Responsible research and innovation in innovation value chains: focus on the catalytic role of non-governmental organizations

ABSTRACT

This study explores the role of NGOs in Innovation Value Chains (IVCs) and Responsible Research and Innovation (RRI). It investigates how NGOs' presence influences RRI adoption in ecosystems and whether their strong RRI focus enhances its diffusion. Agent-Based Modelling and Simulation are employed for analysis. Findings reveal that NGOs promote RRI when adopting intermediate RRI values for consortium evaluation, whereas strict criteria hinder RRI diffusion. Highly RRI-oriented NGOs foster knowledge dissemination and network diversity, despite limited IVC involvement. These insights offer guidance for Research Funding Organizations to promote RRI and assist Research Performing Organizations in managing dynamic innovative networks.

KEYWORDS:

• Responsible research and innovation
• innovation value chains
• complexity
• agent-based modeling
• NGOs

Introduction

The Innovation Value Chain (IVC) view presents innovation as a three-phase process that involves idea generation, idea development, and the diffusion of developed concepts (Hansen and Birkinshaw 2007). Across all the phases, innovation can be considered a complex, dynamic, and collective process (Owen, Bessant, and Heintz 2013) in which many agents are involved in building new knowledge, networks, and social interactions (Felt et al. 2007; Geels 2002).

Frameworks such as the Triple (Etzkowitz and Leydesdorff 2000) and the following Quadruple Helix (Carayannis and Campbell 2009) emphasize the heterogeneity of agents involved in innovation processes. In the Triple Helix, the three primary actors are academia, industry, and government. These actors are viewed as interdependent and engaged in a process of co-evolution, where their interactions and collaborations can lead to innovation and economic growth. In addition, the Quadruple Helix model includes a fourth actor: civil society. This comprises non-governmental organizations (NGOs), community groups, and other social actors who may have a stake in the innovation process (Cornett 2009; Lindberg, Lindgren, and Packendorff 2014).

Since innovation emerges from interactions within these complex networks of actors and relations, it is in such a framework that responsible research and innovation (RRI) principles must be located (Arnaldi and Neresini 2019; Owen, Macnaghten, and Stilgoe 2012). The RRI has become increasingly important since its birth (Owen and Pansera 2019), proposing itself as a new inclusive approach that mitigates the potential adverse effects of research and innovation (von Schomberg 2013). In this sense, diffusing RRI principles among involved agents is crucial to ensure that innovation is responsible and meets societal needs.

Considering the actors involved in the Quadruple Helix model, while the roles of academia, industry, and government in promoting RRI have been widely studied in the literature (e.g. Gurzawska, Mäkinen, and Brey 2017; Marschalek et al. 2017; Silva et al. 2019), the role of civil society, particularly NGOs, has received comparatively little attention. NGOs operate independently from the government and are typically non-profit and non-partisan (Lewis 2010). Such organizations are critical members of civil society and have the potential to play a significant role in promoting RRI.

According to Ahrweiler et al. (2019b), NGOs are recognized as fundamental elements for realizing RRI in the present research and innovation processes. Their involvement is considered an essential pathway toward successful RRI implementation (Cavallaro et al. 2014). In addition, NGOs have attracted increasing attention concerning their impact on creating networks (Fisher 1997) and adding specific scientific capabilities and expertise, demonstrating their social responsibility (Sutcliffe 2011). Furthermore, their presence in the network increases the diffusion and sharing of knowledge, sharing new capabilities and abilities that other agents do not have (Ahrweiler, Frank, and Gilbert 2019a; Böschen et al. 2020; Cavallaro et al. 2014; Lang and Griessler 2015; Sutcliffe 2011; Szarka 2013).

With a particular focus on the diffusion of RRI practices, we wonder how the presence of NGOs pushes agents toward pursuing RRI practices. In addition, we aim to understand if cooperation with NGOs ensures better performance in knowledge dissemination and inclusiveness of agents involved in the IVCs opened to RRI. More in depth, the research questions are: How do NGOs enhance the diffusion of RRI practices in IVCs networks? How do NGOs ensure better performance in knowledge dissemination and inclusiveness within innovation networks?

To this aim, we adopt an Agent-Based Modeling (ABM) and related simulations to investigate the evolution of networks of IVCs in industrial contexts, modeling the role of NGOs and the openings to RRI keys and practices. AMB is used as a research approach to consider the heterogeneity of actors populating the system and the massive non-linear relationships among them, letting collective patterns or behaviors emerge from the bottom up (Gilbert and Terna 2000).

Particularly, we adapt the agent-based model named I AM RRI SKIN model (developed as part of the research activities related to the Horizon 2020¹ carried out under the EU’s SwafS-12-2017 call; European Commission 2020) to investigate the behavior of NGOs and the role they play in diffusing RRI in networks of IVCs. The contribution of NGOs exchanging RRI practices among the various components of the created networks is analyzed. In addition, the model proposes a characterization of the agents’ knowledge base coherent with the context of additive manufacturing (Cozzoni et al. 2021). This implies that the model can study IVCs that open to RRI principles and practices in industries. At the same time, the model is not perfectly based on a specific empirical case, it is instead a typification (Boero and Squazzoni 2005), that refers to a class of similar empirical phenomena, thus allowing for theory grounding or testing, and also for following specifications in other real fields. Our study makes a dual contribution. Methodologically, the novel contribution is represented by modeling NGOs within IVCs networks with highly innovative projects informed by RRI. From a theoretical perspective, we emphasize the importance of avoiding overly rigid RRI evaluation strategies, offering practical guidance for NGOs and industrial partners in adopting responsible practices.

