Responsible innovation ecosystem governance: socio-technical integration research for systems-level capacity building

ABSTRACT

Calls for a ‘systemic turn’ in Responsible Innovation and Responsible Research and Innovation (R(R)I) stem from unease with engagement research. Engagement research structures science-society interactions to align research and innovation processes with societal considerations. As this research often focuses on micro practices of actors at discrete events or in bounded environments, it tends to neglect the systemic nature of these processes. We introduce the innovation ecosystem concept to account for complexity, openness, and mutual learning in responsible innovation governance. Responsible innovation ecosystem governance refers here to the capacity of diverse actors to reflect on socio-ethical horizons in different streams of the ecosystem. For systems-level capacity building, we discuss three adaptations of Socio-Technical Integration Research (STIR): STIR workshops curated by engagement agents, multi-stream engagement through brokers, and a multi-method research design addressing networked responsibility. This methodological design introduces an ecosystem perspective in R(R)I engagement research.

KEYWORDS:

• Responsible innovation
• innovation ecosystem
• socio-technical integration
• reflexive governance

Introduction

Engagement research on the relationship between science, technology, and society has gained considerable support from funding agencies, regulatory bodies, and ethics commissions, especially in Europe and North America (Von Schomberg and Hankins 2019). Under the labels Responsible Innovation and Responsible Research and Innovation, here summarized as R(R)I, a number of approaches, methods, and practices have been developed with an ongoing focus on public and stakeholder involvement as well as reflexive learning among technoscientific experts (Fisher et al. 2015; Reijers et al. 2018; Schuijff and Dijkstra 2020). Engagement research in R(R)I has philosophical roots in applied ethics, where recent approaches embed ethicists within sites of science and technology development to encourage ethics deliberations (Van Den Hoven, Miller, and Pogge 2017; Van Der Burg and Swierstra 2013). Similar approaches have been developed in the ‘engaged program’ of Science & Technology Studies (STS) (Sismondo 2008), which combines the analysis of science and technology as socio-cultural practices with an ambition to actively participate in democratizing these practices. The overarching aim is to render processes of research and technology development more sensitive to societal, ethical, and environmental concerns (Caswill and Shove 2000; Downey and Zuiderent-Jerak 2021; Gjefsen and Fisher 2014; Zuiderent-Jerak and Jensen 2007).

Engagement thus appears as an umbrella concept for multiple types of participatory research. Public and stakeholder participation can take the forms of public and citizen consultations (Chalmers et al. 2014; Ketzer et al. 2020), deliberative mini publics (Capurro et al. 2015), and stakeholder scenario workshops (Roßmann 2021; Schulz-Schaeffer and Meister 2017) to explore different actors’ prospective views on emerging science and technologies in domains like energy (Groves, Sankar, and John Thomas 2018), transportation (Cohen, Stilgoe, and Cavoli 2018), and healthcare (Decker et al. 2017). Engagement with technoscientific experts and industry actors often seeks to stimulate reflexive learning in long-term interdisciplinary collaboration (Aicardi, Reinsborough, and Rose 2017; Balmer, Bulpin, and Molyneux-Hodgson 2016; Fisher et al. 2015), workshops (Lehoux et al. 2020; Long et al. 2020; Wickson, Strand, and Kjølberg 2015), and card-based discussion groups (Felt, Fochler, and Sigl 2018; Urquhart and Craigon 2021). While it has become commonplace to assert that such engagement research has contributed to a ‘participatory turn’ in science governance, we now observe a shift in the literature involving increasingly louder calls for a ‘systemic turn’ (Aris and Willems 2023; Braun and Könninger 2018; Chilvers, Pallett, and Hargreaves 2018; Nieminen and Ikonen 2020).

These calls stem from an unease with the common design choices in engagement research. As individual actors or groups of actors tend to be the target of engagement, their interconnections with actors who indirectly shape their activities, responsibilities, and rooms for maneuvering receive limited attention (Timmermans et al. 2017). Examples of such indirect influences include policy makers setting the framework conditions for a research project and corporate competitors who influence market dynamics around a novel technology. Moreover, the context of engagement is often defined by the boundaries of a group or organization (Dabars and Dwyer 2022; Long et al. 2020; Pansera et al. 2020; Schuurbiers 2011), which ignores that actors move across contexts. Scientists, for example, do not only work in the university or company laboratory, but participate in conferences, international consortia, and policy advice (Keating, Cambrosio, and Mackenzie 1992). In engagement activities not restricted to bounded environments but taking place in multi-stakeholder events, the emphasis is often on single one-off events, which does not allow for attending to the evolving nature of research and development over time (Chilvers and Longhurst 2016; Chilvers, Pallett, and Hargreaves 2018). Lastly, engagement studies with an analytical focus on micro practices (Fisher 2019; Shilton 2012) risk neglecting emergent phenomena, ripple effects, and indirect impacts on meso- and macro-levels of research and innovation. In other words, they fall short of systemic complexities of research and innovation involving interdependencies between actors with distributed and oft-asymmetrical agencies (Ceicyte and Petraite 2017, 2018).

In response to these shortcomings, scholars introduced the ecosystem concept. Braun and Könninger (2018) identify the eco-systemic approach as one among several others that aim to account for the complexities of public participation. They consider Chilvers, Pallett, and Hargreaves' (2018) ‘ecologies of participation’ as one such approach. The term ‘ecology’ emphasizes the interdependencies between practices, technologies, and political settings of participation. Stahl (2021) suggests that the ethics of Artificial Intelligence (AI) could be understood from an ecosystem perspective to derive recommendations for handling ethical dilemmas. Using the ethics of AI as a case study in a later article (Stahl 2022), he proposes that literature on innovation ecosystems in management and organization studies, on the one hand, and R(R)I discourses and practices, on the other, could mutually inform each other. R(R)I could shift innovation ecosystem management aiming at economic success toward responsible governance integrating ethical and social awareness into the system. In turn, innovation ecosystem literature could help R(R)I scholars consider the ‘systems character of research and innovation and how to deal with it’ (9). In the light of the abovementioned methodological limitations of existing R(R)I engagement studies, Stahl calls for the development of governance approaches to create ‘responsible innovation ecosystems’ (1).

