· · · Institutional Repository

In many cases, congestion is a recurrent problem that exists in electricity lines since the demand, in general, steadily increases whereas lines capacity cannot increase in small increments but only in multiple units of line capacity. Congestion in electricity lines hampers the interconnection of local markets because the interconnection line capacity is lower than the optimal. Hence it limits the European Union goal to form a unique electricity market while liberalization is spread. Moreover, congestion could jeopardize the benefits of liberalization, like it happened in California and in England, where Producers manipulated the market by bidding strategically, in other words, they withheld capacity and/or ordered higher prices than marginal cost. A better understanding of the new liberalized electricity market in order to support decision-makers to formulate sound policies is thus needed. Given the complexity of these markets, simulation seems to be a promising tool to provide an understanding thereof

Agent-based modelling is a popular and suitable tool for exploring the possible states of so-called socio-technical systems. These systems consist of both technical artefacts (the physical infrastructure, e.g. pipelines), and many social artefacts (relevant actors and institutions, e.g. end-users and governments), which are intertwined with each other and strongly interact (De Bruijn & Herder, 2009). The quality of the model output strongly depends on the quality of the rules of an agent, i.e. the lines of code that describe how an agent behaves. Slightly different rules on agent-level, may lead to significantly different outcomes on system-level (Bousquet, Cambier, Mullon, Morand & Quensiere, 1994; Levine & Fitzgerald, 1992). Thus, valid rules for agents are crucial for a valid analysis of socio-technical systems as a whole. When modelling a socio-technical system with an agent-based model, some agents represent social artefacts, and thus must simulate real-life behaviour. However, in many cases, rules that describe social phenomena, are not based on empirically tested, theoretical models, and agents display unrealistically simplistic behaviour (Jager & Janssen, 2003). This restricts the analysis with respect to social behaviour, and may even lead to an invalid system analysis. In earlier research it is suggested that gaming simulations can be used to improve the realism and diversity of agent behaviour. However, this application of games has not been examined extensively. This research aims to acquire insight in whether this application of games is possible and feasible. The central research question thereby is: To what extent can gaming contribute to the definition of realistic behavioural rules for agents in an agent-based model, within the context of modelling socio-technical systems? An extensive literature research shows several problems and challenges with agent-based modelling. These include fundamental problems with the currently used methods used for gathering information about realistic behaviour (i.e. interviews and literature research). Several characteristics of gaming can help to reduce some of these these challenges, providing theoretical evidence that there is a potential for synergy between the two methods. Based on a structural comparison of the methodological processes of agent-based modelling and gaming, it appears that there are several possibilities in which gaming can contribute to the definition of realistic behavioural rules of agents. In this thesis we elaborate on using gaming as a data collection tool. The gap between global knowledge about how agents behave and the implementation of precise rules in models is large. In case games can function as a valid data collection tool, the collected data can function as a basis for rules, which helps to overcome, or at least to reduce, design and formalization problems. When a game is used as a data collection tool, there are three basic requirements that must be met: one must be able to generate valid data, to measure the desired data, and to analyse the data. Several aspects with regard to fulfilling these goals, affect the validity and the costs of the data collection tool. Furthermore, choices within one design (game design, design data collection, or design data analysis), may have implications, direct or indirect, on other design choices, both within that design and in the other two. This interconnectedness makes the design process of the data collection tool as a whole very complex. Neither the game, nor the data collection method, nor the data analysis method is per definition leading in the decision which is the most suitable design. Whether, and how much the data collection tool can contribute to the definition of realistic behavioural rules, is very context dependent. The proposal for using games as a data collection tool, and the proposed design framework have been done solely on a theoretical basis. The development of a useful and feasible tool, however, should also have a decent practical basis. Therefore, this work should be seen as the first of many research projects on this topic. The lessons learned from the application of this tool can and should be used to improve the proposed design framework.

