2.1 Complexity

Systemic risks originate in complex adaptive systems, such as ecosystems, markets, the energy grid or the internet. Broadly defined, a system is an entity of interacting elements, and it is defined by boundaries, which set it apart from other systems. As systems become more differentiated, they also become more complex. Increasing complexity is a major trend of the evolution of all systems, be they biological, social, cultural, technological or economic (Laubichler & Renn, Reference Laubichler and Renn2015). These complex systems are composed of many interconnected parts. Complexity can be defined as a ‘causal chain with many intervening variables and feed-back loops that do not allow the understanding or prediction of the system's behaviour on the basis of each component's behaviour’ (Aven et al., Reference Aven, Ben-Haim, Andersen, Cox, Droguett, Greenberg, Guikema, Kröger, Renn, Thompson and Zio2018, p. 5). Increasing interconnectivity therefore results in increasing complexity.

Complex systems can be characterised by the diversity of their parts, density of their interconnections and their size, i.e. their number of parts and connections (Ladyman & Wiesner, Reference Ladyman and Wiesner2020). Increasing any of these three dimensions makes the system more complex. Complex systems become dynamic when feedback between interrelated parts and/or the system's environment generates the emergence of unexpected system behaviour (Taleb, Reference Taleb2020; Zio, Reference Zio2016). Especially complex adaptive systems, i.e. systems that involve many components that adapt and change their properties and behaviour as they interact – are at the heart of important contemporary problems (Holland, Reference Holland2006). Complex adaptive systems depart from linear trajectories of development due to feedback mechanisms. Therefore, interconnectivity increases adaptation but also vulnerability to external stressors, leaving the system prone to cascading effects and regime shifts (Helbing, Reference Helbing2013). Ripple effects may even spread beyond the originating systems, leading to transboundary effects of contagion. The COVID-19 pandemic not only illustrates how pathogens rapidly spread thanks to the global networks of international passenger transport but also the cascading effects of the pandemic. The pandemic affected all our society's vital systems: from the food supply to our economic and financial systems, and from education to culture and social life. Similarly, climate change is caused by emissions in energy and other predominantly human dominated systems. The resulting changes in the climate change system then can have cascading negative effects on biodiversity (Habibullah et al., Reference Habibullah, Din, Tan and Zahid2022), food security (Wheeler & von Braun, Reference Wheeler and von Braun2013), human migration and armed conflict (Scheffran et al., Reference Scheffran, Brzoska, Kominek, Link and Schilling2012). As energy systems are beginning to decarbonise, the effects of warming may impact the ability to produce green energy (Groundstroem & Juhola, Reference Groundstroem and Juhola2021), further exemplifying the feedback mechanisms in the complex systems.

2.2 Uncertainty

The category of uncertainty comprises several aspects, such as statistical variation, measurement errors and ignorance (van Asselt, Reference van Asselt2000) which challenge the strength of confidence put in cause-and-effect relations. Systemic risks originate in highly complex systems which entail various interdependencies and feedback mechanisms. This makes it difficult to establish the relation between cause and effect, i.e. which events or hazards cause systemic risks and how these risks propagate in society. Furthermore, it is difficult to forecast the effects of manipulating these systems by governance interventions due to stochastic effects (Tannert et al., Reference Tannert, Elvers and Jandrig2007). In addition, systemic risks originate in dynamic systems that are sensitive to initial conditions. Small differences in initial conditions result in large differences at a later stage. Consequently, each systemic risk develops due to its own set of conditions. This means that ‘for predicting any event at any level of precision, all sufficiently past events are approximately probabilistically irrelevant’ (Werndl, Reference Werndl2009, p. 197). Thus, generalisations rather pertain to the underlying mechanisms that cause systemic risks than specific events. Evolutionary approaches towards analysing the complex dynamics in social-ecological systems highlight processes of emergence, self-organisation, adaptation and niche construction as key for understanding system change (Currie et al., Reference Currie, Borgerhoff Mulder, Fogarty, Schlüter, Folke, Haider, Caniglia, Tavoni, Jansen, Jørgensen and Waring2023; Renn et al., Reference Renn, Laubichler, Lucas, Kröger, Schanze, Scholz and Schweizer2022).

Additionally, systemic risks develop over long periods of time and come with sudden regime shifts marked by tipping periods or points. The temporal component of regime shifts is nearly impossible to determine, although some indicators signalling transitions have been identified (Kopp et al., Reference Kopp, Shwom, Wagner and Yuan2016; Scheffer, Reference Scheffer2010; Scheffer et al., Reference Scheffer, Bascompte, Brock, Brovkin, Carpenter, Dakos, Held, Van Nes, Rietkerk and Sugihara2009). Extrapolations of future developments from past actuarial data are inhibited by tipping points. The emerging question of how can we ‘logically go from specific instances to reach general conclusions’ (Taleb, Reference Taleb2010, p. 40) fundamentally challenges inductive approaches to science.