The reminder of the paper is the following: Section ‘Literature background’ discusses the theoretical background of the work; Section ‘Agent-based modeling and simulation approach’ illustrates the characteristics of the methodological approach; Section ‘The proposed model’ describes in detail the modeling of NGOs in IVCs; Section ‘Simulations and results’ reports the design of the experiments and simulation’s results; discussions and conclusions end the paper.

Literature background

RRI and innovation value chains

Responsible Research and Innovation (RRI) is an approach that aims to promote responsible and sustainable innovation by encouraging public participation and transparent communication about the goals, risks, and impacts of innovation (Kuzma and Roberts 2018) and has recently been integrated into industrial innovation processes (e.g. Arnaldi and Neresini 2019; Gurzawska, Mäkinen, and Brey 2017; Lubberink et al. 2017; Martinuzzi et al. 2018; Van de Poel et al. 2020). Indeed, in recent years, also driven by the policy debate, several researchers have looked at RRI as a catalyst for innovation processes, studying its implementation from the perspective of strategic innovation management (e.g. Cozzoni et al. 2021; Panciroli, Santangelo, and Tondelli 2020; Thorstensen and Forsberg 2016).

To model the process by which firms gather the knowledge needed to develop an innovation and therefore create growth and added value, Hansen and Birkinshaw (2007) propose the notion of the Innovation Value Chain (IVC) as a general framework within which almost all firms’ innovation activities can be considered (Ganotakis and Love 2012). The same authors define the IVC as a sequential, three-phase process that involves idea generation, idea development, and the diffusion of developed concepts (Hansen and Birkinshaw 2007). Innovative initiatives can emerge anywhere in the local ecosystem or the organizational ecosystem and move from one organization to another, adding value through transactions or with cooperation between agents involved (Ponsiglione, Quinto, and Zollo 2018); therefore, recent studies suggest that different IVCs can coexist by forming networks in which innovation born in one organization can boost multiple innovations in different networks as well (Vinogradov, Nguyen, and Jensen 2021).

Using this perspective, administrative and policy debate has encouraged identifying and exploiting openings for RRI practices in IVCs (European Commission 2017) by seeking to implement RRI from inside the industry rather than imposing RRI on it from above (e.g. Arnaldi and Neresini 2019; Tabarés et al. 2022). IVCs divide the business innovation process into three distinct stages (Hansen and Birkinshaw 2007). Seminal work by Stilgoe, Owen, and Macnaghten (2013) has shown how including RRI criteria for moving to the next stage can anticipate previously unexplored impacts, applications, and problems. Therefore, the stage-gate approach in R&D and innovation (Grönlund, Sjödin, and Frishammar 2010), and therefore in IVCs, can include and create openings for RRI in the assessments during the ‘gate’ and during the ‘stage’ to anticipate the eventual assessment (European Commission 2017; Nordmann 2014). In this way, accountability and innovation assessment tools lose their retrospective character but become proactive with the expectation of increasing social acceptability, sustainability, and desirability (Wender et al. 2014).

According to Cozzoni et al. (2021), RRI plays a key role in IVCs by promoting inclusive and sustainable innovation, which involves different stakeholders in creating and distributing innovations in the marketplace. Indeed, RRI plays an essential role in IVCs as it promotes active stakeholder engagement and transparent communication about innovation’s goals, risks, and impacts (Kuzma and Roberts 2018).

RRI is involved in all three phases of IVCs. First, in the ideation phase, RRI identifies the needs and expectations of stakeholders involved in IVCs (Stilgoe, Owen, and Macnaghten 2013). In this way, innovative ideas and solutions that address stakeholder needs and concerns emerge, identifying and assessing the innovation’s potential environmental and social impacts early in its development (Wender et al. 2014). However, there is no shortage of uncertainty about the effectiveness of RRI at this stage, which could result in an onerous restriction and cause delays in the innovation process resulting in a competitive disadvantage (Kuzma and Roberts 2018).

In the prototyping and development phase, RRI can evaluate the prototype and define the features and functionality of the innovative product or service. Adopting an RRI approach at this stage can bring benefits such as increased end-user confidence and adoption of the innovative product or service, achieving better social and economic outcomes (Van de Poel et al. 2017). However, at this stage, Kuzma and Roberts (2018) identify potential negative publicity (through national media) as the main cause of investor aversion and, therefore, as the main barrier to the adoption of RRI practices.

At the commercialization and diffusion stage, RRI can be used to define marketing and distribution strategies to ensure that the innovative product or service is marketed and distributed ethically and responsibly (Gurzawska, Mäkinen, and Brey 2017). In addition, RRI can be used to monitor and assess the impacts of innovation on society and the environment after the innovative product or service is launched in the market (Wender et al. 2014). Once the product is on the market, there seems to be less resistance to implementing aspects of RRI; in fact, innovators and developers see inclusion as a way to control the narrative and inform people about the availability and benefits of the technology (Kuzma and Roberts 2018).

In addition to changing the managerial strategy within a stage of IVCs, RRI impacts decision-making strategies in establishing an innovative network and in the transition from one stage of IVCs to another. In fact, as shown by the work of Cozzoni et al. (2021) in the gates, both technical and commercial criteria (Hansen and Birkinshaw 2007) and criteria referring to the dimensions of RRI must be met (Owen, von Schomberg, and Macnaghten 2021; Randles, Tancoigne, and Joly 2022; Stilgoe, Owen, and Macnaghten 2013).

According to the literature discussed above, Figure 1 shows how the role of RRI is integrated in the IVCs development process through stages and gates.

Figure 1. RRI and IVC (Authors’ elaboration).

NGOs in IVCs under RRI considerations

According to Cavallaro et al. (2014) the active involvement and engagement of end-users and civil society stakeholders are essential for effectively implementing RRI and ensuring innovation that is more ethical and beneficial for society.