Building on the emerging ecosystem perspective in R(R)I, this article develops a methodological approach to engagement research for realizing responsible innovation ecosystem governance in practice. We consider the innovation ecosystem as a conceptual lens guiding methodological development. This lens helps us to recognize that interconnectedness between actors increases complexity, which can give rise to unanticipated system-level effects as actors in different ecosystem streams participate in steering the system through mutual learning (Autio and Thomas 2014; Moore 1993). While the ‘innovation system’ emphasizes the heterogeneity of actors and the micro-, meso-, or macro-constellations they may form (Nieminen and Ikonen 2020), the innovation ecosystem concept deployed here focuses on mutual dependencies between various innovation-related actors from companies, universities, research institutes, and other organizations at the meso-level of research and innovation (Adner 2017; Carayannis and Campbell 2009). We suppose that in such a meso-level system – larger than an organizational entity but smaller than a national innovation system – it is feasible for a group of social scholars to contribute to innovation governance through engagement research activities. Furthermore, the innovation ecosystem concept speaks to policymakers, funding agencies, industrial representatives, and city officials worldwide who often present the import of the Silicon Valley innovation ecosystem model as a regional boost to economic growth and increased quality of life (Gonçalves et al. 2020; Piqué, Jasmina Berbegal-Mirabent, and Etzkowitz 2018; Reynolds and Uygun 2018; Van Agtmael and Bakker 2016). To these parties, engagement activities that rest on the ecosystem concept may be attractive because these activities promise to provide support in replicating a ‘best-practice model’ (Pfotenhauer and Jasanoff 2017, 419) of innovation locally. Although this article can inform decision-makers in designing innovation ecosystems with a view to societal considerations, it is primarily written for an audience of R(R)I scholars and adjacent research communities.

To develop a practical approach to responsible innovation ecosystem governance, we develop methodological adaptations of Socio-Technical Integration Research (STIR). STIR is an engagement research method that approaches responsible governance as a ‘capacity’ (Fisher 2007) of scientists, engineers, and other relevant actors to integrate the societal dimensions of science and technology development into their work. As actors’ capacities are shaped by wider political, institutional, and material structures, STIR assesses these capacities within their structural conditions. For this purpose, it usually embeds a social scholar for 12 weeks in a group of technoscientific experts to facilitate reflexive dialogues with selected members and to analyze the ways in which actors take societal considerations into account in day-to-day practices. In light of its focus on individual actors, bounded environments, and micro practices, STIR can be regarded as an example of an engagement method neglecting the systemic nature of research and innovation (Krabbenborg 2013). In this article, we ask: how can STIR be adapted to support responsible innovation ecosystem governance? In answering this question, we develop more general recommendations for adjusting R(R)I engagement methods to the governance of innovation ecosystems.

To render STIR applicable to innovation ecosystems, we first review literature in management and organization studies to define the innovation ecosystem concept and elaborate on the concept’s implications for the development of governance approaches. Next, we conceptualize governance by drawing on literature in transformation and transition research, which has taken up the challenge to study and contribute to changes of wider socio-technical systems (Heyen and Brohmann 2017; Köhler et al. 2019). More specifically, we use the concept of ‘reflexive governance’ (Voß, Bauknecht, and Kemp 2006) to introduce an ecosystem perspective into R(R)I engagement research. After grounding responsible innovation ecosystem governance theoretically, we propose three methodological adaptations of STIR: STIR workshops curated by engagement agents, multi-stream engagement through brokers, and a multi-method research design that complements STIR with stakeholder research to address networked responsibility. These adaptations allow us to create a research design which induces reflexive learning processes among actors across an innovation ecosystem. By accounting for complex interdependencies, co-responsibilities, and distributed agencies in an open system, this design promises to build systems-level capacity for responsible innovation ecosystem governance. We finish by providing recommendations for advancing an eco-systemic turn in R(R)I engagement research.

Innovation ecosystems in management and organizational studies

Policymakers, industrial representatives, and city officials worldwide often present innovation as the key ingredient for economic growth and increased quality of life (Pfotenhauer, Juhl, and Aarden 2019). An iteration of the promising tale of innovation revolves around innovation ecosystems, ranging from Silicon Valley in the United States (Piqué, Jasmina Berbegal-Mirabent, and Etzkowitz 2018) and Porto Alegre in Brazil (Gonçalves et al. 2020) to the ‘Innovation Valley’ (ZRR 2021, 13) emerging in the Rhenish area in Germany. Perceiving Silicon Valley as the best-practice model of innovation ecosystems, governments, universities, and companies have made attempts to replicate it, for instance, in Silicon Fen in the United Kingdom and Silicon Plateau in India (Pfotenhauer and Jasanoff 2017). In management and organizational studies, this best-practice model is generally described as a successful entrepreneurial region where high-tech firms and their subsidiaries closely collaborate with universities to transform scientific ideas into marketable products and services (Cheyre, Kowalski, and Veloso 2015; Engel 2015; Ooms et al. 2015). Management and organizational scholars acknowledge that the Silicon Valley model is one among multiple types of innovation ecosystems. These types have been distinguished from each other based on their industry (high-tech, middle-tech, low-tech), spatial scope (regional, national, transnational), and innovation (disruptive, radical, incremental), to name but a few examples of typological criteria (Klimas and Czakon 2021). Across the board, the term ‘innovation’ bears positive connotations – the association between innovation and commercial technologies is seen as profitable and desirable.

For companies, regions, and universities to get at the forefront of innovation and actualize its positive effects, the innovation ecosystem literature suggests that a broader support structure is needed. With governments providing funding for such support structures to be developed, innovation ecosystems appear to be purposefully designed rather than naturally grown (Oh et al. 2016; Sun et al. 2019). This begs the question of what the prefix ‘eco’ signifies. Reviews on the use of the innovation ecosystem concept (Gomes et al. 2018; Klimas and Czakon 2021; Valkokari 2015) trace the ecological metaphor back to Moore (1993). Moore suggests to view companies as members of an ecosystem cooperating and competing around a new innovation. These ecosystems can thrive or perish in society as a whole. He argues that ecosystems – both in the business and in the natural world – develop in ‘evolutionary stages’ (76). Accordingly, phase models and life cycle analyses of innovation ecosystems have sought to explain how ecosystems evolve (Cantner et al. 2020; Dos Santos, Zen, and Bittencourt 2021; Rabelo and Bernus 2015).