As goals are set to increase the share of renewable energy production within the province of Overijssel, the Netherlands, the project group Groen Gas Salland is founded and has taken the initiative to explore the opportunities of green gas production through a biogas infrastructure within the region of Salland. Since within the region of Salland (intensive) livestock farming is practised, it is assumed that the utilisation of manure for the production of green gas by means of the anaerobic digestion process has a considerable potential. To comprehend whether the manure can be made available for energy production, it is necessary to learn how manure is currently used and valued by the local farmers within Salland. Due to intensive livestock farming the manure production exceeds the local demand for manure-based fertilisers, which is considered a problem as this (local) abundance of manure urges for a manure distribution system of which the costs are high. Furthermore, circumstances within the (intensive) livestock farming sector are changing especially due to the amendment of policies that monitor farming activities. Changes within these institutional rules and especially the perceptions that these changes occur unpredictably, affect the decision making of local farmers and will influence the condition that underlie the manure distribution system. In order to explore the potential for manure-based energy within this complex system, we developed an agent-based model by means of the MAIA framework. Based on our research we found that the energy potential is low due to several related causes. A clear expectation of what will be gained is lacking due to uncertainties about both the amount of subsidy that will be granted and the order of the specific technology costs. Furthermore, since a sufficient production capacity is required, many small-scale to middle-scale farms will not consider manure-based energy production. A considerable investment capital is required as well, this puts a lot of pressure upon the farmers as they often experience at the same time poor financial situations and ongoing changes in the policies. We state that farmers do not obtain a clear benefit concerning the production of manure-based energy as current issues with respect to the abundance of manure are not solved and meanwhile the produced digestate, almost identical in volume and composition to manure, had to find its way back to the manure distribution system. Although cooperation between farmers should not be taken for granted, we found in case farmers do cooperate, the potential for manure-based energy production to be highly increased. In order to increase the potential, we recommend to create a clear benefit from a farmer's point of view, increase their understanding with respect to manure-based energy and reduce institutional barriers, especially with respect to regulations that control subsidy.

Julia Kasmire is writing her PhD dissertation on technology transitions in the Westland greenhouse area. She is looking whether Universal Darwinism (UD) can explain these processes and could lead to could policy recommendations. As a prelude to her research this report focuses on, the less ground braking, technology diffusion in the Westland greenhouse area. UD is not taking as a point of departure to be able test without prejudice whether the concept holds. In this report an answer is given to the following main research question: How does a technology diffuse through a network of greenhouse farmers in the Westland and which factors are important to this process? This is done by answering three sub questions. The first sub question is the following: What methodological individualistic framework on technology diffusion in the Westland can be developed? To answer this question first an overview is given of the literature on technology diffusion from which a preliminary framework is constructed. Central in this framework are the decision individual greenhouse farmers make. Based on their own characteristics, the characteristics of the environment and the available information on the characteristics of technologies decisions are made. This framework has been taken as a point of departure for an interview with a greenhouse farmer and subsequently questionnaires were sent to other greenhouse farmers. After this the framework was enriched with the information retrieved from the interview and the questionnaires. Important results were the fact that greenhouse farmers may be considered boundedly rational, that they value information from peers quite high and that an innovativeness/risk attitude is a differentiating characteristic. How can technology diffusion in the Westland be modeled? For the second research question an Agent Based Model (ABM) was developed. The model combines the physical and social layers of technology adoption. In the physical layer technologies are acquired and used as well as products are sold. In the social layer agents communicate about their satisfaction of their currently owned technologies. Because the agents are heterogeneous and they can only communicate about sets of technologies (not the individual benefit of one technology), they send limited information. Depending on the characteristics of the technologies the model behavior may result in technology diffusion. This has been discussed when answering the last research question: Which factors are important to technology diffusion in the Westland? Four experiments with the ABM have been performed and discussed. The first two experiments show that, at standard settings, technology diffusion does occur for technologies that are beneficial for all companies. The second experiment shows that having a high percentage of innovators (10-20%) is slowing diffusion down at first but leads to higher diffusion rates in the long run. Valuing the opinion of peers highly leads to high diffusion rates in the short run but is not the best strategy in the long (when self obtained knowledge is the most valuable). The fourth experiment shows that talking to many peers is beneficial for overall diffusion but that this effect diminishes with an increasing number of peers with who is communicated. A general important conclusion for the third sub question is that time is a very important variable in technology diffusion. The characteristics of the greenhouse farmers have different effects at different times and this research cannot tell at which point in time one should look exactly.

