight how the tensions emerge in each of these strategies and to explore the pathways they offer to answer them.

4.1. Procedural theories and the co-construction of the context

As shown in section 1, RRI approaches that have a strong procedural overtone establish a close link between conditions that theoretically have to be fulfilled and the possibility of achieving responsibility in actual R&I practices. Although compliance with such conditions would not be a guarantee against the possibility of moral luck (Williams 1981; Grinbaum and Groves 2013), that is, bad outcomes resulting from well-intentioned decisions, all together, they are meant to help individuals and institutions face many obstacles and sketch responsible pathways of development for R&I.

Such a proceduralist perspective avoids some of the issues raised by outcome-oriented norms: because procedural approaches do not identify a definite objective to achieve (e.g. gender equity or the reduction of CO2 emissions), they are less susceptible to value conflicts. Although authors disagree on the exact set and number of conditions of RRI, they tend to mostly agree on the content of these conditions. However, practically implementing these conditions can raise numerous challenges.

For instance, taking the condition of reflexivity: almost all theories insist on a critical and reflexive process to revise judgments and frameworks of assessments. But they often neglect the practical conditions through which reflexivity can emerge. How to enhance individual capacities for reflexivity? At a systemic level, what institutional design should be implemented to foster individuals’ and organizations’ capacity for reflexivity (Wynne 1993, 2011)? Existing propositions, such as embedded social scientists, ethical training for scientists and innovators or funding criteria that give ethical assessment and responsibility a binding power are illustrations of such institutional settings.

But implementing second order reflexivity would need to go a step further. According to the context, the type of innovation, the historical and cultural background of a specific community, the interpretation and application of norms will differ significantly. To have the chance of being effectively implemented and followed by actors, norms cannot be decided in an abstract or pre-determined way and must, it is suggested here, result from a co-construction of the issues and answers at stake.

Such a co-construction of the context (including the relevant problems to be addressed, the information about them, the outcomes to be assessed and the possible solutions) is a fundamental condition for an effective application of norms (Maesschalck 2001; Lenoble and Maesschalck 2003). Nor can this “reflexive” stance rest on a purely rationalistic exercise. Rather, the construction of norms and the process of negotiation has to leave room for discussing the effects of norms. When applied to institutional settings, this would mean exploring the kind of institutional learning by which the design of participation and deliberation processes could be corrected and modified according to their capacity to ensure (a) the quality of dialogue and (b) the possibility of critically assessing one’s own perspective. For instance, is it possible to design processes of open innovation or crowdsourcing¹³ practices, which seek to integrate non-traditional actors of innovation, to enhance the possibility of challenging stakeholders’ framings? Answering such practical questions will help to design participatory and deliberative structures that allow social actors to reflect on their own perspectives and judgments.

Turning to the condition of responsiveness, section 1 recalled that most RRI approaches assume that individuals or R&I processes must constantly adapt to a changing environment. But, in this view, there is no normative horizon that can help to define what the good pathway to follow is. Having a capacity to respond quickly can prevent from sticking to wrong patterns of solutions, but it does not necessarily lead to morally “good” answers (Pellé and Reber 2016). Rather, from a constructivist perspective on norms, the way in which good solutions are said to be good derives from a collective production of norms, through participation and deliberation.

Yet, in relying on the condition of inclusion, the problem of defining a collective normative horizon has only been displaced: in a procedural framework, participation and overall deliberation are given the power to reveal the norms we have to follow. As Richardson puts it (1999, 244): “moral deliberation may be understood as epistemically oriented; that is, it is oriented not toward the achievement of some nameable goal, but toward generating right answers about what we ought to do.” In this view, deliberation and participation are not aimed at securing specific outcomes but at forging our normative horizon (our political ends). Deliberation is acknowledged to be a fallible process that secures pluralist and collective judgments about norms (without having the epistemic quality of leading to morally good outcomes).