In management and organizational studies, a further motivation to use the ecosystem concept has been to understand economic activities in terms of the self-organizing properties of natural ecosystems (Valkokari 2015). Along these lines, researchers draw on complex systems theoretical approaches. They propose that innovation ecosystems could be portrayed as complex adaptive systems (Ritala and Almpanopoulou 2017; see also Peltoniemi 2006). In these systems, agency is typically not located at the level of an individual but at the level of complex interactions between organizations or actors. Through interaction, actors participate in mutual learning, which helps the system adapt to a changing environment (Nylund, Ferras-Hernandez, and Braun 2019). This means that interactions between actors shape the innovation ecosystem as a whole. If these interactions form a ‘managed network’ (Aarikka-Stenroos and Ritala 2017), actors are tied together around a central node (Adner 2006, 2017; Gobble 2014): a key actor or a focal firm playing a major role in steering these interactions and shaping the system.

Pitfalls and opportunities of the innovation ecosystem concept

There are several aspects of the innovation ecosystem concept that ask for critical scrutiny. First, the positive framing of ‘innovation’ obscures its darker sides: the aspiration toward economic growth in technological innovation is said to be at odds with the urgency to solve today’s societal and environmental crises (De Saille et al. 2020); innovation tends to be of benefit to some while putting others at a disadvantage (Jasanoff 2016; Laurent 2019); the modern obsession with innovation leads to the neglect of essential activities of maintenance and care (Vinsel and Russell 2020). Second, the analogy between natural and innovation ecosystems hides the abovementioned politics of innovation behind a veil of seemingly ‘natural’ evolutionary processes (Oh et al. 2016). The reference to the mechanism of natural selection through ‘survival of the fittest’ (Moore 1993, 86) makes the innovation ecosystem concept vulnerable to misuse for justifying the outcomes of innovation processes. Such misuse can have abhorrent consequences, as demonstrated by the history of social Darwinism (Claeys 2000). Third, the concept of a system may evoke an association with cybernetic models of circular causal processes to be controlled through strategic management (Merkel, Brückner, and Wagener 2016). This association raises concerns with centralized control and risks neglecting the complexity and non-linearity of interactions in an innovation ecosystem.

Considering the pitfalls of the innovation ecosystem concept, one may question why we did not replace the concept with an alternative. Chilvers, Pallett, and Hargreaves (2018), for instance, introduce ‘ecologies of participation’ to move away from bounded and reductionist understandings of socio-technical systems. However, we refrain from replacing the concept because the innovation ecosystem literature comprises a key insight for this article: innovation ecosystems are recognized to be subject to change and intervention. As evolution depends on the interdependencies between choices of all actors in the system, management and organization scholars use the terms ‘governance’ (Dos Santos, Zen, and Bittencourt 2021), ‘orchestration’ (Dhanaraj and Parkhe 2006), or ‘choreography’ (Ferraro and Iovanella 2015) to develop strategies for aligning incentives, deciding on positions and roles, and guiding mutual learning. Most of the management and organizational literature considers economic success as the end of innovation ecosystem governance (Adner 2006; Autio and Thomas 2014; Gobble 2014; Provan and Kenis 2008). Only a few studies relate innovation ecosystem governance to ethical and social concerns (Carayannis et al. 2021; Dos Santos, Zen, and Bittencourt 2021; Zygiaris 2013). In reference to the ‘quadruple helix model of innovation’ (Etzkowitz and Leydesdorff 2000; see also Carayannis and Campbell 2009), Dos Santos, Zen, and Bittencourt (2021) emphasize the importance of formalizing the participation of civil society in interactions with the actors of the triple helix – academic institutions, businesses, and governmental agencies – to better understand and integrate citizens’ needs. We build on this proposal by combining the innovation ecosystem literature with approaches to the responsible governance of science and technology in R(R)I.

For this purpose, we define general features of the innovation ecosystem concept. Whereas innovation ecosystems as empirical phenomena are multiple in nature with different actor configurations, levels of maturity, geographies, and types of innovation (Klimas and Czakon 2021; Tsujimoto et al. 2018), we use the innovation ecosystem as an abstract concept with general features applying across empirical cases. Accordingly, we deploy the innovation ecosystem concept as a lens that guides our view in analyzing and developing governance practices. Our lens is triangular-shaped. This means that we define three conceptual features by drawing on our review of the innovation ecosystem literature in management and organizational studies. In this way, we set the framework conditions for adapting R(R)I engagement methods to innovation ecosystem contexts.

We consider the following features as having implications for such methodological adaptations: complexity, openness, and mutual learning. A complex system consists of interdependent, non-linear relations between multiple actors whose interactions can have unexpected effects at the systems level. For governance approaches to monitor and flexibly adjust to such effects, they cannot only target the micro practices of individual actors or groups but must adopt a multi-level perspective. Furthermore, we think of innovation ecosystems as open rather than closed systems. As innovation ecosystems are subject to change, their boundaries are continuously redefined in a process of negotiation between policy programs, local visions, diverging definitions of a technology, as well as the aims of the researcher using the innovation ecosystem concept analytically (Stahl 2021). Ongoing research on ecosystem boundaries supports the identification of moving targets of governance. The evolving boundaries and nature of the system also stem from the mutual learning of actors whose decisions both provoke and respond to changes in the system. Through guidance of mutual learning in innovation ecosystems, decentralized governance practices can be aligned with R(R)I.