Considerations of knowledge strength are therefore important in this context. Subjective probability is ‘an exercise of judgement, especially under conditions where actuarial data is unavailable or meagre’ (Rosa, Reference Rosa, Renn and Walker2008, p. 111). It expresses degrees of beliefs about events or unknown quantities based on some knowledge. First, the analysis of systemic risks is framed at the outset by the demarcation of a system's boundaries (Juhola et al., Reference Juhola, Filatova, Hochrainer-Stigler, Mechler, Scheffran and Schweizer2022a). Bearing in mind that boundaries are analytical (not empirical) constructs, the answers to the questions of what counts as the ‘inside’ of the systems and what as its ‘outside’ is based on suppositions that could hide ‘critical assumptions and therefore provide a misleading description of the possible occurrence of future events’ (Aven, Reference Aven2013, p. 46). Thus, the definition of system boundaries a priori entails an exercise of judgement. Second, judgements about the state of knowledge are made regarding model deployment. Models are inevitably entrenched in disciplinary thinking. Furthermore, epistemic uncertainty comes from taking estimates from well-controlled experiments or from modelling and applying them to specific contexts that may differ from the context where knowledge was developed in ways that inhibit transferability (Dietz, Reference Dietz2023; Rosa, Reference Rosa1998).

Decisions about which models are appropriate for the analysis of systemic risks already determine the frame of analysis and limit the spectrum of potential results. Although quantitative methods which deal with uncertainty exist, e.g. Monte Carlo simulation or Bayesian updating, these methods alone are insufficient for an analysis of systemic risks because quantifiable uncertainties are interspersed with unquantifiable uncertainties. For instance, unquantifiable uncertainties are associated with problem framings, model structures, system boundaries and value-ladenness (van der Sluijs et al., Reference van der Sluijs, Craye, Funtowicz, Kloprogge, Ravetz and Risbey2005, p. 482).

This situation has been termed deep uncertainty. Cox (Reference Cox2012) describes deep uncertainty as a situation in which ‘trustworthy risk models giving the probabilities of future consequences for alternative present decisions are not available; the relevance of past data for predicting future outcomes is in doubt; experts disagree about the probable consequences of alternative policies – or, worse, reach an unwarranted consensus that replaces acknowledgement of uncertainties and information gaps with groupthink – and policymakers (and probably various political constituencies) are divided about what actions to take to reduce risks and increase benefits. For such risks, there is little or no agreement even about what decision models to use’ (Cox, Reference Cox2012, p. 1607). Therefore, modelling systemic risks requires not only multi-method model architectures but also a genuine effort at prudent judgement and interdisciplinary integration.

The COVID-19 pandemic's diverse impacts gave rise to emergent processes and phenomena. Agent-based models aim to demonstrate emergent effects due to system dynamics as they simulate the behaviour of individual or collective agents (e.g. Billari et al., Reference Billari, Fent, Prskawetz and Scheffran2006; Bonabeau, Reference Bonabeau2002). Agent-based models thus allow assessing the impact of economic shocks by considering adaptive human behaviour and disaggregated information for targeted policy interventions (Poledna et al., Reference Poledna, Miess and Hommes2020a, Reference Poledna, Rovenskaya, Crespo Cuaresma, Kaniovski and Miess2020b). For instance, the agent-based model developed by Poledna et al. (Reference Poledna, Miess and Hommes2020a) has been used to forecast the macroeconomic effects of the COVID-19 pandemic at the sectoral level in Austria (Poledna et al., Reference Poledna, Miess and Hommes2020a). These promising developments hopefully spark further research on interdisciplinary coupled models. With regards to climate change, uncertainty related to modelling efforts has been recognised for a long time now (Reilly et al., Reference Reilly, Stone, Forest, Webster, Jacoby and Prinn2001), but consideration and search of consensus on uncertainty in climate change (Adler & Hirsch Hadorn, Reference Adler and Hirsch Hadorn2014) or energy systems modelling continues to be a challenge (Plaga & Bertsch, Reference Plaga and Bertsch2023). Further, a recent review of uncertainties related to adaptation to climate impacts, reveals that multiple sources of uncertainty affect adaptation decisions (Moure et al., Reference Moure, Jacobsen and Smith-Hall2023). While previously scientific uncertainties have been at the centre of study, increasingly attention is paid to uncertainties in terms of the role and function of the decision maker and timeframes associated with the decisions (Moure et al., Reference Moure, Jacobsen and Smith-Hall2023). A review of US local level adaptation plans shows that while the presence of uncertainty is acknowledged at the local planning level, no specific measures are taken to address it (Woodruff, Reference Woodruff2016).