If most RRI approaches attempt to apply such an ideal of deliberative democracy to processes of research and innovation, many issues remain in the dark. Firstly, in procedural deliberative theories (such Habermas’ discourse theory, for instance) the intrinsic dynamic of rational argumentation is meant to help build a common ground for norms. But RRI theories remain unclear concerning such a dynamic, its complex mechanisms and limits. They often neglect the strategic dimension of deliberation (Selin 2007; van Oudheusden 2014; de Hoop, Pols, and Romijn 2016), but also the question of the grounds on which mitigating between conflicting values or that of the normative framework through which value conflicts (or conflicts of value systems) can be resolved (lexicographic order, costs/benefits, majority rule, etc.).

Secondly, inclusion is sometimes interpreted as mere consultation (Sutcliffe 2011; EC report 2013). In this case, the focus is on social acceptability gained through an increased legitimacy of research and innovation products, and on meeting societal needs in order to avoid market failures by revealing consumers’ preferences. Consultation is potentially a first step as it helps unveil social actors’ preferences, interests and values. However, it falls short of co-construction, since actors are not influential or decisive participants, only mere spectators whose voices are collected without any real impact on the final decisions (see e.g. Fung 2006).

In other RRI approaches, participation and/or deliberation is given a prominent role (Owen, Macnaghten, et al. 2013; Sykes and Macnaghten 2013; Stilgoe, Owen, and Macnaghten 2013). But the practical conditions through which a mutually responsive process of technology development can be implemented are often left aside. For instance, the conditions and requirements of legitimate deliberation are not really explored. Sykes and Macnaghten (2013) propose practical conditions for a “good” public dialogue (related to a pluralist way of setting the agenda), drawn from their study of various deliberation experiments; they also defend the possibility to have a nuanced discussion (and not only a battle between pros and cons). Nevertheless, deliberation here appears to have the same meaning as “public dialogue,” and public dialogue as such does not ensure a quality argumentation (Blok 2014), as has been illustrated by the French national debate on Nanotechnology (Laurent 2010), for instance.

Yet, the quality of a deliberative process is a key feature of its efficacy and the literature on political theory and political philosophy has been keen to provide various conditions to ensure fruitful experiments in public engagement and deliberation (Fishkin and Luskin 2005; Thompson 2008; Cohen 1989, 1997; Mansbridge 1983; Rawls 1971, 1993; Habermas [1981] 1984–1987; Steiner 2012, among others). Jürg Steiner (2012), who discusses and confronts these various conditions with empirical evidence, considers the following requirements: the participation level of the community members, the use of narratives to formulate an argument, the type of justification used (based on a reference to the common good or on self-interest), the level of respect between participants, the appropriate level of transparency (when early phases of deliberation need confidentiality), the status of the force of the better argument or the need for truthfulness. While not all of these requirements will necessarily be of equal relevance in the context of RRI, discussing them could help clarify the settings of participation and/or deliberation processes and the way in which they are expected to affect R&I developments. For instance, what institutional design is the most appropriate in a specific context: mini public, consensus conference, workshops, forums, etc.? By contrast, there are far fewer discussions in the RRI literature on the practical criteria and the structure of a good deliberation in R&I processes.

To sum up thus far, procedural conceptions of RRI identify relevant conditions for RRI defined in a pluralist manner since they do not rest on any substantive definition of responsibility. Yet, they often forget to discuss how these conditions will be applied in specific contexts. The specific forms that broad requirements such as responsiveness or reflexivity should take needs further theoretical and empirical work. This would help to understand how they effectively shape R&I practices and inform particular social devices towards greater responsibility.