Public dialogs reflect a deep mistrust of citizens not only of companies, but also of governments, in terms of lack of endorsement of technology, product choices, etc. This lack of trust poses challenges for governments and companies in convincing citizens that the scientific and technological investments they support serve the greater public good rather than being driven solely by financial or personal interests (Sutcliffe 2011).

According to Kuzma and Roberts (2018) several barriers to adoption of RRI practices are present in innovative projects. Gurzawska, Mäkinen, and Brey (2017) define incentives and practices that can overcome these barriers in the industry context, including the involvement of NGOs as an actor capable of providing diverse expertise and perspectives. Indeed, NGOs avoid a top-down imposition of RRI in Triple-Helix systems (Wakunuma et al. 2021), playing the role of a key intermediary actor in promoting RRI in industrial settings (Arnaldi and Neresini 2019).

The emergence of the Quadruple-Helix framework, building upon the Triple-Helix model, recognizes NGO as intermediaries between citizens, communities, and the other three helices of academia, industry, and government (Carayannis and Campbell 2009). Within the Quadruple-Helix framework, NGOs bring forth the perspectives, concerns, and aspirations of civil society, providing a voice for stakeholders who are often underrepresented in traditional decision-making processes (Wakunuma et al. 2021; Cavallaro et al. 2014).

Therefore, NGOs have become increasingly interested in this area and in sharing knowledge about how governments and companies do what they do, as new approaches are needed to involve all groups in thinking about the choices and decisions that are made. An essential aspect of RRI is to establish a mutual understanding among the European Commission, governments, businesses, and NGOs, which is crucial for instilling confidence in the public and other stakeholders regarding the safety and efficacy of innovative systems, processes, and products (Sutcliffe 2011).

The entities involved in RRI are the same as those involved in any research, development, and innovation (R&D&I) process (i.e. industry, researchers, civil society organisations (CSOs), policy-makers – including representatives of the European Commission – and universities – Gurzawska, Mäkinen, and Brey 2017). One of the challenges of RRI is to be more inclusive in involving the public at all stages of research and innovation, which could imply the participation of the public and other stakeholders – such as CSOs or NGOs – at the very beginning of the process in shaping a vision of the future toward which innovation can be directed (Cavallaro et al. 2014).

Moreover, the earlier the NGOs participates in an innovation network, the more influence it can have and the greater its impact on the results (Ahrweiler et al. 2019b). It is therefore evident that the timing of NGOs participation and its influence on the project is relevant. NGOs play a key role in the implementation of inclusive business models through which industry seeks to target the ‘base of the pyramid’ (Cavallaro et al. 2014).

According to Böschen et al. (2020), the governance of a research project in which an NGO participates is influenced by two main variables:

1. The social interaction scheme determines the authority of the NGO within the project and defines its role and activities. This variable distinguishes between projects where NGOs have limited involvement in the main decision-making processes and projects where NGOs take the lead.

2. The importance of the NGO for knowledge production describes the variables that might lead to a difference in influence in knowledge construction within a project given the participation of the NGO.

Following the above premise, some questions about NGOs in innovation networks arise: What is the reason for project consortia to involve NGOs as partners? What do they expect their specific contribution to be? Do they get a unique and coherent role in project consortia? Are NGOs the main or only drivers of RRI within projects?

The work of Ahrweiler et al. (2019b) aims to understand the role of NGOs within the networks of innovation value chains and to see the real contribution they make. For this purpose, a quantitative online survey was conducted. The data were collected from projects funded under the Competitiveness and Innovation Programme 2007–2013, in Information and Communication Technologies (ICT) under the Policy Support Programme of the European Commission’s 7th Framework Programme. These projects were per se innovation-oriented, and we can assume that the survey’s results are quite significant for the characterization of NGOs with respect to their role in IVCs. The questionnaire showed that NGOs were partners in 53% of all projects. In all projects with NGO participation, at least two NGOs participated. The survey shows that NGOs can choose whether to take part or not in a project. If NGOs enter into a network project, there will always be more than one NGO involved. In addition, NGOs decide to participate in projects that are civil-oriented and work in the same field as themselves. When an NGO enters a project its involvement in all phases is about the same. It’s important to note that this survey highlights the fact that including NGOs as partners in consortia is only partly motivated by the expectation that these would provide RRI competences.

Cavallaro et al. (2014) indicate the inclusion of NGOs within a project as a key element and catalyst for RRI practices. Greater adoption of RRI practices results in greater inclusiveness and heterogeneity of working groups (Cozzoni et al. 2021; Fitjar, Benneworth, and Asheim 2019; Van den Hoven 2013). Therefore, the involvement of NGOs indirectly implies a variety of actors involved, which ensures a richer discussion and better decisions (Cozzoni et al. 2021). Moreover, several authors emphasize the incidence of NGOs in knowledge production and dissemination due to the particular capabilities and abilities they possess; in fact, cooperating with a NGOs allows to access and exploit specific scientific data and expertise including data, not directly accessible by other actors of the ecosystem (Ahrweiler, Frank, and Gilbert 2019a; Böschen et al. 2020; Cavallaro et al. 2014; Lang and Griessler 2015; Sutcliffe 2011).

Agent-based modeling and simulation approach

ABM is the methodological approach used in this research. Over the past two decades, ABM has been increasingly applied as a research approach in various domains and fields, such as macroeconomics (Cincotti, Raberto, and Teglio 2022), management and organizational studies (Ponsiglione et al. 2021), supply chains (Massari and Giannoccaro 2021), and innovation diffusion (Kiesling et al. 2012) among others. ABM has proven to be a powerful methodology for studying Complex Adaptive Systems (Holland 2006), such as, networks, value chains, markets, economies, production clusters, and innovation ecosystems, by representing a group of heterogeneous and autonomous decision-making entities with bounded rationality – referred to as agents – interacting in a virtual shared environment. These interactions are massive and non-linear, allowing the emergence of collective properties, behaviors, and patterns from the bottom-up (Gilbert and Terna 2000).