2.3 Ambiguity

Ambiguity stems from unresolved issues of complexity and uncertainty. Uncertainty and complexity give rise to controversies about knowledge base, differences in framing of the problem, and inadequacies of the institutional arrangement at the science–policy interface (van der Sluijs et al., Reference van der Sluijs, Craye, Funtowicz, Kloprogge, Ravetz and Risbey2005), all of which result in ambiguity. Ambiguity denotes the variability of interpretations based on identical observations or data assessments (Renn et al., Reference Renn, Laubichler, Lucas, Kröger, Schanze, Scholz and Schweizer2022). A plurality of viewpoints for evaluating data exists under conditions of ambiguity. More precisely, two kinds of ambiguity can be identified. Interpretative ambiguity refers to a legitimate range of interpretations of data and research results. Several interpretations of the results are possible with stochastic statements, all of which are based on the same evidence even though they suggest different governance measures. For instance, given the logistical constraints on the rollout of the COVID-19 vaccine, decisions had to be made which susceptible groups of the population to vaccinate first. Most countries decided to start with the most elderly. Another route would have been to start the rollout with persons who have weak immune systems irrespective of their age, e.g. persons who are receiving cancer treatment. Decisions such as these must be made based on incomplete knowledge and often under time pressure. Similar observations have been noted in relation to climate policy (Millner et al., Reference Millner, Dietz and Heal2013). Decision-making in such situations is aided by heuristics. Heuristics are mental strategies for situations in which individuals act under conditions of limited information (Simon, Reference Simon1991). Heuristics therefore play an important role in decision- and policy making by allowing people to focus on specific aspects (and ignoring others) of a task or problem thereby reducing the complexity of decision-making (Gilovich & Griffin, Reference Gilovich, Griffin, Gilovich, Griffin and Kahneman2002; Kahneman, Reference Kahneman2011). Heuristics for climate change adaptation have emphasised an incremental approach for governing, including no-regret options in order to retain flexibility (Nalau et al., Reference Nalau, Torabi, Edwards, Howes and Morgan2021).

Normative ambiguity refers to heterogeneous normative assumptions that lead to divergent opinions on policies and management options (Renn et al., Reference Renn, Klinke and van Asselt2011). Systemic risks often touch upon common pool problems for which normative ambiguity plays a role (Renn et al., Reference Renn, Laubichler, Lucas, Kröger, Schanze, Scholz and Schweizer2022). Common pool problems are the result of overconsumption and depletion of common resources due to the selfish individual behaviour of a few (McCay, Reference McCay, Ostrom, Dietz, Dolšak, Stern, Stonich and Weber2002; Ostrom, Reference Ostrom1990, Reference Ostrom2010b, Reference Ostrom2010a). If all users were to restrain themselves, the resources could be sustained. A dilemma arises, however. If some people limit their use of resources and others do not, then the resource still collapses, leaving the first set of people without the short-term benefits of taking their share (Dietz et al., Reference Dietz, Dolšak, Ostrom, Stern, Ostrom, Dietz, Dolšak, Stern, Stonich and Weber2002). (In fact, their restraint results in a bigger slice of the cake for the free-riders who use up their share in addition to their own.) These problems are therefore closely related to the ‘free rider problem’ which occurs when those benefiting from public goods or common resources do not pay for them. Systemic risks often play out as a tragedy of the commons, a prominent example being greenhouse gas emissions resulting from carbon-intensive industries and lifestyles.

Normative ambiguity also surrounds the COVID-19 pandemic. For instance, governance measures implemented in many countries around the world put severe restrictions on people's lives. Basic human rights such as freedom of movement have been curtailed for the greater good. People whose occupation allows a smooth transition to home office work can continue with their jobs whereas others face short-time work or unemployment. Issues of normative ambiguity and equity ensue. To overcome normative ambiguity in relation to climate change, there have been calls for ensuring accountability of actions remains, as well as ensuring ethical interactions among knowledge exchange processes (Fleming & Howden, Reference Fleming and Howden2016) and recognising different forms of knowledge (Latulippe & Klenk, Reference Latulippe and Klenk2020).

3.1 Analysis of systemic risk

Governance of systemic risks is concerned with the analysis of tightly coupled systems, their various interdependencies and the resulting dynamics. Risk analysis here investigates feedback mechanisms between components of a system at the intra-system level and at the interaction with other systems at the inter-system level which result in transboundary cascading effects. Three levels of abstraction can be differentiated to map out the complex embeddedness of nested systems.

First, at the level of system I, analysis is concerned with the intricacies of the risk-emitting system, such as a financial or technological system. In many cases, the systemic risk is being emitted from a convergence of systems. For instance, the convergence of the biohazard SARS-CoV-2 with the societal dispositions of complexity and functional differentiation, e.g. in the form of global supply chains, gave rise to the systemic properties of the COVID-19 pandemic, which make this pandemic stand out compared to previous pandemics, such as the Spanish flu and the plague. Thus, risk analysis needs to capture the dynamics of feedback mechanisms and the stochasticity of non-linear systems at both the intra- as well as the inter-system level. Modelling these complex interrelations is (even at this level) a challenging enterprise which calls for interdisciplinary cooperation and a modular approach towards model frameworks, with a focus on systemic thinking (Groundstroem & Juhola, Reference Groundstroem and Juhola2021).