Responsible governance through systems-level capacity building

The abbreviation R(R)I binds together Responsible Innovation, an intellectual movement with academic roots (Brundage and Guston 2019), and Responsible Research and Innovation, a public policy discourse originating in the European Commission’s Science in Society program (Owen, Macnaghten, and Stilgoe 2012). As the meanings of and boundaries between those labels have remained contested, we follow Smolka (2020) and Shanley (2021) in using R(R)I to refer to the community as a whole. The community is founded on the observation that innovation is not beneficial for everyone, everywhere, all the time. For example, praised Silicon Valley innovations like the smartphone and mobile apps are accompanied by privacy concerns (Lai and Flensburg 2020), environmental impacts (Laser 2020), psychological distress (Dai, Tai, and Ni 2021), hidden labor (Gray and Suri 2019), and adverse effects on social relations (Albury et al. 2017). Against the backdrop of a history of scholarly work in STS criticizing innovation policies for neglecting societal considerations (Shanley 2022), R(R)I calls for shifting the focus of attention from risk assessment and regulation to the social shaping of research and innovation processes through ‘innovation-governance’ (Felt et al. 2007, 10). This form of governance distributes responsibility for the processes and outcomes of science and technology development across governments, researchers, innovators, other stakeholders, and citizens.

R(R)I can be considered as an extension of a broader set of discourses and practices related to the governance of science and technology development. They include but are not limited to applied ethics (Van Der Burg and Swierstra 2013), ELSI and ELSA programs (Balmer et al. 2016; Zwart, Landeweerd, and Van Rooij 2014), technology assessment (Rip and Robinson 2013; Guston and Sarewitz 2002), anticipatory governance (Barben et al. 2008; Guston 2014), and reflexive governance (Voß, Bauknecht, and Kemp 2006; Voß, Smith, and Grin 2009). In the three last-mentioned approaches, governance has been conceptualized as a ‘capacity’ (Guston 2014, 226; Guston and Sarewitz 2002, 95; Voß and Kemp 2006, 16) of actors – authorizers, sponsors, scientists, engineers, and other stakeholders – to reflect on their roles and responsibilities in attending to the societal dimensions of science and technology development. We focus here specifically on reflexive governance because it explicitly takes a systems character of this capacity into view.

Reflexive governance is rooted in transformation and transition research (Elzen, Geels, and Green 2005; Heyen and Brohmann 2017; Köhler et al. 2019), where it is a general approach to govern sustainable transitions subsuming transition management and strategic niche management (Voß and Bornemann 2011). The governance challenge lies in transforming socio-technical systems, for instance, to adopt ‘smart’ agendas for mobility and urban development (Köhler et al. 2019, 15). Reflexive governance sees the capacity to influence the transformation of socio-technical systems as being distributed between different but interdependent levels (individual, organization, network, sector), arenas (academia, public discourse, companies, policymaking), and a broad range of actors. By focusing on systems, it directs attention to how individuals interact to organize themselves and thus participate in governing the emerging organizational patterns. If they become aware of their role, position, and participation in governance, they can take responsibility for co-steering the system in a way and direction that is responsive to societal concerns (Rip 2006). This governance is reflexive in two senses. First, it is carried out by actors who recognize themselves as having the capacity to govern the system of which they are part (Smith and Stirling 2007). Second, such reflexive actors realize that participation in governance requires them to handle diverging viewpoints, distributed agencies, and unintended effects of actions within a complex system. In other words, they have the capacity to reflexively engage with the dynamics and conditions of governing the system in which they are embedded (Voß and Kemp 2006).

By capacity we mean that actors are generally able and competent to act in particular ways and that they have the human qualities to do so (Kroesen, Darson, and Ndegwah 2015). We introduce the concept of ‘systems-level capacity’ to acknowledge that responsible innovation ecosystem governance cannot be done by an isolated team of managers, but involves actors in the entire ecosystem (Garst et al. 2022; Radatz et al. 2019). If these actors become aware of the ways in which they shape the research system, they develop ‘reflexivity as an institutional capacity’ (Rip 2006, 89). According to Grimpe et al. (2020), reflexivity as an institutional capacity can be cultivated through what they coin ‘collaborative reflexivity’ (720). Collaborative reflexivity is an interactional practical achievement of actors working in an organization. The authors analyze how funding managers enact collaborative reflexivity regarding science-policy-society relations through ordinary activities, such as telephone conversations and meetings with colleagues, researchers, and stakeholders. Based on the analysis, the authors recommend an organization wishing to enhance institutional reflexivity to open up learning spaces, for instance in workshop settings, where collaborative reflexivity can be cultivated. We propose that opening up such learning spaces at multiple governance levels, in different arenas, and within diverse actor groups can build systems-level capacity for the social steering of science and technology in innovation ecosystems. For this purpose, we develop methodological adaptations of STIR that put responsible innovation ecosystem governance into practice.

Methodological adaptations of STIR

STIR is an R(R)I engagement research method that approaches the responsible governance of science and technology development as a capacity for socio-technical integration, that is ‘any process by which technical experts account for the societal dimensions of their work as an integral part of this work’ (Fisher and Maricle 2014, 74). To study and cultivate this capacity, STIR embeds a social science or humanities scholar, also referred to as an ‘embedded humanist’ (Fisher and Mahajan 2010, 216), into a technoscientific environment to conduct dialogues that structure reflection on the societal dimensions of expert decision-making. The embedded humanist deploys a decision protocol that structures dialogic interaction with a technoscientific expert in four conceptual components: opportunity, considerations, alternatives, and outcomes. The protocol treats individual decisions as the unit of analysis for documenting, probing, and assessing how experts can, do, and could respond to a variety of political, social, material, and other considerations. STIR dialogues usually take place in the ‘midstream’ of a project, when decisions are neither fully determined by ‘upstream’ agendas, nor limited to an instrumental ‘downstream’ approach to their implementation (Fisher, Mahajan, and Mitcham 2006, 490). By intervening in the midstream of research and innovation, STIR supports experts in becoming reflexively aware of socio-technical integration in routine decision-making. This can result in reflexive and practical modulations in how experts account for societal dimensions in technoscientific workflows (Fisher and Schuurbiers 2013).