The advantages in using this approach when studying complex systems mainly rely on the possibility of considering the heterogeneity characterizing actors in the system, with their different attributes, objectives and behaviors, thus reducing the need to make simplifying assumptions about reality (Gallegati and Kirman 2012). Furthermore, an ABM allows to reproduce in a virtual environment the interactions among the heterogeneous and autonomous (without a top-down control) agents, observing what happens at macro-level in terms of emergence of macroscopic regularities that cannot be known a priori decomposing the system in its constituting parts. In fact, agents dynamically evolve by adapting to the emergent structures or events they have produced by interacting with other agents and with the environment (Gallegati and Kirman 2012). Agent-based models incorporate both micro-level behavioral rules and macro-level emergent properties of the system under investigation, thus making possible to build computational laboratories to conduct virtual generative experiments (Epstein 1999). This approach can be used for different purposes, like theory development, testing, or illustration (Edmonds et al. 2019). More recently, ABM has also received increased recognition as a tool supporting policy development (Bruch and Atwell 2015). What makes useful ABM for this aim is mainly the possibility to take into account the heterogeneity of agents and their dynamic adaptation, accounting for the uncertainty and the role of randomness (Edmonds and ní Aodha 2019).

According with the premise above, ABM is adopted in this paper to study the dynamic evolution of IVCs in industries adopting additive manufacturing as enabling technology and considering openings to RRI. The reasons for using this specific approach are related to the need of considering the heterogeneity of agents populating the system under investigation, their non-linear and massive interactions, and the uncertainty of the research and innovation journey they are involved in (Vermeulen and Pyka 2016). Furthermore, ABM provides us with a ‘rich and detailed account of the process of a system’s unfolding in time, and not just the final state of the system’ (Wilensky and Rand 2015, 36). One key requirement of our model is that it should capture the dynamics of the collaborative networks in IVCs through time. The ABM method fits well with this requirement.

The model proposed in this paper has not been designed to replicate a specific real-world case, but aims to represent a broader class of cases, and as such, can be classified as a typification (Boero and Squazzoni 2005). This implies that, at this stage, the agent-based model requires an intermediate level of reference to empirical world and cannot be used to design specific ad hoc policy measures. At the same time, the model could be used for theory development or testing, and could be enriched and extended in the near future, extending also the adoption of empirical driven data for calibration, in order to support policy advice.

Research context

The proposed agent-based model was developed within the I AM RRI project (Cozzoni et al. 2021) to simulate and visualize the behavior of IVCs and to foster a general understanding of potential RRI openings. The additive manufacturing industry was chosen as a reference context to develop our model, given the consortium’s industry knowledge of components deepened through literature research and use cases.

The additive manufacturing industry is characterized by its inherent complexity, stemming from various factors (Rehnberg and Ponte 2018). The supply chain complexities in additive manufacturing, including raw material sourcing, post-processing, and quality control, pose significant challenges (Laplume, Petersen, and Pearce 2016). In the social sciences, the management of the entire additive manufacturing workflow has gained increasing interest, with a focus on networking (Johns 2022; Rehnberg and Ponte 2018) and socioeconomic impact (Birtchnell and Urry 2016; Huang et al. 2013). In addition, the additive manufacturing industry is characterized by intricate networking and interrelationships among various actors, including manufacturers, suppliers, researchers, and regulatory bodies (Johns 2022). Navigating these connections and understanding their impacts on the surrounding environment, such as sustainability concerns and regulatory compliance, adds another layer of complexity to the industry’s operations (Huang et al. 2013).

In summary, the complexity of the additive manufacturing industry stems from the heterogeneity of actors involved, its interconnection and dynamic nature, and its interaction with several external factors. These characteristics make it a Complex Adaptive System (Holland 2006) in which the ability to adapt and respond to change is critical for industry’s survive.

The proposed model

The model here proposed has been originally developed under the context depicted above and has been implemented in the programming language NetLogo² (v.6.1.1.). It can be consulted and downloaded from Github.³

It is a ‘double-industry’ model in which agents operating through additive manufacturing (AM tech companies, Suppliers, Customers, Research Institutions, and OEMs) can belong to the Automotive Industry, the Biomedical applications industry, or both, and the IVCs evolve in different phases (Idea Generation and Product Development⁴), each with a different duration. Agents’ inclination to ethical values is modeled through four endogenous variables representing three thematic areas of RRI identified by the European Commission (2012): public engagement, open-access, ethical thinking, and gender equality. These RRI keys have been chosen (among six different keys) considering the possibility to more easily collect data to measure them and for the availability of an operationalization provided by the context of the I AM RRI project. These RRI characteristics influence agents’ decisions, the IVCs’ performance in which they participate, and their way of interacting with the environment and other actors. The characterization of knowledge endowment of the agents in the system and of some learning and diffusion mechanisms have been reproduced according to the SKIN model (Gilbert, Ahrweiler, and Pyka 2014), a well-known agent-based model designed to analyze the dynamic evolution of innovation networks in knowledge intensive industries. The SKIN model has been chosen as a base to develop the original model in I AM RRI project, because it has been considered a reliable and adequately diffused model in the scientific community to represent a robust core to be further adapted in this research, as happened in other studies (Ahrweiler, Frank, and Gilbert 2019a). The name chosen for our model reflects this link with SKIN and is I AM RRI SKIN. The key component taken from SKIN is the agents’ knowledge base, called the Kene. The Kene is made up of a set of triples, each consisting of a Capability (broad domain of technological or scientific knowledge), Ability (a more specific skill in the knowledge domain), and an Expertise (level of experience associated with the Capability-Ability pair). Starting from its knowledge, each agent elaborates and extracts a fuzzy innovation hypothesis that must be refined through cooperation with partners. Agents’ knowledge is a function of the agent’s economic resources, the agent type, and the industrial sector to which they belong (automotive or biomedical). Agents’ strategies, e.g. choice of partners, open-access publication, or willingness to stay within a network, are not only cost-oriented but also RRI-oriented.