Second, at the system II level, institutional and organisational structures, as well as the regulatory arrangements that engulf risk-emitting systems come into focus. Modern organisations are susceptible to ‘normal accidents', which start out small before cascading through tightly coupled systems (Perrow, Reference Perrow1984). Risk analysis is therefore concerned with organisational structures as well as organisational risk management protocols and practice. Furthermore, risk-emitting systems are surrounded by a vast number of institutions at various levels of scale and responsibility. Institutional fragmentation leads to a spread of authority across multiple state and non-state institutions with various political and functional overlaps. As a result, responsibilities and legitimacy become clouded (Widerberg, Reference Widerberg, Pattberg and Zelli2016). System II analysis therefore must focus on mapping out institutions and networks concerned by the risk-emitting systems on the system I level. Institutional mapping will lead to a clearer picture of responsibilities, legitimacy and accountability, thereby fostering effectiveness of governance.

Third, system III analysis is concerned with societal risk controversies that engulf risk-emitting systems and their institutional apparatus. Risk controversies speak of epistemic controversies, stakeholder and public interests, diverging beliefs and values, as well as manifestations of power. These controversies need to be mapped out to analyse the complex workings between public risk debates, more covert channels of risk proliferation (Beck & Kropp, Reference Beck and Kropp2011), mechanisms of risk perception and modes of risk amplification and attenuation. Furthermore, maps of risk controversies support argument-based tools for conflict resolution, which can serve as guidance for decision-making (Beck, Reference Beck2016). Mapping risk controversies will neither reconcile power plays nor competing epistemic claims. Yet, it may improve understanding among parties and therefore can fundamentally enhance transparency of risk governance (National Research Council, 2008).

In sum, analysis of systems I, II and III provides the epistemic foundation for risk governance of systemic risks. Demarcation of system boundaries is an issue that needs to be settled for each systemic risk individually. Nevertheless, some universal features stand out. Risk analysis of systemic risks needs to be concerned with interdependencies within and between systems as these cause cascading effects that surpass system boundaries. System modelling is therefore tasked with replicating system complexity (insofar as it is necessary), associated interdependencies and dynamics as well as potential cascading effects. Modelling systemic risks is also concerned with the task of demonstrating tipping points and early warning signals that herald transition periods (Lucas, Reference Lucas2022; Lucas et al., Reference Lucas, Renn and Jaeger2018). Research on network analysis approaches system dynamics from a perspective of evolving network dynamics. This framing focuses on connections among agents as a dynamic and evolving system. From a governance perspective, this approach seems promising as network scholars analyse the pathologies and potentials of network dynamics for reflection, innovation and social learning (Henry, Reference Henry2020, Reference Henry2023; Masuda et al., Reference Masuda, Liu, Reddy, Frank, Burford, Fisher and Montambault2018).

Modelling system I, II, III complexities and interrelations requires interdisciplinary collaboration and a modular model framework. This also calls for a transdisciplinary approach towards modelling and scenario building including experiential and tacit knowledge. However, model-based risk assessment and foresight are accompanied by uncertainties, only some of which are quantifiable (Funtowicz & Ravetz, Reference Funtowicz and Ravetz1990). Systemic risk modelling therefore needs to consider empirical and subjective uncertainties which are associated with ‘problem framings, model structures, assumptions, system boundaries, indeterminacies, and value ladenness’ (van der Sluijs et al., Reference van der Sluijs, Craye, Funtowicz, Kloprogge, Ravetz and Risbey2005, p. 482). Van der Sluijs et al. propose the NUSAP (Numeral Unit Spread Assessment Pedigree) method to complement state-of-the-art quantitative uncertainty methods with systematic qualitative assessments that address the methodological, epistemological and societal dimensions of uncertainty (van der Sluijs et al., Reference van der Sluijs, Craye, Funtowicz, Kloprogge, Ravetz and Risbey2005, p. 482). NUSAP is a notational system originally proposed by Funtowicz and Ravetz (Reference Funtowicz and Ravetz1990) which provides an analysis and diagnosis of uncertainty in the knowledge base of complex policy problems. This approach aims to capture both quantitative and qualitative dimensions of uncertainty by means of the five qualifiers of the NUSAP acronym, i.e. numeral, unit, spread, assessment and pedigree (van der Sluijs et al., Reference van der Sluijs, Craye, Funtowicz, Kloprogge, Ravetz and Risbey2005, p. 482). Assessment and pedigree constitute the more qualitative side of the NUSAP expression. Assessment expresses qualitative judgements about the information. Pedigree conveys an evaluative account of the process of information and knowledge production (van der Sluijs et al., Reference van der Sluijs, Craye, Funtowicz, Kloprogge, Ravetz and Risbey2005).