STIR studies have mostly taken place in well-defined, delineated environments, such as laboratories (Conley 2014; Fisher 2007; Smolka, Fisher, and Hausstein 2021), companies (Flipse and Van De Loo 2018; Flipse, Van Der Sanden, and Osseweijer 2013), and universities (Crow and Dabars 2020; Dabars and Dwyer 2022). An exception is a multi-sited research project by Richter et al. (2017), which aims at ‘moving STIR from the lab to the city’ (378). This project was designed to build capacities for socio-technical integration in the context of energy systems by conducting six coordinated STIR studies in university, commercial, and policy settings in Portland and Phoenix. Preliminary results suggest that the effects of STIR remained confined to the local settings in which the studies had taken place (Levenda 2022). There was no communication across these settings to shed light on shared responsibilities in the respective city, nor did the project install infrastructures that would support the spread of the STIR method throughout the cities for systems-level capacity building. We address the limitations of the project by Richter et al. (2017) in adapting STIR for responsible innovation ecosystem governance.

We select STIR for responsible innovation ecosystem governance for two reasons. First, STIR aligns well with the reflexive governance literature in the sense that it also considers technical experts like scientists and engineers to engage in ‘de facto governance’ (Rip 2006, 83). They are assumed to always already participate in co-steering research and innovation processes through socio-technical integration without necessarily being aware of their participation (Glerup 2015; Glerup, Davies, and Horst 2017). Instead of introducing societal consideration from outside into technoscientific workflows, STIR was designed to study and assess technical experts’ own capacities for socio-technical integration (Fisher 2019). In doing so, STIR has been shown also to enhance experts’ reflexive awareness of socio-technical integration, which enabled them to deliberately modulate emerging governance patterns (Flipse, Van Der Sanden, and Osseweijer 2013; Lukovics and Fisher 2017; Schuurbiers 2011; Smolka, Fisher, and Hausstein 2021).

Second, STIR has been applied in more than 80 studies worldwide (Fisher et al. 2022; Smolka et al. 2022). These studies indicate that STIR can assess and enhance experts’ capacities for socio-technical integration. Capacity building was captured in the reflexive, discursive, and material changes that could be related to the regular use of the decision protocol (ibid.). This track record stems from the high rate of interactions between the embedded humanist and the experts involved in a STIR study (Fisher 2019). R(R)I engagement methods, such as Constructive Technology Assessment (CTA), that rely on more time-efficient interaction formats (e.g. workshops) are equally successful in stimulating reflexive changes, but, to the best of our knowledge, practical changes have been documented less decidedly (Konrad 2021; Rip and Van Lente 2013). As studies have documented practical effects of STIR dialogues in a variety of empirical settings, they have widened the expert category to include, for example, city officials (Richter et al. 2017), project managers in industry (Flipse and Van De Loo 2018), and students (Pickering, Fisher, and Ross 2022). We therefore assume that STIR can be successfully used with a diversity of actors within an innovation ecosystem.

Despite these reasons for selecting STIR from a plethora of R(R)I engagement methods, we do not claim that STIR is the only method suitable for responsible innovation ecosystem governance. Quite the opposite, other R(R)I engagement methods can both inspire and be inspired by our methodological adaptations of STIR. More specifically, we draw on CTA to adapt STIR to innovation ecosystems. Krabbenborg (2013) described CTA as going beyond the common focus of R(R)I engagement methods on specific actor groups in bounded environments, for example the laboratory floor. According to her, CTA acknowledges that ‘scientists act on many more ‘floors’ than the laboratory floor, and encounter different groups of actors’ (168). Therefore, it takes multiple floors into consideration in building reflexive capacities for diffusing and embedding new technologies in society. CTA relies on ethnographic insertion in different contexts and multi-stakeholder workshops to facilitate and study exchange between the different places where technology development is carried out and affects society (Rip and Kulve 2008; Schot and Rip 1997).

Our adaptations of STIR are informed by CTA because Voß and Kemp (2006) point out that CTA could be considered as a reflexive governance approach. Through multi-sited research and engagement, CTA initiates procedures that expose problem definitions, assessment criteria, and action strategies of different actors within a system. Actors can thus adapt their definitions, criteria, and strategies in a process of mutual learning. In reflecting on and practically handling the diversity of problem definitions, the ambiguity of assessment, and the contingency of effects of human actions, actors reflexively engage with the conditions of governance in complex systems. In the following, we lay out how concepts and methods of CTA aligned with reflexive governance help us to develop three methodological adaptations of STIR: (1) rethinking the role of the embedded humanist through the concept of the engagement agent, (2) incorporating multi-stream engagement in midstream modulation, and (3) embedding STIR in a multi-method research design to shift from a focus on individual experts to networked responsibility among stakeholders. To present each adaptation, we first introduce relevant concepts and methods, after which we discuss potential challenges to the method, and, finally, we provide suggestions for responding to these challenges.

Engagement agent in curated STIR workshops

The concept ‘engagement agent’ was introduced by Te Kulve and Rip (2011) to denote social scholars preparing stakeholder engagement activities who can serve as ‘linking pins’ (704) among multiple levels of a particular research and development domain. Drawing on Conley’s (2014) STIR studies, Conley and Fisher (2019) suggest that embedded humanists can become engagement agents if they ‘traverse multiple “streams” and “work at more than one level” of innovation processes’ (247). We propose that an engagement agent can intervene at multiple levels through the use of regular STIR workshops (Figure 1). These workshops can be considered as ‘bridging events’ (Rip and Robinson 2013, 4) in which bridges are created between actors from different streams of an innovation ecosystem to understand each other’s values and perspectives.

Figure 1. STIR workshops curated by an engagement agent.

In curated online STIR workshops (Smolka and Fisher 2022), actors are grouped in pairs to use the STIR protocol for mapping each other’s work-related decision-making processes under the guidance of an engagement agent. The goals of the workshop are twofold: to widen actors’ value horizon by increasing their understanding of the socio-ethical concerns relevant to other streams of the innovation ecosystem, and to teach them a dialogic reflection tool for self-use with colleagues in local work contexts. STIR can be deployed as a reflection tool for self-use because it views actors as having resident capacities for socio-ethical reflection. Actors can thus expand their own socio-ethical considerations by engaging in reflexivity inducing questioning (ibid.). ‘[C]onscious reflection on assumptions, values and the basic premises of the system that is supposed to be in need of change’ (Beers and Van Mierlo 2017, 433) is an important component of reflexive governance. The regular use of the STIR protocol in curated workshops and local work contexts creates and maintains learning spaces, where the focus is less on problem-solving, but on working toward desirable ends. As actors enact collaborative reflexivity in these spaces, they can build capacities for participating in the responsible governance of the innovation ecosystem in which they operate.