The thesis research was conducted at the section E&I (energy and industry) of the faculty of Technology, Policy and Management. The world around us is complex and this complexity is only expected to increase as technological development spurs on. This complexity arises from social and technical parts that interact with each other and co‐evolve over time, resulting in emergent system behaviour and structure. Today’s industrial systems are interesting examples of socio‐technical systems that contain large aggregates of technical components and actors that make economic and strategic decisions, which shape the physical world. It is impossible to predict or steer future developments of these systems because of the non‐linear interaction and feedback loops between the different social and technical components. However, insight about the fundamental behavioural modes and possible evolutionary pathways can be gained by building a computer model. In this thesis agent‐based modelling (ABM) was used; in ABM, an agent represents an independent socio‐technical component in the system. It has a state that represents the physical and economic assets and a behavioural component in which reasoning and decision‐making takes place which in turn influence the state of the agent and interactions with the outside world. Maasvlakte II, the new extension of the Rotterdam harbour area that is currently under construction is an ideal case study to expand experience with ABM; the ambition of the Port of Rotterdam is to realise sustainable chemical facilities, like a synthesis gas cluster that uses different types of feedstock to produce a variety of clean end products, like fuels and value‐added chemicals. The problem with previously built ABMs is that the technical detail represented is too limited to apply it to syngas networks; the technical part of the agents is currently modelled as a black box and generic facilities are taken as a starting point. The question is whether this lack of detail makes a significant difference for the structure of the syngas cluster that evolves and the agent decision‐making. The main question of the research was therefore: “What is the added value of Agent based modelling of large scale syngas networks at Maasvlakte II with ‘grey‐box’ models representing syngas technologies instead of black‐box models?” The research method consisted of a literature study of the different syngas technologies and building the agent‐based model. Developing the ABM entails conceptualisation and formalisation of the syngas technologies in an ontology (the system decomposition method), making a storyboard of the agent’s behaviour and transcribing this to (pseudo‐)code. There was unfortunately not enough time left for full implementation and running of the model. The main added detail that was proposed is a description of the composition of the gasifier feedstock and syngas, and a more detailed description of technologies that may influence this composition; i.e. treatment units and reactors are described separately, and descriptions of reactors are diversified to several reactors that produce low or high quality syngas. The general conclusion from the research is that the expected added value of adding technical detail to the black box descriptions of syngas technologies is related to the increased realism of network connections, while unfeasible connections are ruled out by the constraints that are imposed by adding technical detail in describing different reactor types and including the syngas composition. The description of treatment units and impurities in the composition of streams also make it better possible to evaluate the sustainability of the system and economic viability of routes from feedstock to end product. The amount of work that should is involved in adding technical detail to the agentbased model is however considerable, and the added value is partly offset by the problems that were found during the research. The three main problems are: There is a myriad of different syngas technologies, and it is very easy to get lost in the details, while it is hard to conceptualise them and formalise them. This is because all detail levels are somehow connected. The major problem of this class was to decide how the mass balances should be dealt with; the optimal solutions lies in between using a few standard input and output compositions without calculating anything, or detailed calculation of the composition as is done with flowsheeting programs like ASPEN. The second main problem was that the technical detail added does also imply that agents will have to reason about the design criteria that are added, which means that the more technical detail is added, the more (pseuco‐)code has to be written to capture agent reasoning and decision‐making. The final main problem was that in some cases this reasoning and decision‐making is so complex that it can hardly be captured in (pseudo‐)code, for example strategic behaviour and contract negotiation. The largest danger is falling in the trap of (having the intention of) adding too much detail. The main recommendation is thus to define the level of detail per case study, and at a level relevant for the context of that case study; it is better to have a partial model than no model at all. For the research of syngas network evolution over time one should focus on the interface between syngas producing and syngas consuming technologies, which means including the composition of syngas and factors that influence this composition, like different reactor types and gas treatment technology. When evaluating the sustainability of a syngas network, one should also include pre‐treatment of gasifier feedstock. A second recommendation is to conduct a parameter sweep for certain variables of which it is unsure how they influence the decision‐making of the agents or when the behaviour of the agents is too complex to capture with simple (pseudo‐)code. The final recommendation to be made when conducting agent‐based modelling is to be aware of the fragile balance between too much and too few detail; the system should be defined in such an extent that the behaviour of the agents is restricted in a realistic way while leaving enough degrees of freedom for the system to exhibit complex and emergent phenomena. Future research should focus on solving the problems with mass balance calculations by taking a chemical engineering point of view and learning from how flowsheeting programs have tackled these problems previously, and figure out what the breath and depth of the interplay is between technological details, social decision‐making, and economic aspects of the model.