3.2 Governance principles for systemic risk: reflection, iteration, inclusion, transparency and accountability

Governance of systemic risks is also concerned with procedural considerations of governance. Concepts such as participatory governance (Fischer, Reference Fischer2006; Fung & Wright, Reference Fung and Wright2001, Reference Fung and Wright2011) and governance through networks (Torfing, Reference Torfing and Levi-Faur2012) focus on which kind of (institutional) actors will engage with each other in which ways. At a meta-level, governance research focuses on modes of governance and their self-reflexive and provisional nature. For instance, Kuhlmann et al. (Reference Kuhlmann, Stegmaier and Konrad2019), based on a review of the scholars in the field, propose tentative governance, which aims to create spaces for probing and learning instead of stipulating definitive targets, as a meta-criterion that permeates other forms of governance (Kuhlmann et al., Reference Kuhlmann, Stegmaier and Konrad2019, p. 1091). In a similar vein, experimentalist governance continuously reflects on its assumptions, treats all solutions as provisional and corrigible, as well as considers and compares alternative approaches to advancing general aims (Sabel, Reference Sabel, Heckscher and Adler2006). Experimentalist governance focuses on multi-level concertation that relies on networking various types of decision-makers at several levels (Sabel & Zeitlin, Reference Sabel and Zeitlin2008). Consequently, ‘experimentalist governance structures have emerged as a widespread response to turbulent, polyarchic environments, where strategic uncertainty means that effective solutions to problems can only be defined in the course of pursuing them, while a multi-polar distribution of power means that no single actor can impose her own preferred solution without taking into account the views of others’ (Zeitlin, Reference Zeitlin and Zeitlin2015, p. 11). By the same token, adaptive governance starts with the premise of multilevel governance of nested, coupled systems. Adaptive governance implies power sharing that promotes participation by means of self-organised social networks that draw on various knowledge systems and experiences (Folke et al., Reference Folke, Hahn, Olsson and Norberg2005; Juhola, Reference Juhola2023).

Tentative, experimentalist and adaptive governance concepts, together with the inclusive risk governance approach mentioned above, provide stepping stones for governance of systemic risks. These include the governance principles of reflection, iteration, inclusion, transparency and accountability. Reflection considers the volatile nature of systemic risks. The opacity of systemic risks relates foremost to uncertainties about tipping points and (to a somewhat lesser extent) cascading effects. In this respect, systemic risks resemble black swan events (Taleb, Reference Taleb2010). Black swans are extreme events with regards to their potential impacts that are associated with a low probability of occurrence based on current knowledge and beliefs (Aven, Reference Aven2013, p. 49). Several types of black swan events are relevant to the discussion on systemic risks. The first type of black swan events denotes risk phenomena totally unknown to science (unknown unknown risks). Risk governance can only offer generic guidance for this type of event, such as constant vigilance, continuous research and the enhancement of the risk-absorbing system's resilience. The second type of black swan event relates to unknown known risks. These events have not been considered by those carrying out risk analysis, yet they are known to other experts or stakeholders (Aven & Krohn, Reference Aven and Krohn2014).

Thus, the process of governing systemic risks demands continuous reflection of the underlying premises not only of system I, II and II risk analysis but also of the governance process itself. Continuous reflection requires iteration. The third type of black swan event is on the ‘list of known events in the risk analysis but judged to have negligible probability of occurrence, and thus [is] not believed to occur’ (Aven, Reference Aven2015, p. 84). The COVID-19 pandemic falls under the latter category with the caveat that there was no error about negligible probability of occurrence on the part of risk science. For instance, the German parliament had received a briefing document in 2013 that sketches out a potential scenario of a SARS virus pandemic for Germany (Deutscher Bundestag, 2013). In this document, the pandemic is expected to occur once every 100–1,000 years and the anticipated adverse effects on human health, the economy and democracy are supposed to be tremendous (Deutscher Bundestag, 2013, p. 56). This risk analysis conducted by the Robert Koch Institute and other German federal agencies has been verified by the outbreak of the COVID-19 pandemic in 2019.

Furthermore, the systemic risks of pandemics have been pointed out by Ian Golding and Mike Mariathasan in their seminal work The Butterfly Defect. Goldin and Mariathasan (Reference Goldin and Mariathasan2015) suggest that globalisation causes societal interdependencies that give rise to systemic risks such as pandemics. Thus, knowledge about probability and adverse effects of pandemics in general, and SARS mutations in particular, had been available prior to the COVID-19 pandemic. Yet the pandemic turned out to be an ‘extreme event relative to current knowledge and beliefs’ and thus qualifies as a black swan event according to Aven's definition (Aven, Reference Aven2013, p. 49). For climate change, this means broadening the perspective from single extreme events to considering to what extent climate change poses an existential risk to humanity (Huggel et al., Reference Huggel, Bouwer, Juhola, Mechler, Muccione, Orlove and Wallimann-Helmer2022; Magnan et al., Reference Magnan, O'Neill and Garschagen2023). The same dispositions that made contemporary societal systems susceptible to systemic risks also brought about unprecedented levels of human welfare and prosperity to an increasing number of people. Risk governance that aims to ameliorate the effects of this latter type of event must consider beliefs and value orientations for finding trade-offs between risks and benefits in iterative deliberative processes.