By encouraging STIR workshop participants to use the method with colleagues in their local work contexts, the capacity for socio-technical integration can be cultivated at the systems level. This thesis is supported by literature on dynamic capabilities which denote an organization’s ability to build competencies for dealing with changing environments (Helfat et al. 2007; Teece 2009). Salvato and Vassolo (2018) theorize that an organization’s dynamic capabilities result from the amplification of individual capacities for creatively responding to new situations throughout a system. Amplification occurs when actors engage in (formalized) dialogues, which helps create and transfer new knowledge and skills. As studies indicate that this theory holds for the cultivation of organizational capabilities for value-sensitive innovation (Kokshagina 2021; cf. Garst et al. 2021; Watson et al. 2017), we assume that dialogic STIR interactions can function as a social mechanism for spreading capacities for socio-technical integration through an innovation ecosystem.

Yet, in the case of innovation ecosystems, having a diversity of actors – for example, from multiple companies, research organizations, and municipal agencies – participate in workshops to learn about STIR may provoke one of the many tensions highlighted in research on R(R)I in competitive environments (Blok, Hoffmans, and Wubben 2015; Brand and Blok 2019; Van De Poel et al. 2020). This tension pertains to the ‘open innovation paradox’ (Bogers 2011): although companies recognize the benefits of knowledge sharing for collaboration, they are reluctant to do so because knowledge leakage to other companies can lead to a loss of competitive advantage. Fear of technology ‘theft’ may prevent industrial actors from tapping into the collaborative potential of STIR due their reluctance to share work-related decisions, strategic considerations, and underlying motivations with any ‘outsider’ of their company.

Although deliberate pairing of STIR dialogue partners who have already established a relationship of trust through prior collaboration could alleviate such fears, the open innovation paradox may not be fully resolved. Instead, curated STIR workshops are generative of an ‘experimental situation [that] can lead to new and unexpected questions, and problem-solving – including the creative production of research tools and methods’ (Delgado and Åm 2018, 5) – for studying and negotiating open and closed innovation practices (Chesbrough 2003).

Multi-stream engagement through brokers

If mutual learning is the principal driver of systems-level capacity building for reflexive governance, it becomes paramount who is involved in that learning and who takes responsibility for creating and maintaining learning spaces (Voß, Smith, and Grin 2009). Reflexive governance acknowledges that all actors in a system can adopt this responsibility, but that ‘some actors may contribute more to the self-organization than others’ (Rip 2006, 87). Actors who are willing to contribute more to the self-organization of learning spaces than others are the target participants in STIR workshops. These actors can help ensure that capacity building, rather than remaining restricted to STIR workshops guided by an engagement agent, will further permeate the innovation ecosystem.

To facilitate multi-stream engagement – the parallel use of the STIR protocol in multiple streams that cut across arenas, governance levels, and actor groups of the innovation ecosystem – workshop participants are expected to take on the role of ‘brokers’ (Figure 2). In Wenger’s (1998) work on communities of practice, brokers are individuals who can introduce elements of one practice into another. The concept was taken up in STS, on the one hand, to denote experts who communicate their knowledge to potential users (Pielke 2007; Turnhout et al. 2013), and, on the other hand, to specify the intermediary role of social scholars in interdisciplinary projects (Pohl et al. 2010; Wittmayer and Schäpke 2014). As it is a demanding task to know practices of different disciplinary communities and mediate between them, Koskinen (2008) identifies characteristics of good brokers: motivation to expand, use, and distribute knowledge gained in foreign domains to others; an attitude of care for others to succeed; social competencies to empathize with others and develop trustful relationships. These characteristics mirror the ‘ethos of engagement’ (Fisher 2018) in STIR, which involves creating interpersonal conditions conducive to collaborative reflection and open sharing. Skillful brokers can create such interpersonal conditions and carry engagement activities forward.

Figure 2. Brokers from different stakeholder groups participate in curated STIR workshops and transfer knowledge and skills to their local work contexts.

To identify actors with characteristics of good brokers, the engagement agent needs to ‘move about’ (Rip and Robinson 2013, 48) the different streams of an innovation ecosystem to insert herself in a variety of work floors, ranging from university labs and production lines to municipal offices. For this purpose, patchwork ethnography (Günel, Varma, and Watanabe 2020; Tsing 2004) can be useful because it approaches ethnographic knowledge production as a multi-sited, short-term, and fragmented undertaking. Through patchwork ethnography, the innovation ecosystem is constructed as an object of inquiry by ‘following the people’ (Marcus 1995, 106; Latour 1988) and their connections to other relevant actors. As the ecosystem may change over time, patchwork ethnography involves an ongoing process of searching for and co-defining the flexible boundaries of the system. In this process, potential brokers are identified who can help scale up capacity building for responsible innovation ecosystem governance.

However, the search for brokers might be complicated by the reluctance or hesitation of research units, engineering groups, and start-ups to invest time and staff in engagement activities, in particular when these activities are as demanding as brokering and their profitability is doubtful (Brand and Blok 2019; Lubberink et al. 2017). Presentations on STIR at multiple levels of an innovation ecosystem can mitigate such doubts. Providing evidence that the method can enhance efficiency, creativity, and scientific robustness has shown to persuade actors to participate in STIR (Flipse, Van Der Sanden, and Osseweijer 2013). Such communication approaches the pursuit of instrumental ends as a pre-condition for reflexive socio-technical integration in competitive environments (Schuurbiers 2011). For instance, companies can be motivated to reflect on how to increase the sustainability of their products if sustainable products fulfill consumer demands and thus promise to increase sales (Garst et al. 2017). Furthermore, it is important to closely monitor the effects of engagement research by documenting related reflexive and practical changes in context-sensitive narratives (Mertens 2009) and quantitative survey results (Radatz et al. 2019). With the help of such results, actors within an innovation ecosystem may appreciate the relevance of responsible governance, which may keep them engaged in the long run.

