Keywords

5.1 Transformation of Critical Infrastructure Systems

Complex systems, such as energy or transportation systems, constitute vital components of modern society’s critical infrastructure. It is imperative to maintain their stability, even during times of crisis or significant transformation, such as sustainability transformation (Orwat 2011; Engen and Morsut 2023).Footnote 1 The transition to a sustainable state presents a complex issue, as safe and reliable operations must be guaranteed not just after the transition, but also during the phase of coexistence between old and new elements, such as electric vehicles and cars with combustion engine or electric generation devices operated partly by fossil fuels and partly by renewable energies.

Hence, uncertainty prevails during this transition period, and it is challenging to foresee individuals’ actions and reactions toward changing circumstances or political interventions, as well as the system dynamics that may result from uncoordinated individual actions. Will a substantial number of people opt for electric cars or heat pumps and thereby inadvertently cause increased volatility risks to the energy system? Will there be sufficient charging stations for electric vehicles, and how many will be required during each phase of the transition? The availability of charging stations (and affordable charging rates) may prompt individuals to consider buying an electric vehicle. However, the low demand for EVs could hinder providers from installing charging stations if there is no economically viable business case. This situation resembles the classic chicken-and-egg problem. Similarly, how many privately owned and operated photovoltaic devices can be integrated into the electric grid while simultaneously feeding in large amounts of electricity on a sunny day?

Numerous tensions can emerge between safety and sustainability, as well as between stability and change (Nawaz et al. 2019; Agora Energiewende 2019). To manage distributed systems effectively in the future, sophisticated real-time governance measures will be necessary (Weyer 2019). For instance, demand-side management of energy grids can balance out supply and demand based on the available capacities in power consumption and production (Paulus and Borggrefe 2011; dena 2012).

5.2 A Sociological Perspective

The queries posed above could be addressed solely by engineers. However, since the dynamics of complex infrastructure systems depend not only on technical parameters but also on people’s actions, a sociological viewpoint might help to comprehend the social dynamics of complex socio-technical systems. Some individuals may behave as early adopters, swiftly embracing new options, while others may resist rapid change or opt to wait before adopting novel alternatives. Most reports cited above do not consider the role of human behavior but instead rely on aggregated data, such as load profiles in the energy grid.

The question at hand is how to assess or predict the ability to execute sustainable transitions that uphold safe and reliable operations, particularly during critical phases that involve numerous uncertainties. Is it possible to peer into the future and assess the effective functioning of complex socio-technical systems with respect to sustainability, safety and reliability?

5.3 Agent-Based Modeling of Complex (Adaptive) Systems

Agent-based modeling (ABM) has emerged as a novel approach to studying the structures and dynamics of complex, adaptive systems, including the economy and climate (Resnick 1995). ABM permits the creation of artificial systems on a computer screen and experimentation with various what-if scenarios. Running a complex system on the computer enables the observation of nonlinear interactions, their impact on system dynamics and possibly the chaotic behavior of systems that evolve unexpectedly and instantly enter a new and irreversible state (Richter and Rost 2004). Compared with real-world experiments (cf. Bleicher 2023), this approach avoids endangering society and helps coping with uncertainties by evaluating sustainability transformation strategies before their implementation.

An illustration from climate research demonstrates a feedback mechanism, which arises from nonlinear interactions and ultimately leads to an irreversible process (cf. Fig. 5.1).

Fig. 5.1
A block diagram has the following flow, C O 2 emissions, global warming, and permafrost soil unfreezes with a loop to global warming via C O 2 emissions.

Nonlinearity—the example of permafrost soils. Source Author’s own work

Full size image

Global warming results in increased temperatures, particularly in regions like Siberia where permafrost soil thaws and emits additional greenhouse gases, exacerbating global warming. At certain tipping points, these processes accelerate in a self-dynamic way and become irreversible. Such developments cannot be comprehended solely through linear thinking but require concepts from complexity research.

Similar effects may arise in socio-technical systems, such as traffic congestion or energy system blackouts, which are nonlinear results from the aggregated behavior of numerous individual actors. No actor intentionally causes these effects, but rather contributes to them through their individual decisions, such as purchasing radiators to cope with high gas prices during the current energy crisis.

5.4 ABM of Socio-technical Systems

Social scientists have also adopted this method to investigate the dynamics of social systems, such as the spread of rumors, infections or innovations (Epstein and Axtell 1996; Van Dam et al. 2013; Weyer and Roos 2017). Agent-based modeling (ABM) enables the inclusion of diverse and heterogeneous social actors. Additionally, performing simulation experiments with artificial social systems allows social scientists to explore potential paths to the future and examine various policy intervention strategies.