A final issue is raised by the procedural dimension of RRI approaches, which holds implications for responsibility. In the Rawlsian scheme (Rawls 1971, 1993), for instance, members of society cannot contest inequalities resulting from a fair process of distribution (satisfying the two principles of justice elaborated under the veil of ignorance). Once primary goods are fairly distributed and social and economic inequalities are to be arranged so that they are both at the greatest benefit of the least advantaged and ensuring fair equality of opportunity,¹⁴ the distribution of resources is said to be fair, even if there are remaining inequalities. Similarly,¹⁵ in procedural approaches of RRI, social actors would not be able to hold researchers, scientists, actors of innovation, etc. responsible for possibly bad outcomes stemming from a process of R&I that satisfies the conditions of responsibility. Going back to the issue of moral luck, this would mean that if a damage occurs but results from a responsible process of decision-making (inclusive, reflexive, transparent, responsive and anticipatory), responsibilities¹⁶ will be difficult to ascribe. More generally, the conditions of RRI cannot secure the outcomes of R&I processes, which sometimes would not be desirable. Moreover, by avoiding assessing the moral dimension of the outputs of R&I processes, procedural approaches neglect whole sets of morally relevant information. Yet, as stated by Stirling (2008) in the case of participation, for instance, well-designed top-down governance of technological development can sometimes lead to morally better outcomes than participatory and deliberative approaches. This is perhaps where RRI approaches that add to their proceduralist structure a concern for the moral quality of R&I outcomes or for the moral dispositions of social actors can fare better.

Outcome-oriented conceptions of RRI are more focused on whether certain events will come about or not. Achieving the European pillars, for instance, would ensure a certain level of gender equity in R&I practices or a satisfactory knowledge-flow about R&I activities. However, such a view has to face important issues related with the determination of good in a pluralist way, that is, with the boundaries of the conceptions of responsibility that is held. As the brief discussion above on the European pillars suggests (section 2), in a list of “good practices” or “good norms” what to include, what to reject? More generally, how to ensure that a specific conception of the good promoted by substantive norms of responsibility will be subject to a broad agreement? Compared to procedural structures, theories that defend identified outcomes or even definite principles without inscribing them in a proper theory of justice will be even more subject to unresolved value conflicts. The European debate on GMO’s for instance illustrates how substantive norms of responsibility (e.g. allowing or not the growing of new GMO’s species) can be subject to conflicting positions. In this context how to collectively define what is responsible?

4.2. Interpreting responsibility as care

This is where RRI approaches that rely on an interpretation of responsibility based on virtue and more precisely on care offer an interesting alternative to both ensuring a pluralist determination of the normative horizon and providing a framework suited for effective implementation.

Elsewhere (Pellé 2016, 2017a, 2017b; Pellé and Reber 2016), I carefully analyze how RRI can be fruitfully understood through the lens of care ethics. Let me mention three main aspects here. Firstly, by interpreting responsibility as care, RRI approaches are anchored in a strong normative framework that is focused on specific attention to the demands and the needs¹⁷ of other relevant and affected parties, and to the quality of answers that are provided to these demands. The goodness and rightness of good caring practices is not identified a priori by a principle or by a calculus. They do not stem from the application of a procedure or by the attempt to secure a specific outcome. Rather, care ethics is a context-dependent approach that puts a specific emphasis on connectedness and the relational aspect that exist between a care-receiver and a care-giver (Gilligan 1982; Tronto [1993] 2009; MacIntyre 1999; Engster 2007, among others).

Tronto ([1993] 2009) for instance, insists on some conditions of good care that imply the competencies and dispositions (virtues) of the care-giver but also the ability to respond of the care-receiver, who has an active role in defining what is appropriate to him or her. These relational dispositions can be extended to R&I actors (Pavie 2014), but also between social actors and non-human entities such as institutions or the environment (Groves 2014) and can lead to principles of governance that tend to favor care relationships between these entities (through training and education, appropriate design for funding criteria and institutions, organizational culture, etc.). Grounded on the interconnected needs of social actors instead of being based on reciprocal rights for instance (Adam and Groves 2011), the care perspective leads to action through the elaboration of norms that are sensitive to the context. Compared to procedural approaches, a care-based definition of responsibility ensures a certain level of applicability: norms are elaborated as close from the practices and specific situations in which they apply and at the same time stem from a normative horizon – to take care of the “objects” (human and non-human) affected by R&I¹⁸ – that gives a firm orientation to action.