Networked responsibility in stakeholder analysis and engagement

As STIR takes expert decisions as the unit of analysis, the method tends to individualize responsibility for reflexive socio-technical integration. The underlying assumption is that individual responsibility for orienting decisions and related practices toward socially desirable outcomes is at the root of systems change. Although the individual is recognized to be constrained by institutional, material, and political structures, STIR considers large-scale changes of these structures to be grounded in actors’ capacities for responsible decision-making (Fisher 2019). A focus on individual actors, however, neglects that systems are characterized by ‘networked responsibility’ (Timmermans et al. 2017). In an innovation ecosystem, actors have co-responsibilities for considering the broader implications of research and technology development. Capturing existing interlinking responsibilities and further developing and maintaining them helps attain socially responsive research and innovation (ibid).

To this end, we propose to complement STIR with stakeholder analysis and stakeholder engagement activities in a multi-method research design (Figure 3). A process-oriented stakeholder analysis (Popa, Blok, and Wesselink 2020) that takes responsibility as the unit of analysis traces existing responsibility relationships, patterns, and practices. Such a process-oriented approach acknowledges that responsibilities change over time along with the life cycle of an innovation ecosystem (Dos Santos, Zen, and Bittencourt 2021). By tracing change, we can better understand how these responsibilities interrelate across stakeholder groups whose boundaries are difficult to draw – especially in innovation ecosystems where activities like research, development, and commercialization cut across organizations (Godin 2009).

Figure 3. Multi-method design of engagement research for systems-level capacity building.

To make research on networked responsibility available for engagement, the evolving findings of the ongoing stakeholder analysis need to inform the design of STIR workshops and scenario workshops. Scenario workshops take the evolving nature of innovation ecosystems as a resource for engagement. This means that these workshops can facilitate democratic processes in which divergent (potential) stakeholders reflect together on future development paths of the innovation ecosystem and its responsible governance. They provide a space for anticipation and mutual learning between the actors operating at different streams of an innovation ecosystem who influence each other without necessarily being aware of these influences.

In reflexive governance, scenario workshops are considered as a relevant method for attending to potential path dependencies in complex systems (Voß and Kemp 2006; Voß, Truffer, and Konrad 2009; Weber 2009). Multi-actor, multi-level dynamics characterized by interdependencies may result in organizational patterns and structural configurations emerging as unintended side-effects of actors’ strategic actions. These patterns and configurations may then shape the variety of directions open to future changes. For this reason, the evolution of an innovation ecosystem is typically path-dependent. Although the complex dynamics within an innovation ecosystem make it impossible to predict emerging paths with certainty, these paths can be anticipated. Anticipation refers to the exploration and appraisal of alternative development paths spurred by actions taken today. In scenario workshops, the range and form of perceptions and expectations of various actors can be revealed and feed into the construction of scenarios for different development paths. Such scenarios can raise awareness of future structural changes that may be provoked by seemingly minor decisions in the present. As scenario workshops are a key component of CTA, a number of case studies have been published that can inspire the methodological design of these workshops (for an overview, see Rip and Van Lente 2013).

In designing scenario workshops, it is important to keep in mind that a critical issue pertaining to stakeholder engagement in competitive environments is the divergence and potential conflict between visions, goals, motives, and values among stakeholders (Blok, Hoffmans, and Wubben 2015; Blok and Lemmens 2015). For example, tensions may arise between market interests in economic profit and public interests in social goods (Van De Poel et al. 2020). Technology developers may promise many possible applications and profitability of innovations so as to convince potential stakeholders to join an innovation ecosystem. Firms, suppliers, and societal stakeholders, however, may find it difficult to envision their role and potential gains in the innovation ecosystem (Dattée, Alexy, and Autio 2017). Each ‘actor involved has only a limited view of the whole – which may be incommensurable with constructions of others’ (Smith and Stirling 2007, 280). Such incommensurability is a condition of reflexive governance by actors who recognize that the future is an indeterminate outcome of negotiations between conflicting values and worldviews.

Conflicting values and worldviews thus feature as a point of departure for engagement research rather than as an obstacle to be overcome. Through conflicts, contestation, and controversy, matters of concern are brought to light that call for empirical analysis and stakeholder negotiations. Negotiation processes do not necessarily aim at consensus, but can also result in an agreement to disagree, acknowledging the legitimacy of opposing views and the impossibility to come to terms with dissent in a conclusive manner (Delvenne and Parotte 2019). Hence, multi-stakeholder scenario workshops serve to reveal that a technology can have contested impacts on society – impacts that researchers and innovators may not have anticipated. To create a space in which stakeholders dare to express their views, the abovementioned brokers can be invited to join scenario workshops so as to draw on the social capital developed through regular STIR interactions. Incorporating informal socialization activities (e.g. outings, walkshops) as well as formal social mechanisms (e.g. world café, open space methods) into scenario workshops may also help explore and work with tensions in stakeholder engagement research (Lipmanowicz and McCandless 2013).

Conclusions and recommendations for future research

The methodological adaptations of STIR presented above aim to realize responsible innovation ecosystem governance in a double sense. On the one hand, they study the capacity of innovation ecosystems to integrate societal considerations into research and innovation processes. On the other hand, engagement research can build systems-level capacity for responsible governance by making actors operating in multiple streams of an innovation ecosystem reflexively aware of the ways in which their decisions co-steer governance. By drawing on the reflexive governance approach in transformation and transition literature, we introduce an ecosystem perspective into R(R)I engagement research. Although we concentrate on STIR in this article, we suggest that the methodological adaptations developed here could inform adjustments of other engagement research methods for responsible innovation ecosystem governance. These methods have been criticized for falling short of considering the systemic qualities of innovation due to their focus on the micro practices of specific actors at discrete events or within bounded environments.