Implementing a sociological model of a socio-technical system, such as the transportation or energy system, requires three key components:

  • agents who represent typical actors and their decision-making rules;

  • the contextual framework, including social, technical, political and institutional structures that define the boundary conditions for the actions of agents;

  • rules for interaction between agents themselves as well as their context.

5.4.1 Agents

Agents have properties, preferences and strategies, which resemble real actors. Data needed for modeling agents is mostly gathered by means of surveys (cf. Postmes et al. 2023). This method makes it possible to construct typical agent types, such as eco-friendly or comfort-oriented agents. With modern simulation software, one can parametrize each agent differently (referring to age, sex, income, agent type, preferences, routines, car ownership, daily tasks, etc.), so that large populations of heterogeneous agents can be generated for experimentation at the computer screen.

The decision rule in models of artificial societies is simple. When faced with alternative options, such as taking the bus, car or bike, agents select the option that best aligns with their individual preferences (Konidari and Mavrakis 2007). The concept of subjective expected utility (SEU) encompasses situational parameters, as well as personal expectations and preferences (Esser 1993; Velasquez and Hester 2013). In a comparable situation, the environmentally conscious individual would possibly opt for cycling, while the comfort-seeking individual favors driving a car.

5.4.2 Context

The so-called landscape is the second component of a simulation model, which is necessary for agents’ movement. Its configuration is influenced by the area of inquiry. Typically resembling a checkerboard, an infrastructure network often consists of nodes, representing residential buildings, workplaces, crossroads, train stations, bus stops, as well as edges that connect them, like roads, bike tracks or public transport railways. Together with available technologies such as cars, bicycles, public transport, car sharing and more, this context shapes the possibilities and limitations for all individuals, offering opportunities such as nearby bike rentals, while also imposing restrictions, such as the prohibition of cycling on highways.

Every contextual element has properties, some of which are “natural”, such as the maximum number of cars permitted on a residential road, and others that are politically defined, such as the limit on CO2 emissions or the amount of city toll charged for that road. These properties provide policymakers with a significant tool for intervention, including the option to increase the city toll for combustion engine cars or, ultimately, ban their use.

The same holds true for technologies, which exhibit specific characteristics, including bikes’ low pollution levels in comparison with cars as well as their lower speed. These characteristics can be altered by political policies such as implementing speed limits on cars or introducing new technologies such as the e-bike that enhances the bike’s speed and range.

5.4.3 Interaction

The last aspects of an ABM include rules for the interactions between agents, but also between agents and context. In a transportation system, agents typically adhere to their individual lanes and sustain a safe distance when approaching other agents. Roads can affect agents through various mechanisms, such as speed limits or tolls. In turn, agents not only occupy a road for a short period of time, but their presence also leads to wear and tear of the road alongside polluting the environment.

Every transportation system user affects system dynamics by altering parameters such as the number of cars on a given road section. Therefore, they have an indirect impact on other users who may opt for public transport when roads become congested.

5.4.4 System Dynamics

A multitude of autonomous actions, influenced by the present system state at time t, leads to a self-organized system dynamic. The emergent outcomes of this process are hard to predict but make up the subsequent system state at time t + 1. Agent-based modeling can depict the dynamic interaction between the micro-level (agents’ actions) and the macro-level (system state). The outcomes of these actions, such as traffic congestion, may be unforeseen and are not included in the agents’ strategies, but emerge as a nonlinear product of their autonomous and uncoordinated actions.

5.4.5 A Sociological Perspective

This approach to modeling complex infrastructure systems may resemble the methods employed by engineers when analyzing the causes of traffic congestion (Schreckenberg and Selten 2013). Nonetheless, from a sociological standpoint, it is crucial to avoid treating human agents as mechanical components that behave identically in a perfectly rational manner. Rather, they must be regarded as conscious individuals who act in accordance with personal preferences. Sometimes, decisions may appear irrational, such as taking a car for a short one-kilometer trip. However, these everyday practices are important to consider when attempting to understand the dynamics of socio-technical systems by analyzing the interplay between the micro- and the macro-level.

Sociological theory of action and macro–micro–macro models are essential components in creating artificial societies that represent real societies, particularly in cases of sustainability transformation (Hedström and Swedberg 1996; Ostrom 2010; Esser 1993).

5.5 Simulation of the Governance of Complex Systems (SimCo)

The simulation framework SimCo has been developed at TU Dortmund University, starting in 2012. Its primary aim is to promote and advance governance research, which previously relied heavily on case studies and was limited by the “governance trap” (Grande 2012). According to Grande, this trap resulted from a lack of understanding of social mechanisms that constitute social systems and enable external influences.