Once the various networks are established, the consortium members proceed through the innovative stage-gate process in which the process evaluation becomes predictive and not just reflective. The evaluation is carried out by the regulatory, standardization, and funding bodies who, in addition to assessing the technical quality of the idea, consider the RRI inclination of the networks. The process consists of two phases: Idea Generation (duration 3 ticks – the temporal unit of the simulation cycle) and Product Development (duration 12 ticks). During both phases, the mechanisms of diffusion of knowledge and RRI values among the networks members are manifested, taking into account the inertia to change in the organizations. At the simulation end, after the evaluation by the regulatory bodies or the standard organizations, the networks can concretize in a start-up or dissolve. The outputs of the simulation are divided into three macro-areas: social (e.g. the trend or the average of RRI values), economic (the average capital of the networks), and strategic (e.g. the number of agents involved in the networks, the percentage of agents that survive each phase, the number of start-ups created).

In this paper, as mentioned above, we are particularly focused on considering the role of NGOs in IVCs’ evolution and in diffusing RRI keys. As a consequence, the original I AM RRI SKIN model has been further developed introducing the new actors in the system. According to what depicted in theoretical sections of this paper, we consider this new type of agent as provider of valuable data set for eventual partners. In fact, we recall that the major contribution NGOs make in a project is connected to their ability in providing data and knowledge. NGOs are also specialized entities in sourcing of raw materials/software/goods needed for project purposes. In addition, they are by nature experts in public engagement, know the needs of society very well and cooperate with large banks and states to obtain funds for social purposes (Ahrweiler, Frank, and Gilbert 2019a; Aldashev and Navarra 2018; Fisher 1997; Szarka 2013).

Unlike other agents, the knowledge possessed by the NGOs was not modeled as a function of belonging to the Automotive or Biomedical sector since no empirical difference was found; however, the modeling through the Kene and the extraction of the Innovation Hypothesis remains unchanged.

An NGO may initiate an innovative project or join an already established network, increasing capital and knowledge as a result of cooperation in a successful network.

An NGO pursues projects with a social purpose; therefore, it only decides to join a network if it develops a socially-oriented idea. To model this aspect, we decided to use the RRI keys as a proxy for the network’s sensitivity to social issues. Therefore, at the simulation beginning, the values of the network RRI keys are calculated through the average of the participant’s RRI values. Ultimately, an NGO participates in an innovative network if it has adequate RRI levels that exceed a threshold value that the user can also choose.

Moreover, Ahrweiler, Frank, and Gilbert (2019a) show that in a network, if present, the NGOs are not less than two.

In the first case that could be configured, an already established network asks an NGO to join the innovative project, which only agrees after assessing the RRI inclination level of the consortium. If this assessment is successful, the importance given in the consortium to public engagement and open access is increased.

In a second case, it is an NGO that initiates an innovative project and searches for potential partners while the mechanisms for updating RRI values and disseminating knowledge remain the same. If a network manages to include only one NGO, not reaching the minimum of two, it dissolves.

The Figure 2 schematically represents the NGO selection process.

Figure 2. Find NGO partners.

Simulations and results

RRI is considered a strategy that pursues long-term goals, often ‘elusive and difficult to measure’ (Pansera et al. 2020, 409), affecting the whole innovative process and indirectly the product of the process (Cozzoni et al. 2021; Van den Hoven 2013). To answer the research questions, we evaluated the performance of the innovation ecosystem through three main indicators:

• RRI spreading.

• heterogeneity of innovative networks.

• knowledge dissemination.

The spread of RRI values was assessed through the average RRI values reached by the generated networks. While heterogeneity was modeled as the ratio of the number of different breeds (class of agents) within the network to the total number of members. (1) $Heterogeneity={\displaystyle \frac{{\sum }_{breed\in Network}breed}{numberofpartners}}$(1)

Starting from the definition of Kene (Gilbert, Ahrweiler, and Pyka 2014), knowledge of the i_th agent (k_i) was represented by the length of its capabilities vector. To evaluate the diffusion of knowledge, we have considered the average knowledge of the agents participating in an innovative project, represented as: (2) $\overline{k}={\displaystyle \frac{1}{nFirms}\sum _{i\in \{partneringFirms\}}{k}_i}$(2)

Then, to evaluate the effect of NGOs presence, an independent variable called NGO-attendance represents the RRI value of a network required by the NGOs to enroll the innovative project.

To test our model, we assume that an increase in the NGO-attendance variable (set at 0.4, 0.7, and 1.0) corresponds to fewer NGOs participating in innovative projects $(F(2,897)=3882,759,p<0.00010,4\to meanNGO=1980,7\to meanNGO=1601,0\to meanNGO=58.).$Table 1 reports the parametrization and the considered control variables (in Table A1 of the Appendix, instead, we reported a synthetic description of all the I AM RRI SKIN variables).

Table 1. Experimental design.