Our world is a complex socio-technical system-of-systems (Chappin & Dijkema, 2007; Nikolic, 2009). Embedded within the geological, chemical and biological planetary context, the physical infrastructures, such as power grids or transport networks span the globe with energy and material flows. Social networks in the form of global commerce and the Internet blanket the planet in information flows. While parts of these global social and technical systems have been consciously engineered and managed, the overall system-of-systems (SoS) is emergent: it has no central coordinator or manager. The emergence of this socio-technical SoS has not been without consequences: the human species is currently facing a series of global challenges, such as resource depletion, environmental pollution and climate change. Tackling these issues requires active policy and management of those socio-technical SoS. But how are we to design policies if policy makers and managers have a limited span of control over small parts of the global system of systems?

Infrastructure systems for energy, water, transport, information etc. are large scale socio-technical systems that are critical for achieving a sustainable world. They were not created at the current global scale at once, but have slowly evolved from simple local systems, through many social and technical decisions. If we are to understand them and manage them sustainably, we need to capture their full diversity and adaptivity in models that respect Ashby's law of requisite variety. Models of evolving complex systems must themselves be evolving complex systems that can not be created from scratch but must be grown from simple to complex. This paper presents a socio-technical evolutionary modeling process for creating evolving, complex agent based models for understanding the evolution of large scale socio-technical systems such as infrastructures. It involves the continuous co-evolution and improvement of a social process for model specification, the technical design of a modular simulation engine, the encoding of formalized knowledge and collection of relevant facts. In the paper we introduce the process design, the requirements for guiding the evolution of the modeling process and illustrate the process for Agent Based Model development by showing a series of ever more complex models.

Overzicht van producten en diensten van het Shared Service Centre ICT van de TU Delft in samenwerking met onderzoekers.

Exactly predicting the future of an evolving large scale socio-technical system is impossible. Yet, if we are to sustainably manage the industrial and infrastructure systems our society depends on, we must understand how the actions we take today will affect the evolution of these systems. Simulating how the social and technical networks co-evolve over time allows us to explore possible system futures. This knowledge can help today’s decision makers to steer the system away from undesirable evolutionary pathways. Creating models that capture the complexity of socio-technical systems co-evolution requires multiple formalisms to be encoded in a modeling framework that itself evolves. This thesis presents a method for creating Agent Based Models that suitably represent complex evolving systems. The method involves a co-evolution between the technical aspects of model development, the social process involving the stakeholders, the collection of relevant domain knowledge and the encoding of facts. Through seven case studies the method is demonstrated to yield subsequent generations of richer and ever more useful simulation models.