Focusing on governance issues, SimCo does not address physical details, such as the dimensions and lengths of bus stops, and instead puts emphasis on social mechanisms that shape and influence individual behavior (Adelt et al. 2018). Thus, the network, representing a transportation system, comprises nodes and edges, with freely programmable dimensions (as stated above). As a general-purpose framework, SimCo aims to explain the dynamics of systems resulting from the interaction of heterogeneous agents that make autonomous decisions—and conversely, to explain agents’ behavior as an outcome of their individual preferences and situational constraints. SimCo is one of the efforts to systematically translate a macro–micro–macro sociological model into an agent-based model (Esser 1993).

SimCo has been utilized for various experiments on risk management and sustainability transformation, primarily involving road transportation (Philipp and Adelt 2018; Weyer et al. 2019, 2020). Several what-if scenarios have been investigated, analyzing the effects of external interventions on the individual transport mode and route choices of various agent types. The primary outcome of these experiments is this: the most effective strategy for political interventions, such as mitigating risks of congestions or emissions and promoting system change toward sustainability, is to adopt the governance mode of soft control. This mode uses incentives rather than harsh measures such as bans that are typical of strong control (Weyer et al. 2020).

5.6 Experiments

Two experiments, one relating to transportation and the other to energy, will illustrate the value of the ABM approach. The initial experiment will highlight the relation between sustainability and social acceptance, while the subsequent one aims to examine the balance between sustainability and safety.

5.6.1 Political Regulation of Urban Transportation

Several political strategies for regulating urban transportation have been considered, including enhancing bicycle comfort (hypothesis H1), increasing the number of bicycle tracks (H2), implementing a city tax for cars (H3), introducing speed limits for cars (H4) or reducing prices for public transport (H5). These options have been examined incrementally by adjusting relevant parameters, represented on the x-axis of Fig. 5.2 (Philipp and Adelt 2018). This graph illustrates the effects of all five measures on (mean) emissions, represented on the y-axis of Fig. 5.2.

Fig. 5.2
A multi-line graph of emissions short versus depth of intervention. It plots the H 1 bicycle comfort increase, H 2 bicycle edge increase, H 3 car costs increase, H 4 car speed reduction, and H 5 P T costs reduction lines with linear and decreasing trends. The depth of intervention also has a table with values.

Effects of political regulation on emissions in transportation. Source Philipp and Adelt (2018: 44)Footnote

Reproduced with permission. This figure is excluded from our open access license.

Full size image

Furthermore, the individual satisfaction of agents in accepting the five options was measured by calculating their subjective expected utility (SEU) of the journeys traveled (represented on the y-axis of Fig. 5.3).

Fig. 5.3
A multi-line graph of S E U versus depth of intervention. It plots the intersecting bike comfort line with a decreasing trend and fluctuations, the road taxation and speed limit lines with decreasing trends, and 1 more line with a linear trend. The depth of intervention also has a table with values.

Acceptance of various policy measures. Source Philipp and Adelt (2018: 46)Footnote

Reproduced with permission. This figure is excluded from our open access license.

Full size image

As depicted in Fig. 5.2, there is no discernible effect from either decreasing public transport tariffs (H5) or expanding the bicycle network quantitatively (H2). However, implementing a speed limit (H4) proves effective but faces significant challenges regarding public acceptance (cf. Fig. 5.3).

Surprisingly, a minor improvement in cycling comfort (H1), through initiatives such as bike storage facilities, traffic signal priority and charging stations, has a significant impact comparable to speed limits. In addition, these measures are relatively inexpensive and easy to implement, and public acceptance is high (cf. Fig. 5.3). Finally, a city tax (H3) is an effective measure that is more accepted than other alternatives.

However, the two tipping points (highlighted by the blue arrows in Fig. 5.2) are most intriguing as they indicate a nonlinear progression. These tipping points can be explained by the behavior of distinct agent groups implemented in SimCo. Certain agents, who have a general preference for cycling but are discouraged by the lack of comfort, immediately switch if, for instance, secure storage options are available (a 10% increase). If cycling were as comfortable as driving a car, for instance, through covered and heated cycle paths during winter and free e-bike rentals (resulting in a 90% increase), it is likely that other groups would switch, at least in this hypothetical scenario. This may be an improbable assumption, but it demonstrates the significance of a sociological perspective, which considers the diverse actions of various agent groups.

To better understand the impact of policy measures, it is important to consider that different groups of agents may react in varying ways. This can lead to unforeseen, aggregated effects that are difficult to interpret as not all agents alter their views simultaneously.