To move R(R)I engagement methods from bounded environments to innovation ecosystems, we have three recommendations. First, to implement R(R)I engagement methods in open innovation ecosystems marked by fuzzy and evolving boundaries, process-oriented stakeholder analysis complemented with patchwork ethnography can help trace the boundary-work in and around the system. An empirical analysis of boundary-work helps to flexibly adapt engagement methods to actors arriving and perishing, negotiations of local ecosystem visions, and evaluations of brokering performance. Second, to enhance mutual learning among the actors within an innovation ecosystem, workshops can serve as bridging events. The workshop design should allow actors to explore views, values, and activities of other actors in the system so as to reveal contested aspects of innovation, networked responsibilities, as well as opportunities for collaboration. Moreover, introducing reflexive tools, such as the STIR protocol, to workshop participants can guide mutual learning toward capacity building for socio-technical integration. Third, as R(R)I engagement in complex systems can have unexpected and unintended effects, the effects of engagement need to be subject to monitoring, documentation, and evaluation. Ongoing evaluation that provides feedback in real time to an engagement agent can help her adjust research activities adaptively and experimentally.

As the evaluation of R(R)I engagement in innovation systems has received limited attention (Nieminen and Ikonen 2020), we suggest that future research could make efforts to address this gap in the literature. We have several suggestions for the evaluation of innovation ecosystem governance. Given that innovation ecosystems are characterized by complex processes and multiple relations between actors, these systems do not match well with linear approaches to evaluation prevalent in prior R(R)I projects (ibid.). In transition and transformation research, by contrast, process-based approaches to evaluation seek to capture and respond to systemic complexity, for instance through ‘reflexive monitoring in action’ (Van Mierlo et al. 2010), abbreviated as RMA. RMA does not happen after the fact but is part of the engagement process. The insights gained from monitoring and evaluation are subsequently experimented with in engagement activities. Klaassen et al. (2020) pair RMA with criteria for desirable impacts of R(R)I engagement as a means to monitor and continuously improve them. Criteria for systems-level capacity building of socio-technical integration could comprise the spread and use of the STIR protocol, different types of learning (cf. Parandian 2012; Van Merkerk 2007), and midstream modulations (Fisher and Schuurbiers 2013) in multiple streams of the innovation ecosystem. As R(R)I engagement changes in relation to the evolution of the ecosystem, these criteria cannot be defined once and for all, but require sustained research, reflection, and multi-stakeholder discussion.

Future research on innovation ecosystem governance, in particular through multi-stakeholder engagement, should also pay attention to power dynamics within the system and the real-world political contexts in which the system is embedded. Power asymmetries in innovation ecosystems, for instance between municipal and corporate actors, can be reflected in de facto governance practices, as well as engagement activities (Voß, Smith, and Grin 2009). For example, economic interests may overdo sustainability concerns in scenario workshops where powerful actors dominate the discussion. There is also a potential to marginalize specific social groups and interests if only actors displaying willingness to cooperate are invited to join participatory processes (ibid). Such methodological decisions tend to stabilize the hegemonic political economy instead of ‘opening up’ (Stirling 2008) alternative social orders, neglected issues, and peripheral perspectives. Voß and Bornemann (2011) make several suggestions for taking an opening-up approach in reflexive governance designs, including procedural arrangements to avoid domination in discourse. A task for future research is to study how innovation ecosystem governance is shaped by power dynamics and how R(R)I engagement research enacts or deflects incumbent political orders.

The approach to innovation ecosystem governance presented in this article calls for empirical implementation. According to Voß, Smith, and Grin (2009), governance design is an experimental process in which a ‘preliminary design’ (292) is redesigned with respect to feedback from its context of implementation. Along these lines, we consider our methodological adaptations of STIR as a preliminary design, one that cannot simply be put into practice, but gives rise to feedback loops of empirical testing, reflection on the implementation experience, and readjustment. While the adaptations of STIR respond to the obstacles to R(R)I in competitive environments that have been identified in prior research (open innovation paradox, resistance to participate, diverging stakeholder values), empirical testbeds will reveal how our methodological design needs to be revised in order to address unexpected challenges arising in specific types of innovation ecosystems.

An agenda for empirical research could be to tailor our preliminary design for responsible governance to different types of innovation ecosystems. Innovation ecosystems focusing on the development of novel computer hardware (Smolka et al. forthcoming), personalized health services (Pombo-Juárez et al. 2017), and high technology for agriculture (Mikhailov et al. 2021) are all characterized by complexity, openness, and mutual learning. However, they differ with regards to actor configurations and levels of maturity. For example, in innovation ecosystems emerging around personalized health services, such as wearable monitoring devices coupled with data analytics, relations between academic researchers and commercial firms tend to be weak because the demand for such services is not clearly articulated. In such ecosystems, multi-stakeholder scenario workshops are of particular relevance when it comes to exploring alternative visions and defining shared responsibilities. In mature innovation ecosystems, such as the Californian agtech ecosystem, in which software development start-ups play an intermediary role between farmers and multinational companies providing machinery and equipment, start-ups could be targeted to adopt the role of brokers. Tailoring multi-method engagement research designs to specific types of ecosystems could be the long-term goal of empirical research on responsible innovation ecosystem governance.

To conclude, we emphasize that responsible innovation ecosystem governance cannot be realized by one approach alone. As the mainstreaming and standardization of R(R)I runs the risk of relegating authority about proper implementation into the hand of a powerful few and of establishing technocratic control over procedure (Frahm, Doezema, and Pfotenhauer 2021; Voß and Amelung 2016), we advocate pluralism in systems-level governance. Along these lines, Stahl (2021, 2022) makes several suggestions for the governance of AI ecosystems, ranging from regulation and legislation over impact assessments to ethics dialogues and training. While we emphasize that both larger national and local measures are relevant mechanisms for the governance of innovation ecosystems, this article provides a methodological contribution to advancing an eco-systemic turn in R(R)I through systems-level capacity building.

Acknowledgements

We thank Erik Fisher, Barbara Grimpe, Philipp Neudert, Lyric Peate, Jan-Peter Voß and the reviewers for their comments on earlier versions of this article.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was conducted as part of the NeuroSys Cluster4Future funded by the German Federal Ministry of Education and Research (Bundesministerium für Bildung und Forschung).