This thesis is concerned with the notion of Darwinism and its application outside of its traditional biological realm, to that of economics. The mechanism of natural selection is based on the co-­‐ existence of variational, hereditary and selective processes within a system (Cordes 2006). This context-­‐independent description of natural selection has been coined ‘Universal Darwinism’, the extent and practical relevance of which is not thoroughly understood. This thesis has aimed to contribute to the understanding of this practical relevance by applying a framework of Universal Darwinism to a specific economic context, that of the Dutch horticulture sector. The innovation and adaptive complexity in this industry makes it highly suitable to respond to the notions brought forth by Universal Darwinism. The competitiveness of the Dutch horticulture sector is under pressure and ways are sought to formulate policies to improve it. By identifying national legislative bodies as the primary stakeholder, the practical relevance of Universal Darwinism is derived from its role in illuminating a deeper understanding of the drivers of economic change and its interplay with the competitiveness of the sector so as to facilitate the formulation of more effective policies. Through an agent-­‐based simulation study that is thoroughly grounded in a Darwinistic ontology, answers are sought to these questions. The primary research question of this thesis has been formulated as: “How can the application of Universal Darwinism help inform policy formulation to improve the competitive position of growers in the Dutch Tomato Sector?” This research question covers a lot of ground and was identified as being the consolidation of three different themes: the Framework of Universal Darwinism, Competitiveness within the Dutch Tomato Sector and Sectoral Policy Formulation. On the one hand, sufficient insight into the framework of Universal Darwinism was required to understand how to apply it to the economy. On the other hand, the drivers of competitiveness for growers within the Dutch Tomato Sector were to be understood. Lastly, these two themes were to be combined within a sectoral policy formulation directed to improving the competitiveness of growers. Therefore, a deconstruction into three separate research questions was found to be appropriate. These were formulated as: “What is the structure of and what are the core drivers of competitiveness within the Dutch Tomato Sector?” (1), “How can a framework of Universal Darwinism be applied within the context of the Dutch Tomato Sector Economy?” (2) and “How do Sectoral policies, aiming to improve the competitiveness of the Dutch Tomato sector, perform and how does Universal Darwinism help interpret the insights achieved and shed light on policy formulation?” (3). In the methodological discussion, it was shown that the framework of Universal Darwinism showed considerable overlap with the ontological commitments of evolutionary economics. This allowed for a set of generalized conditions for evolutions could be identified that allowed for the transplantation of Darwinism to the economic realm. In this transplantation, three assumptions were made with respect to two themes. Firstly, ongoing difficulties exist around the appropriate ‘Unit of Evolution’ within economics. This thesis assumed as a unit of evolution a description of ‘company strategy’. Although many other definitions exist, this definition allows for an intuitive bridge between the behavior and selection of the growers in the model. Secondly, two assumptions had to be made with respect to the ‘Variability of the Selective Landscape’. If the environment of a company is too variable, both the description of evolution and the relevance of any policies derived from it would carry little meaning. These assumptions, together with the data requirements necessary for the model, led to a set of barriers to the practical relevance of Universal Darwinism within economics were constructed. For the current study goals, it was found that Agent-­‐based modeling is the appropriate methodology because it is able to build a model of the economy on Darwinistic foundations. In discussing ontology of an evolutionary agent-­‐based modeling, it was found that the ‘production of agent instructions’ was found to be the key link that distinguishes economic evolution from mere economic change in the economy. The model hypothesis were formulated as: “It is feasible to build an agent-­‐based 4 model within the framework of Universal Darwinism (thereby distinguishing economic change from economic evolution)” (1), “The model can reproduce the stylized facts observed in the Dutch Tomato Sector” (2) and “The model is able to give insights into the relationship between policy formulation and the competitive position of Dutch Tomato Growers” These model hypotheses were chosen in order to collectively help reflect on both the primary research question, as well as the barriers to the practical relevance of Universal Darwinism within the field of economics as a whole. An extensive empirical investigation subsequently looked at the empirical foundations of the model by investigating the structure and core drivers of competitiveness within the Greenhouse horticulture Sector. The first research question was answered in the form of five different Stylized Facts’ of this particular economy and an overview of threats and opportunities / policy measures available to enhance the competitiveness of the Dutch Tomato grower. The stylized facts related to differences in the particular production systems used within this economy, an observed trend of increasing concentration and scale in the market, a positive relationship between performance volatility and company size, the particular price and volume contracting behavior observed between growers and retailers and the importance of company innovation and knowledge exchange within the sector. In the model conceptualization phase, a connection was sought between the economic characteristics of the sector (structure and stylized facts) and the ontological foundations of an evolutionary agent-­‐based model. In addition to formulating model assumptions that set the ground for the model boundaries, the model objective and the concept of grower competitiveness were identified. Grower competitiveness was defined through the notion of the ‘selection of individual tomato growers’ in the model. In this definition, competitive Tomato growers stay ‘alive’ or ‘economically viable’ while uncompetitive Tomato growers go bankrupt or ‘die’ in the model. As such, the model objective was identified as giving insights into the relationship between policy formulation and the selection of individual Tomato growers. A set of 10 different model scenarios were constructed to help illuminate the relationship between policy formulation and the selection of individual Tomato growers. While some of these scenarios rested primarily on the notion of economic change, others included strong elements of economic evolution, thereby giving insights into the role of evolution in the economy. The base case model simulation simulated the stylized facts in the economy well and formed the reference model for subsequent model scenario comparisons. From these comparisons, conclusions were drawn. It was found that innovation stimulation for northern markets for the benefit of competitiveness can only be (initially) effective if the innovation is of such a type or sufficient measures are taken to limit or delay the transplant ability of innovation to southern regions of production while maintaining diffusion in the northern region. The effects of knowledge diffusion were very strong in the model, creating both opportunities and risks for the competitiveness of the northern grower. Further more, Resource and Credit support were found to be an effective policy mechanism to help improve the competitive position of Northern growers. Expansion stimulation , on the other hand, seemed to require very prudent investment and expansionary grower strategies to be effective. Northern growers appeared to do well in case the retailers were more selective in their pricing demands from the growers, indicating the power of innovation as a way of maintain close ties with retail partners. The simulation results led to two pieces of advice for stakeholders in the greenhouse horticulture sector: “Investigate the potential benefits of future sectoral innovation (both incremental in terms of production efficiency gains as well as radical in the form of new market creation). If there seems to be limited innovation possible in the sector, the competitiveness of Dutch Tomato growers runs a serious risk of declining in coming years (or be very costly to maintain) and competition with 5 southern growers is to be avoided using other means (such as assuring other markets than purely northern European ones).” Determine the ease of knowledge diffusion in the sector. The diffusion of knowledge regarding production efficiency depends on the barriers to transplantation that exist either intrinsically in the type of knowledge that is shared or the difficulty of incorporating such knowledge into effective business operations or extrinsically in the ease of information flow between different companies. In terms of study limitations, several barriers and model assumptions and limitations were identified. The most important two barriers were data availability and the environmental variability of an economic system. In terms of the first, several model assumptions and limitations had to be made with respect to different model components due to lack of data availability (for example in how companies change their reactive behavior towards their economic environment over time). In terms of the second, it was observed that data requirements increase as we are forced to take processes into account that are relevant or shorter-­‐time-­‐scales. As such, a balance needs to be found between the time scale of the model (from minutes to years to centuries), the data availability for the model (the shorter the time-­‐scales to be included, the more data required) and the validation of the model (leaving out too many descriptions of reality will limit the model validation). In the conclusions, the study findings were decontextualized from the particular economic context of this study and transplanted upon the research problem initially identified: “What is the practical relevance of US in economics?” Three practically relevant contributions of this a cross-­‐disciplinary bridge were identified: UD helps economists to “structure economic phenomena into a coherent system of interaction” (1), “Identify the scope and limitations of a model, thereby illuminating policy generalizability (2) and “Allow for cross-­‐discipline system comparison, thereby utilizing economic of scope within the realm of science” (3).