Governance of systemic risks therefore requires deliberation and inclusion. Governance implies the sharing of power with and amongst stakeholders and affected publics. Hence, participation, which serves to improve governance on substantive, instrumental and normative dimensions, is a core feature of governance for systemic risks. Participation and communication are also means to achieve transparency. Potential openings and stumbling blocks for governance are mapped out by an analysis of institutional and regulatory settings (system II) as well as societal debates (system III), which both permeate risk-emitting systems (system I). Trade-offs can be negotiated and management options can be identified through deliberation (Bächtiger & Parkinson, Reference Bächtiger and Parkinson2019), which clarifies interests of stakeholders and reveals value orientations of the public, and knowledge co-production (Hochrainer-Stigler et al., Reference Hochrainer-Stigler, Deubelli-Hwang, Parviainen, Cumiskey, Schweizer and Dieckmann2024). Thus, governance of systemic risks aims to establish accountability for risk management measures taken to ward off detrimental effects of systemic risks. However, accountability for systemic risks is difficult to establish and even harder to enforce in practice. Systemic risks are transboundary regarding the demarcation of system boundaries as well as cascading effects that spread across systems. Their modes of action also surpass national responsibilities and jurisdiction, thus calling for multi-level governance (Hooghe, Reference Hooghe and Hooghe1996) and international cooperation. Furthermore, systemic risks often touch upon common pool problems. For instance, climate change can be phrased (albeit in simplistic terms) as a common pool problem that has been caused by the overuse of fossil fuels by some with repercussions for the wellbeing of (almost) all. Classical suggestions for governing the commons focus on the community level, especially on design principles of rule-making and rule enforcement such as manageable size and well-defined borders of the resource system as well as the community (Agrawal, Reference Agrawal, Ostrom, Dietz, Dolšak, Stern, Stonich and Weber2002). Most notably, Elinor Ostrom's findings suggest that governing the commons requires collective-choice arrangements that allow participation by all affected individuals (Ostrom, Reference Ostrom1990). These suggestions have proven to be successful, especially at the community level. Governance of systemic risks, however, cannot be downscaled to the community level due to the transboundary nature of systemic risks which negates institutional containment and analytic compartmentalisation. Hence, effective governance on an aggregate level requires dealing with conflicts and factual information about the environment, human actions as well as information about uncertainty and values (Dietz, Reference Dietz2023; Dietz et al., Reference Dietz, Ostrom and Stern2003; Ostrom, Reference Ostrom1990).

At the same time, supranational institutions have been proposed to facilitate rule enforcement at the international level (Sandholtz & Stone Sweet, Reference Sandholtz and Stone Sweet1998). Empirical observations challenge this proposition. The Paris Agreement within the United Nations Framework Convention on Climate Change, for instance, can be interpreted as an attempt to transfer rule enforcement of climate change mitigation, adaptation and finance back to the national level. By the same token, multilateral agreements struggle with a backswing in the current political climate of international relations towards bilateral agreements. Current practice to curb carbon emissions focuses on a threefold strategy consisting of privatisation, e.g. via an emissions trading system, governmental control and voluntary cooperation. Against the backdrop of the challenges of governing common pool problems, governance of systemic risks needs to explore other paths and their intersections. Governance via networks seems a promising road to travel. Governance networks operate at any level in a multi-level governance framework at various degrees of institutionalisation (Torfing, Reference Torfing and Levi-Faur2012). Governance networks provide a platform for stakeholder engagement, thus facilitating participation (Hirst, Reference Hirst and Pierre2000). Participatory and deliberative approaches in support of cross-national and global governance can achieve a number of distinct goals, e.g. ethical and epistemic functions in addition to providing legitimacy (Bidwell & Schweizer, Reference Bidwell and Schweizer2021). However, those functions can come into conflict with each other which calls for a careful design of deliberative processes that takes into account different deliberative tasks and agents at different moments within a process (Beauvais & Baechtiger, Reference Beauvais and Baechtiger2016).

Furthermore, an important question is which meta-objectives inform the quest for management options. Resilience has been proposed as a guiding principle for risk management of systemic risks. The National Research Council (1996) defines resilience as ‘the ability to prepare and plan for, absorb, recover from, and more successfully adapt to adverse events’ (National Research Council, 1996, p. 16). Resilience acknowledges that systemic risks will materialise, and disruptions will happen. The objective of risk management then is to enhance systems' capacity for recovery and adaptation to ensure their survival and potential improvement through broader systemic changes (Hynes et al., Reference Hynes, Trump, Love and Linkov2020, p. 178f). While scholars refer to resilience as a yardstick for managing risks under conditions of uncertainty, the concept does not provide criteria for assessing which systems and functions should be sustained and which changes in the status quo would qualify as improvements. For instance, Trump and Linkov (Reference Trump and Linkov2020) raise these questions about ‘broader societal resilience to bounce back from the range of outcomes afflicting public health, economics, and general societal harmony and well-being’ (Trump & Linkov, Reference Trump and Linkov2020, p. 173) in the context of the COVID-19 pandemic.