Complex socio-technical systems typically involve nonlinear interactions, which researchers can explore by conducting experiments using computer models that are grounded in a sociological theory of action. Understanding the results of such experiments can be challenging, as they are often not easily comprehensible using linear thinking.

5.6.2 Demand-Side Management in the Energy System

The second experiment was conducted using a simulation of the power distribution grid in a small, sparsely populated residential area with 167 households comprising detached and semi-detached houses (Hoffmann et al. 2020). The agent population was composed based on existing energy end-user typologies (i.e., households). The diffusion of photovoltaic (PV) systems was assumed to be rather high in this future scenario, leading to an increase in power generation volatility (cf. Table 5.1).

Table 5.1 Shares of end-user types and building modernization for 167 householdsFootnote

Adapted from Hoffmann et al. (2020), released under a CC BY 4.0 license.

Full size table

Three experiments were conducted with different modes of governance: decentralized self-organization; distributed, soft control; and centralized, strong control. The latter modes represent two different demand-side management (DSM) concepts: maintaining grid stability and reducing fluctuation risks through financial or other incentives for end-users (soft control) or direct system operator access to controllable devices like PV systems, battery storages and heat pumps (strong control). In comparison, the former mode represents a base scenario of self-organization through independent and uncoordinated decisions of energy end-users.

The aim of these three experiments was to assess the extent to which interventions can enhance system stability on a macro-level. The experiments were conducted over a seven-day period, and the cumulated load of households (in kW) was used as a macro-level indicator to evaluate the effectiveness of the three modes of governance.

A two-day section of the measurement series is illustrated in Fig. 5.4. The black curve (decentral self-organization) shows a feed-in peak at noon of the first day, while no such weather-related generation occurs on the second day. On both evenings, the load increases clearly since electricity consumption is higher during this time of the day.

Fig. 5.4
A graph of cumulated load versus time. It plots the overlapping decentral self-organization, distributed soft control, and central strong control waveforms with intense fluctuations, and peaks and dips.

Cumulated load in kW (two-day section) over time, comparing three modes of governance. Source Hoffmann et al. (2020)Footnote

Reproduced from Hoffmann et al. (2020), released under a CC BY 4.0 license.

Full size image

Interventions aimed at reducing feed-in have some impact by decreasing the total absolute value of the load and resulting in a decrease in grid fluctuations. Centralized and strong control measures exhibit slightly superior results, although differences between central and distributed control are minimal.

Figure 5.5 shows the violin plots that depict the distribution of measured values over the entire duration of seven days. These anomalies are important for assessing the stability of the grid: the interventions, regardless of their type, proved effective in mitigating outliers, i.e., peaks of low or high consumption and feed-in.

Fig. 5.5
A positive and negative violin plot of cumulated load versus decentral self-organization, distributed soft control, and central strong control in decreasing order. Decentral self-organization has the longest load range, while central strong control has the shortest load range.

Cumulated load in kW. Source Hoffmann et al. (2020)Footnote

Reproduced from Hoffmann et al. (2020), released under a CC BY 4.0 license.

Full size image

An additional statistical analysis (not included here) confirms that soft control is adequate and can enhance the stability of the local grid. A considerable proportion of end-users are willing to respond to soft interventions. Therefore, central, strong control should only be utilized in rare circumstances, for instance, when previous soft control efforts have not yielded satisfactory results, and there is an imminent danger to the system’s stability.

5.7 Conclusion

To study the dynamics of complex socio-technical systems, it is necessary to examine not only the technical aspects but also the social components. These social components include real people and their everyday mobility and energy behavior. A sociological theory of action can be used to model these behaviors, which covers people’s subjective decision-making. Other models may also be applicable. Agent-based modeling and computer-based experiments using a sociological system model could enhance our understanding of the interrelation between safety and sustainability, as well as how agents might respond to political interventions aimed at promoting sustainability.

The initial urban transportation experiment highlights the interrelation between sustainability and social acceptance. This relationship is crucial in ensuring safe operations on a broader scale. We have not calculated the number of accidents or congestion length, although it could have been done, too.

The second energy grid experiment has highlighted the requirement for establishing a balance between management approaches and behavioral adjustments to ensure safe and sustainable energy system operations.

ABM therefore is a valuable tool for investigating these complex issues. However, models should not be considered as perfect copies of reality, but rather simplified representations that facilitate experiments with different future scenarios—based on desired goals or states for complex systems. Consequently, agent-based modeling offers a means to test the assumptions behind these scenarios and evaluate the plausibility of transformation pathways regarding safety and sustainability.