Horticulture, a pillar of the Dutch economy, has already achieved remarkable productivity increases through the use of natural gas for heating, lighting and CO2. Further innovative technologies that could aid the transition toward sustainable energy use, including heat/cold storage and deepgeothermal heat sources, are currently in development and spreading. However, there is a need to better understand the processes of technology diffusion in this industrial cluster to help stakeholders retain their competitive advantage and establish the best way to influence the energy future in the region and in the sector. This presentation discusses the experimental results of a series of agent based models of the greenhouse horticulture sector in the Netherlands, simulating the technological innovation decisions of greenhouse growers. Surveys of greenhouse growers suggest that innovation decisions are made on the basis of personal experience and information shared from other growers. In the model, each greenhouse grower must learn how to operate a greenhouse by evaluating their repertoire of technologies, exchanging information with other growers about their technological evaluations and purchasing new technologies to augment, expand or replace the existing selection. The interactions of greenhouse growers and the flow of information between them lead to emergent patterns, including diversity, adaption and complexity, in the diffusion of technologies throughout the community. These emergent patterns of diffusion indicate that technological innovations develop and spread according to evolutionary mechanisms, suggesting that influencing, supporting or advocating the diffusion of sustainable technologies in this sector must also follow evolutionary mechanisms. As an evolving system, the reality of technology, innovation and transitions may require new approaches to management that work with, rather than against, the properties of evolving systems. Survey results, horticulture cluster background, model design and simulation results will be presented and implications for regional industrial management are discussed.

The number of inverse infrastructures (Vree, 2003), that is, user-driven and self-organizing infrastructures, is rising and unsettling policies that are foremost tailored to deal with large-scale and centrally-governed infrastructures (Egyedi et al. 2012). To better understand and address this mismatch, Van den Berg (2012) has developed a complex adaptive systems (CAS) framework for analyzing inverse infrastructures. It is based on and well-fits CASs in physics, mathematics and biology. In this paper we explore the framework’s applicability to three inverse infrastructures, i.e.: Wikipedia (Davis and Nikolic, 2012), citizen-driven waste paper collection (De Jong and Mulder, 2012), and the user-driven roll-out of local glass fiber networks (Weijers, 2012; Nederkoorn, 2012). Applying it reveals that, while the framework’s most basic elements can be identified rather straightforwardly, other elements are often more difficult to identify in human CASs. Our tentative conclusion is that (i) the framework is a good starting point for analyzing inverse infrastructures, and (ii) more case studies are needed to fully understand the conditions under which self-organized emergent behavior of complex infrastructures can be observed.