Suggestions have been made for rephrasing the resilience concept to accommodate normative meta-objectives such as equity (e.g. Logan & Guikema, Reference Logan and Guikema2020) and sustainability (e.g. Berkes & Ross, Reference Berkes and Ross2016) (for an in-depth view on the relationship between resilience and sustainability, see e.g. Derissen et al. (Reference Derissen, Quaas and Baumgärtner2011) and Renn (Reference Renn2020)). However, the relationship between resilience and sustainability when applied to risk management remains a contested issue (Renn, Reference Renn2020). The Brundlandt report published in 1987 defines sustainable development as ‘development that meets the needs of the present without compromising the ability of future generations to meet their own needs’ (United Nations, 1987). Thus, sustainability is about safeguarding the needs of present and future generations all over the world which puts equity and participation front and centre for sustainable development (Grunwald, Reference Grunwald, Newig, Voß and Monstadt2009). Thus, both resilience and sustainability aim to establish humane living conditions for present and future generations. Yet it has been argued that sustainability is the wider-ranging concept as it also strives for intra- and intergenerational justice and calls for due democratic processes of decision-making based on moral principles (Renn, Reference Renn2020, p. 6466). The latter two issues especially require stakeholder and public engagement for assigning and justifying trade-offs between management options. The COVID-19 pandemic has demonstrated that current societies are prone to systemic risks. Resilience and sustainability should guide recovery from the pandemic and adaptation to systemic risks. Thus, this article aims to contribute to resilient functionalities for sustainable societies by proposing an adaptive and participatory risk governance framework for systemic risks.

3.3 A procedural framework for systemic risk governance

Reflection, iteration, inclusion, transparency and accountability serve as guiding principles for the procedural aspects of governing systemic risks. Procedural and analytical dimensions need to be considered for governance of systemic risks. The risk governance framework by the International Risk Governance Council (2017) serves as a reference point for merging both dimensions together (Figure 1). On the one hand, the risk governance framework combines quantitative risk analysis with qualitative assessments. On the other hand, the risk governance framework includes stakeholder participation and public engagement, thus offering a comprehensive (regarding facts) as well as inclusive (regarding values) risk governance framework. The risk governance framework consists of four consecutive phases, i.e. pre-assessment, appraisal, characterisation and evaluation, and management. All phases can be iterated, if necessary. Furthermore, they are linked by continuous modes of stakeholder engagement and public participation. Thus, the risk governance framework provides a platform for implementing the procedural dimension of governance for systemic risks. A system analysis as sketched out above must be performed in each phase to integrate the analytic dimension into the governance process.

Figure 1. Risk Governance Framework, adapted from IRGC (2017).

The pre-assessment phase aims to provide a comprehensive understanding of the risk problem. In this initial phase, system analysis is crucial. At this stage, the analysis need not yet be overly concerned with the interrelations of systems I, II and III. Yet attention should be paid to the underlying assumptions of demarcating system boundaries because they will carry over to the following phases, thus framing the process and its outcomes. Ideally, transdisciplinary teams, which include an interdisciplinary set of scholars and experts as well as stakeholders, reflect on system boundaries and underlying assumptions. These transdisciplinary teams also select conventions and procedural rules for ‘the handling of distributional effects which may cover inter-individual, inter-group, regional, social, time-related and inter-generational aspects’ (Renn, Reference Renn, Renn and Walker2008, p. 13). The COVID-19 pandemic demonstrated that issues of gender inequality (Czymara et al., Reference Czymara, Langenkamp and Cano2020) as well as educational inequality (Andrew et al., Reference Andrew, Cattan, Costa Dias, Farquharson, Kraftman, Krutikova, Phimister and Sevilla2020) are reinforced during lockdown. One of the lessons to be learnt from the pandemic is to pay attention to distributional aspects of governance measures and to decide upon rules for dealing with these issues in the earliest stages of the governance cycle. Scholars also point out distributional issues and injustices with regards to climate change impacts (Pellow, Reference Pellow2018; Whyte, Reference Whyte2020), some of which could be addressed during the planning of adaptation to climate change (Juhola et al., Reference Juhola, Filatova, Hochrainer-Stigler, Mechler, Scheffran and Schweizer2022a, Reference Juhola, Heikkinen, Pietilä, Groundstroem and Käyhkö2022b).