Control Variables
Numbers of Agents		Firm Variables		Environmental Variables
n AM-tech	200	Attractiveness threshold	0.5	Standard organization	5
n Supplier	200	RRI start-up trigger	0.5	Funding	50
		RRI attractiveness	0.5	Regulator	0.5
n Research-inst.	200	Publish open access		Funding RRI	0.5
n NGO	200			Funding quality	5
n OEM	200	Economic threshold	50	RRI cost	30
n Customer	200	RRI open access threshold	0.5	Big firm percent	10
Setting, Explanatory Variable, and Output
Run	Ticks	ngo-attendance	Output
300	30	0.4, 0.7, 1.0	Average RRI values of networks Average heterogeneity of networks Average knowledge of the agents

To evaluate the dynamic evolution of the RRI keys, we conducted 300 simulations, measuring the value of the RRI keys at each step of the simulation for three levels of the NGO-attendance factor. As Figure 3 shows, the average RRI values of the networks increase during the evolution of the innovation process. To assess the statistical significance of the presence of NGOs within an innovative network, we used a One-way ANOVA for steps number 4, 6, and 30, following the Kolmogorov–Smirnov and Shapiro–Wilk normality tests on the collected data. Steps 4 and 6 were chosen because they follow the learning and diffusion mechanisms of the RRI keys, while step 30 represents the final step of the simulation. In all these steps, the NGO-attendance factor produces a statistically significant effect on the RRI values possessed by the networks. Table 2 reports the ANOVA results, considering an α equal to 0.05.

Figure 3. RRI values evolution.

Table 2. ANOVA for RRI values.

One-way ANOVA
	Step 4	Step 6	Step 30
Public Engagement	F(2, 897) = 540.85*	F(2, 897) = 436.86*	F(2, 897) = 710.13*
Gender Equality	F(2, 897) = 344.13*	F(2, 897) = 123.82*	F(2, 897) = 166.84*
Open Access	F(2, 897) = 347.78*	F(2, 897) = 111.03*	F(2, 897) = 163.04*
Ethical Thinking	F(2, 897) = 137.80*	F(2, 897) = 131.80*	F(2, 897) = 118.08*

*p < 0.0001.

Following these results, the analysis was deepened by performing additional post-hoc, such as the Tukey’s HSD Fisher’s LSD, which confirm the statistical difference due to the presence of NGOs in innovative networks. Furthermore, as the descriptive statistics reported in the appendix (Table A2) suggest, the 0.7 level of NGO-attendance represents the value at which a greater diffusion of RRI practices is achieved.

To test the indirect effect of the presence of the NGOs on the diffusion of knowledge of the participants of innovation networks and on the heterogeneity of the working groups, we used a One-way ANOVA by setting the control parameters and modeling the dependent variables as reported in Table 1 and repeating the Kolmogorov–Smirnov and Shapiro–Wilk normality tests. The results of the ANOVA shows that accepting a first kind risk at $\alpha =0.05$, the NGO-attendance produces a statistically significant effect on heterogeneity $(F(2,897)=5,474,p=0.004)$ and increased knowledge of the agents in the model $(F(2,897)=833,65,p<0.0001)$. Specifically, for the highest level of NGO-attendance, we obtain the best average values for both heterogeneity (60%) and knowledge (4.23).

Discussion

From our simulations it emerges that when NGOs adopt intermediate values of RRI in assessing their participation in innovative consortia, they strongly contribute to the dissemination of RRI practices within the ecosystem. At the same time, an overly strict selection strategy based on RRI criteria hinders the effective diffusion of RRI throughout the ecosystem, resulting in suboptimal performance. These results highlight the significance of striking a balance in RRI evaluation strategies for maximizing its impact on innovation and ecosystem-wide diffusion.

Above results find support in several studies who have investigated the role of NGOs in innovation processes. In fact, scholars argue that partners’ selection strategy of NGOs strongly impacts the sustainability and inclusiveness of innovative projects and the diffusion of RRI practices within the innovation ecosystem (Ahrweiler et al. 2019b; De Marchi 2012; Elzen et al. 2011; Hodson and Marvin 2010; Parayil 2003).

Furthermore, the study highlights the complex nature of RRI diffusion within innovation systems. It reveals that an excessively stringent approach to RRI evaluation can inadvertently impede the widespread adoption of responsible practices, limiting the overall potential for positive societal and environmental outcomes. While NGOs play a crucial role in encouraging the adoption of responsible practices (Ahrweiler, Frank, and Gilbert 2019a), their overly stringent approach to RRI evaluation can create unintended consequences and hinder the widespread acceptance of such practices. Indeed, by placing excessive pressure on organizations to meet strict RRI criteria, there is a risk of alienating potential partners and stifling their willingness to engage in innovative projects (Cozzoni et al. 2021).

According to Mampuys and Brom (2015), the main role of NGOs lies in acting as an internal regulatory body, which promptly warns partners of potential technological risks. Through their regulatory pressure, NGOs pursue social goals that push other actors to create a more sustainable and green innovative system (De Marchi 2012). However, as Elzen et al. (2011) state, NGOs lack the normative capacity to fully compel their partners to embrace RRI practices. The management of NGOs should be cautious in adopting an overly rigid approach, as it may overlook valuable contributions and perspectives from stakeholders who might not fully align with the highest standards of RRI but still bring valuable insights and expertise to the table (Gurzawska, Mäkinen, and Brey 2017). It is recommended that the management of NGOs adopt a more flexible and inclusive approach to foster a collaborative and diverse ecosystem of innovation.

NGOs act as initiators and facilitators by orchestrating partners’ efforts, ensuring a right balance of power while making it free the initiative of the partners (Abbott and Snidal 2010; Gurzawska, Mäkinen, and Brey 2017).

Hence, our study highlights the importance of adopting a nuanced and balanced approach to RRI diffusion by NGOs. Encouraging a culture of gradual adoption and continuous improvement can prove more effective in driving long-term change. RRI should be perceived as a collective responsibility, and the establishment of standards should involve co-creation with all stakeholders of RRI (Gurzawska, Mäkinen, and Brey 2017). This approach allows for a progressive integration of RRI while considering the specific contexts and capabilities of different actors within the industry (Yaghmaei 2018).