In parallel, interpreting responsibility in R&I through the idea of care is fundamentally compatible with a pluralist determination of norms. Although virtue ethics is sometimes associated with an “essentialist” approach (e.g. Nussbaum 1993),¹⁹ such a position is not necessary and what is recognized as good caring practices and relationships in RRI can also stem from a collective and deliberative process of assessment. No top-down, expert-driven or authoritative governance is needed to define what virtuous R&I practices would be. Rather, in many care perspectives, definitions of good care relationships emerge from practices, are historically and culturally bounded, susceptible to evolutions and changes and imply the participation of both caregiver and care receivers (Tronto [1993] 2009). In the context of R&I, this would mean that understanding R&I actors’ responsibility as a disposition to care about other human and non-human entities allows for a collective determination of the good, which involves the active participation of the various R&I actors.

Finally, care ethics makes an explicit link between action and responsibility, which reinforces the ability to derive appropriate decisions from norms elaborated through this perspective. When we care about things (ideas, institutions, persons, etc.) we can act to defend them: care implies action. As formulated by Groves (2009, 13), the experience of caring is the root of ethical behavior – that is, “acting in ways which value others in themselves.” Pandza and Ellwood (2013) provide informative insights on the subject. In an ethnographic study of a research group working on nanotechnology, they find that when social actors of science-driven innovation are challenged by uncertainty and changes in the mandate of science, they react by relying on “ethical and strategic” agency (instead of traditional re-iterative agency) so as to develop new organizational capabilities related to responsibility. In other words, in a context of uncertainty and new pressures, actors of innovation sense a change (and might fear it) but they “are also willing to get engaged into development of new capabilities that may help to meet the issue of responsibility” (1122). Such a study provides empirical evidence of the strategic role that virtues play in the scientific actors’ behavior.

Moreover, the ontological responsibility to care about (and for) others and the environment is not limited to the individual. Drawing on Hannah Arendt’s conception of collective responsibility (that differs from guilt) Grinbaum and Groves (2013) defend the idea that scientists are collegially responsible not only for developments they directly contribute to, but also in a wider sense for developments they indirectly contribute to, as a community (Pellé and Reber 2016). Care ethics creates a forward looking link between responsibility and actions that normatively requires – and practically motivates – actors to take care of other actors, institutions and the environment affected by their activities. In this sense, interpreting responsibility in R&I through the idea of care offers a possible pathway of resolution to the tension between pluralism and applicability, mentioned at the outset: care ethics allows the elaboration of norms that can be both collectively defined (taking the various normative systems at stake into account) and practically oriented towards action and decision-making.

Conclusion

This paper has showed that three main types of moral reasoning have influenced most of RRI approaches: proceduralism, consequentialism and virtue ethics. None of the current RRI approaches fully holds one or the other position. Rather, RRI theories borrow elements from these various strategies to ground responsibility. By highlighting these different strategies, the paper attempted to analyze some strengths and blind spots of RRI approaches, concerning one issue specifically: the tension between the concurrent need to hold a pluralist perspective of “the good” and the necessity of defining norms of responsibility that allow practical implementation. I argued that while procedural approaches allow for a pluralistic definition of responsibility where no substantive content is given to the word, they lack practical relevance and applicability, since they do not specify the particular norms that could stem from the general conditions they defend. Outcome-oriented approaches offer a framework in which the normative horizon to be pursued can be identified in a less ambiguous way, insofar as responsibility is defined by the achievement of particular outcomes. While these approaches are more easily applicable since they provide with clear and possibly measurable objectives to be followed, however, they require a form of social agreement and consensus on the outcomes to be attained, which can be difficult to obtain when members of pluralist societies hold various and sometimes conflicting values and value systems. Finally, interpreting responsibility as care can offer an interesting solution to the dilemma. On the one hand, it is a context-dependent approach, where norms of good care are elaborated in relation to the context. As such, it implicates all affected stakeholders and is compatible with a pluralist and collective determination of the norms of good. On the other hand, because it is focused on caring relationships between R&I actors, social actors and non-human entities such as the environment or key institutions, it provides a normative horizon that can guide the elaboration of specific norms, allowing for practical applicability and implementation.