The appraisal phase comprises risk assessment, i.e. hazard identification and estimation, appraisal of exposure and variability, as well as risk estimation, and concern assessment that investigates risk perceptions, societal concerns and socio-political impacts. Risk appraisal includes quantitative, as well as qualitative forms of assessments. Approaches based on multi-model frameworks, such as scenario building, can be deployed to analyse cascading effects and tipping points of systemic risks. Concern assessment calls for participation of stakeholders and the public to investigate perceptions and map out societal concerns. This stage focuses on system analysis at all three levels of system I, II and III; with risk assessment targeting system I and II levels and concern assessment targeting the system III level. The results from both risk assessment, which provides an estimate of the physical harm (this may also include financial losses if the systemic risk originates in the economy), and concern assessment, which investigates the social and (wider) economic implications, need to be integrated for a comprehensive appraisal of a given systemic risk. The COVID-19 pandemic has shown that priority is given to risk assessment while concern assessment takes a backseat. This was understandable in the initial stages of the pandemic when governments needed to act under time constraints and political pressure. However, comprehensive, and systematic concern assessment during the lull between the first and second wave of the pandemic would have unveiled much of the dissatisfactions and concerns that inhibited the success of many governance measures during the second wave. For climate change risks, existing studies stress the need to engage stakeholders in participatory processes to better address and account for the uncertainty of risks in planning and to facilitate social learning (Döll & Romero-Lankao, Reference Döll and Romero-Lankao2017).

The next phase consists of risk characterisation and risk evaluation. The available knowledge regarding systems I, II and III (or lack thereof) is sorted in categories of complexity, uncertainty and ambiguity. Risk evaluation is concerned with judgements on the tolerability, acceptability and the need for risk reduction measures based on the results garnered from knowledge characterisation. The inherent normative dimension of these judgements requires an inclusive approach which reveals and considers the plurality of societal values. Thus, risk characterisation determines the evidence-based component and risk evaluation determines the value-based component for making judgements on the tolerability and acceptability of risks (Renn, Reference Renn, Renn and Walker2008, p. 31). Both components are necessary for making this judgement. Evidence from the handling of the COVID-19 pandemic in Germany seems to suggest that the government tends to rely operationally on the evidence-based component in its decision-making whereas the value-based component is left to the (more time consuming) parliamentary debate. Although this approach adheres to the separation of power between the legislative, executive and judicial branches, the result leads to an asynchronous handling of risk characterisation and risk evaluation in this case. One potential way out of this structural conundrum would have been to conduct citizen assemblies or similar deliberative formats during the lull between the first and second waves of the pandemic to investigate the value-based component.

The results of knowledge characterisation and risk evaluation are channelled into the final phase of risk management. Risk management aims at the design and implementation of actions required to avoid, retain, reduce or transfer risks. This phase consists of decision-making and implementation. The results of risk characterisation and risk evaluation as well as risk assessment and concern assessment form the input material for the selection of risk management options (Renn, Reference Renn, Renn and Walker2008, p. 32). As the SARS-CoV-2 virus cannot be eliminated, people's exposure to the virus needs to be restricted, e.g. by wearing masks and social distancing, and vulnerabilities need to be reduced, e.g. by vaccination and increased capacities for intensive care. Resilience-focused strategies seem to be the method of choice for dealing with the pandemic and other systemic risks. They attempt to enhance the risk-absorbing system to better cope with potential risks (Linkov et al., Reference Linkov, Bridges, Creutzig, Decker, Fox-Lent, Kröger, Lambert, Levermann, Montreuil, Nathwani, Nyer, Renn, Scharte, Scheffler, Schreurs and Thiel-Clemen2014; Renn, Reference Renn, Renn and Walker2008). These strategies aim to establish best practices and analytical support to prevent unmitigated disasters although it is impossible to predict where, when or even whether such a compounding crisis may occur (Trump & Linkov, Reference Trump and Linkov2020). However, the emergence of polarisation and differences in normative beliefs and values in society cannot be reconciled by scientific risk assessments (see Hamilton, Reference Hamilton2024). As personal norms are key to action (Bouman et al., Reference Bouman, Steg and Dietz2021), risk management therefore also requires a broader societal discourse that is aimed at making ambiguities explicit and finding acceptable risk–risk trade-offs, which have also been recognised in relation to climate change (Sharifi, Reference Sharifi2020). For COVID-19, this societal discourse could have taken place in the lull between the first and second wave of the pandemic as scientific risk assessments had been available by then and the societal impacts emerged to be tremendous. The response to the pandemic ideally would comprise a combination of epistemic discourses with experts and agency staff to settle cognitive conflicts and advance risk science, reflective discourses with agency staff, experts as well as stakeholders which will settle evaluative conflicts and develop resilience-focused strategies, and participatory discourses including civil society to reconcile normative conflicts in deliberative formats. Similar calls have also been made for climate change with an emphasis on integration of policies from different sectors and action (Li et al., Reference Li, Wang, Zhao and Qi2021).