According to our results policymakers should not predominantly delegate NGOs to represent the general interest and the social perspective but should consider other actors more. Indeed, as demonstrated in the research conducted by Ahrweiler, Frank, and Gilbert (2019a), various actors, including universities and SMEs, effectively function as Civil Society Organizations to secure additional public funding for the promotion of RRI and to foster increased inclusivity and diversity. Several authors measure the performance of innovative networks adopting RRI practices through a non-financial point of view and the achievement of long-term goals (De Saille 2015; Sutcliffe 2011), such as knowledge diffusion and greater heterogeneity of actors involved (Cozzoni et al. 2021). These goals bring benefits to developing new capabilities (Gonzales-Gemio, Cruz-Cázares, and Parmentier 2020), and leading to faster identification of innovative opportunities (Fitjar, Benneworth, and Asheim 2019). NGOs play a relevant role also with respect to this point (Ahrweiler, Frank, and Gilbert 2019a; Gurzawska, Mäkinen, and Brey 2017).

NGOs support knowledge sharing, involving all groups in shaping decisions (Sutcliffe 2011). They drive inclusive business models targeting marginalized communities (Cavallaro et al. 2014) and contribute unique scientific data and expertise, not easily accessible to others (Ahrweiler, Frank, and Gilbert 2019a; Böschen et al. 2020; Lang and Griessler 2015). Their involvement fosters inclusivity, diversity, and better decision-making within innovation networks (Cozzoni et al. 2021; Fitjar, Benneworth, and Asheim 2019; Van den Hoven 2013).

The statistical evidence presented in this study indicates that when NGOs apply more stringent criteria in evaluating networks (resulting in high NGO-attendance), there is an increase in network heterogeneity and knowledge spreading. However, it also leads to a decrease in the number of NGOs participating in the networks. This suggests that the presence of highly RRI-oriented NGOs could potentially act as a constraint on the innovative dynamics of the networks, limiting their freedom and effectiveness.

While the presence of RRI-oriented NGOs contributes to greater knowledge diffusion and heterogeneity, their limited involvement may compromise the overall performance of the ecosystem, particularly in terms of the widespread adoption of RRI practices.

These findings find support in some previous work. For example, Krabbenborg and Mulder (2015) report that the role of Civil Society Organizations as ‘early warners’ in the early stages of the innovation process can cause criticality in the relationship with scientists and technology developers. A strong attachment to RRI practices undermines an open and effective dialogue among stakeholders by endangering the negotiation on the output of the process (Ahrweiler, Frank, and Gilbert 2019a). For this reason, industrial stakeholders could choose their partners by avoiding RRI-related constraints that are too stringent (Cozzoni et al. 2021). Indeed, NGOs could be perceived as acting as regulatory bodies similar to government agencies (Hutter and O’Mahony 2004; Mampuys and Brom 2015), potentially impeding the innovation process and leading other actors to act autonomously, excluding an additional moment of negotiation and simplifying the process. This strategic partner selection allows for a more streamlined and efficient knowledge dissemination process, as the network can prioritize organizations that share a common vision.

Another reason could be attributed to the concentration of specialized knowledge and expertise within the network when fewer NGOs are involved. With fewer organizations to engage and collaborate with, it becomes easier to establish common objectives, align strategies, and streamline decision-making processes (Nosenzo, Quercia, and Sefton 2015). By selectively involving fewer NGOs, the networks can concentrate efforts on specific areas of expertise or target specific social and environmental challenges, leading to a more focused and impactful approach to knowledge diffusion and inclusiveness (Abbott and Snidal 2010). However, with fewer NGOs involved, there is a risk of limited representation and diversity of perspectives within the network (Cavallaro et al. 2014). This can result in a narrower range of viewpoints and a potential lack of critical voices, leading to the possibility of overlooking important societal, ethical, or environmental considerations (Ahrweiler, Frank, and Gilbert 2019a). NGOs often play a crucial role in holding practitioners accountable and ensuring that responsible practices are maintained. However, when their involvement is restricted, the network may experience diminished checks and balances from one stage of IVCs to the next, potentially leading to reduced control and oversight. Consequently, this situation could elevate the risk of unethical or unsustainable practices going unchecked (Gurzawska, Mäkinen, and Brey 2017).

Finding a balance among NGO involvement, knowledge diffusion, and inclusiveness in innovative networks is of utmost importance. While highly RRI-oriented NGOs contribute to knowledge dissemination and network heterogeneity, their limited participation can hinder network effectiveness, potentially impacting ecosystem performance. Our study suggests that NGOs should consider the perspectives of other partners to incentivize the diffusion of RRI practices. Hence, not only NGOs management but also policymakers, in general, must recognize the diverse needs of industrial partners to promote RRI diffusion in industrial contexts. Selecting partners strategically to avoid overly stringent RRI constraints streamlines knowledge dissemination but may overlook critical perspectives and weaken accountability.

Conclusion

Using an ABM allows us to model an innovative ecosystem without neglecting essential elements of reality, such as agent heterogeneity, uncertainty in interactions, and experiential learning. Our work represents an attempt to create an auxiliary tool for Research Funding Organizations to develop guidelines for promoting RRI practices and to facilitate Research Performing Organizations in adopting RRI best practices and managing innovative dynamic networks. Efforts focused on modeling the behavior of NGOs within highly innovative networks and simulating different scenarios to assess the impact of NGOs’ partner selection strategies on the diffusion of RRI practices, knowledge, and heterogeneity in innovative networks. Although there are previous studies on the role of NGOs within the networks, to the best of our knowledge, their behavior has not yet been modeled. The model presented here, developed considering the automotive and biomedical industry characteristics, can be easily adapted to other contexts and industries in which the actors interact by evaluating ethical components.