Acknowledgments

I would like to thank all the members of the GREAT project and especially Bernard Reber and Philippe Goujon for their fruitful comments. I also would like to thank Richard Owen for his precious help as well as the two anonymous reviewers of the journal who provided challenging comments.

Notes on contributor

Sophie Pellé After completing her Ph.D. in philosophy of economics (Université Paris 1 Panthéon-Sorbonne), the author has completed one post-doctoral position on the ethics of nanotechnology (Université Paris 1 Panthéon-Sorbonne) and is now at Sciences Po (CEVIPOF) and works for the EU funded FP7 project GREAT (Governance of Responsible Innovation) as a post-doctoral fellow.

Additional information

Funding

The research leading to these results has received funding from the European Commission's or Union's Seventh Framework Programme (FP7/2007–2013) [grant number 321480] (project GREAT).

Notes

1. I did not, however, search for any specific dimensions of innovation in relation with responsibility. For treatments of this subject see for instance Pavie, Scholten and Carthy (2014), Pavie and Egal (2014), Ziegler (2015), Blok and Lemmens (2015) and Pellé (forthcoming-a).

2. Including members of FP7 EU funded projects on RRI and scholars deemed relevant from the science & technology studies (STS) community.

3. Schroeder and Ladikas (2015) and Ziegler (2015) adopt a principled view where the Rawlsian difference principle should help allocate funds for R&I and regulate the development of innovation.

4. In the tradition of Kant’s theory, a formalist approach only provides a formal rule to elaborate norms without any reference to content or context. But procedural rules in RRI theories are often substantive and prescribe specific content that procedures should achieve.

5. Which draws on four “p’s”: product, process, purpose and people

6. See Blok and Lemmens (2015), Pavie, Scholten and Carthy (2014), Pellé and Reber (2015, 2016) for a critical reflection about transparency.

7. At early stages of technology, low information prevents from providing relevant assessments and forecasts, while when the technology is more mature, better information is available but the possibility of influencing technological design and pathway of development has been reduced (Collingridge 1980).

8. Both authors differ though, among other points, on their interpretation of the substantive dimension.

9. Cf. EU funded GREAT project. http://www.great-project.eu/

10. See Schroeder and Ladikas (2015) for elements of comparison between the European 6 pillars and the US “broader impacts”.

11. Although the process through which the Treaty has been approved, mostly without the recourse to referendum and after the rejection of the European Constitution by several national referendums, is subject to criticism.

12. Article 2 of the Treaty of Lisbon (2007).

13. Crowd sourcing designates new ways for individuals or companies to obtain needed services, ideas or content by soliciting contributions from a large group of people, usually an online community, rather than from traditional employees or suppliers. For suggestions about its application to responsible innovation, see Winickoff, Jamal, and Anderson 2016).

14. And assuming that members of society are “reasonable” citizens.

15. But the comparison is only a limited one, since RRI approaches do not rest on a complex structure such as Rawls’ theory.

16. Interpreted as blameworthiness or liability (Pellé and Reber 2016)

17. Even if the concept of need has to be critically reinterpreted in the case of R&I practices (Pavie 2014; Pellé forthcoming-a, forthcoming-b).

18. This is a broad definition of the idea of care. See Groves (2014), Engster (2007) or Pellé (forthcoming-a, forthcoming-b), among others, for a discussion on various definitions and aspects of these definitions.

19. “Nussbaum (1993, 2000, 2006) has constantly argued in favour of an approach of justice based on the capabilities that human beings are able to exercise. She defends a list of basic capabilities that allows for human flourishing in a way that is supposed to transcend cultural and historical variations.”