Our results lead us to formulate a further recommendation for the NGOs’ management figures: a greater moderation in strategic evaluations does not mean a departure from RRI practices but rather a rapprochement to the institutional logic of the remaining actors, favoring the learning of valuable capabilities and inclusiveness. Indeed, striking a balance in the evaluation and involvement of NGOs within innovative networks is essential for maximizing the impact RRI practices on knowledge diffusion and inclusiveness. An excessively strict approach to partner selection based on RRI criteria can impede the effective dissemination of RRI and hinder innovation performance. A nuanced and balanced approach is needed, fostering a gradual adoption of RRI practices and considering the specific contexts and capabilities of different actors within the industry. This approach fosters a culture of continuous improvement and co-creation, enabling the collective establishment of RRI standards and promoting the gradual integration of RRI practices to facilitate long-term transformative change.

NGOs play a crucial role in the dissemination of knowledge and the promotion of inclusion within an industrial context. Their involvement contributes to greater inclusivity, diversity, and better decision-making within innovation networks. However, a careful balance must be maintained to ensure that the concentration of specialized knowledge and limited NGO participation does not lead to a lack of representation, diverse perspectives, and weakened accountability.

Clearly, this work is not without limitations. Although the internal verification and validation of the model with respect to the literature were successful, the lack of longitudinal data on the performance of innovative networks does not allow us to compare our simulations directly with empirical data; this limitation opens to future studies that could use empirical data collected in other industrial contexts or particular projects funded through the Horizon Europe program. Furthermore, the paper focuses specifically on the diffusion of RRI practices and the role of NGOs in knowledge dissemination and inclusiveness within networks of IVCs in the context of Industry 4.0 technologies. This narrow focus may limit the generalizability of the findings to other domains or contexts.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 IAMRRI. Webs of Innovation Value Chains of Additive Manufacturing under Consideration of RRI. Available online: https://www.iamrri.eu/ (accessed on April 27, 2023).

2 https://ccl.northwestern.edu/netlogo/ (accessed on July 25, 2022).

3 Github I AM RRI SKIN link: https://github.com/GradoZeroTeam/IAMRRI.

4 The conceptual model includes also an Innovation Diffusion phase, but it has not been developed further at this stage of the research.

Appendix

Table A1. Description variables I AM RRI SKIN model.

nAM-tech	the number of AM-tech companies initially
nSupplier	the number of Suppliers initially
nResearch-inst	the number of Research institutions initially
nOEM	the number of OEMs initially
nCustomer	the number of Customers initially
nNGO	the number of NGOS initially
big-firm-percent	number of firms with 10 times initial capital of rest
attractiveness-threshold	how attractive a firm must be before it becomes a partner
regulator	requirements of regulators to be filled in (depends on ethical thinking)
standard-organization	requirements of standard organization (depends on idea quality)
total-funding-amount	how many funds are available at the start of the simulation
funding	funding for deserving network
funding-quality	quality required to receive funding
funding-RRI	RRI threshold required to receive funding
RRI-start-up-trigger	RRI value threshold that triggers start up
RRI-cost	cost to support ethical values
economic-threshold	economic threshold to be exceeded in order to publish open access
RRI-open-access-thresholds	RRI open access value threshold to be exceeded in order to publish open access
initial-capital	the capital that a firm starts with (except research-inst)
initial-capital-for-big-firms	the amount of start-up capital for those firms set to be ‘big’
max-ih-length	maximum length of an innovation hypothesis
incr-research-tax	tax paid for one step of incremental research
refining-IH-cost	tax paid for every partnership per step
min-partners	minimum number of partners
net-phase-1	networks in idea generation phase
net-phase-2	networks in product development phase
net-pahse-3	networks in market diffusion phase
%survived-1	Percentage of survived firms in idea generation
%survived-2	Percentage of survived firms in product development
industry	sector in which the organization operates (Medical, Automotive, Broker)
initial capital/dimension	Agents’ Initial resources or the dimension of Research-insts
start-project?	The agent is able to start and drive an innovative project
state	i internal agents’ timing
open-access	Agents’ inclination to the Open-access
public-engagement	Agents’ inclination to the Public engagement
gender-equality	Agents’ inclination to the Gender Equality
ethical-thinking	Agents’ inclination to the Ethical-thinking
RRI-value	List of open-access, public engagement and ethical-thinking values
agent-public-open-access	number of agents that made an open-access publication
max-partners	Maximum number of partners that each agent can have
partners	List of agents that are in the same network
previous-partners	List of agents that were in the same network
cap-capacity	maximum Kene length
IH	Innovation Hypothesis
capabilities	Agents’ Capabilities
max-veiled	maximum number of Capabilities not possessed by the agent, but necessary
abilities	Agents’ Abilities
expertises	Agents’ Expertise
advert	Agent capabilities list
RRI-net	RRI values of network
net-partners	list of partners
max-partners-same-breed	maximum number of partners of the same breed
who-net-partners	list of ID partners
net-dimension	number of partners in the network
focal	the network’s agent creator (start-project)
invested-capital	sum of the capitals invested in the project by the partners
phase	phase of the evolution in which the network is working.
quality	quality of network’s product

Table A2. Descriptive statistics RRI values diffusion.

Mean RRI Value
	Step 4			Step 6			Step 30
ngo-attendance	0.4	0.7	1.0	0.4	0.7	1.0	0.4	0.7	1.0
Public Engagement	0.61	0.66	0.58	0.69	0.71	0.63	0.88	0.86	0.78
Gender Equality	0.58	0.63	0.60	0.61	0.65	0.64	0.71	0.75	0.75
Open Access	0.58	0.63	0.58	0.64	0.67	0.64	0.83	0.84	0.80
Ethical Thinking	0.64	0.67	0.65	0.65	0.68	0.67	0.71	0.74	0.73