6 Synthesis
6.3 Overall conclusions and future perspectives
6.3.1 Simulating the process of social mobilisation: Solution concepts “at work”
It is the genuine purpose of the method of ABSS to describe the processes underlying collective social phenomena and to scrutinise their implications by simulation. In this respect, the abstract ABSS HAPPenInGS‐A explored and validated the macro‐level consequences of the HAPPenInGS theory. The case study simulations HAPPenInGS‐N and SoNARe went a step further and investigated the effectiveness of typical solution approaches for public good dilemmas. In doing so, simulations start out from a model representation of the situation characteristics presently found in the respective case and explore possible futures by simulation. The initial conditions observed in both case studies correspond to a situation where public good provision is unsuccessful due to largely defective behaviour. The simulation experiments assess possible intervention scenarios which aim on changing the boundary conditions in the models systematically such that collective action may emerge. The results of such experiments contribute to the understanding of processes of social mobilisation in general and yield policy recommendations on promising intervention measures in particular.
The types of interventions investigated in the two case study exercises are diverse but they correspond to typical solution concepts for social dilemmas. The case of neighbourhood support considers information campaigns. The basic idea is that during information
campaigns individuals actively refine their knowledge about useful behaviours in the dilemma situation. HAPPenInGS‐N demonstrated when and how this individual‐level process results in the emergence of new social norms, new habits and subsequently in collective action. The case of land reclamation investigates the effect of financial incentives. SoNARe showed that limited but focused distribution of incentives best fosters collective action.
Clearly, the types of solution approaches considered in HAPPenInGS‐N and SoNARe match well the circumstances of the respective case. For the farmers in the Odra region financial considerations play an important role while the case of neighbourhood support focuses on voluntary behaviours which lack this economic dimension. However, the common conclusion for such HAPPenInGS models is that focused intervention is more effective that unselective intervention. This was shown for selective information campaigns in HAPPenInGS‐N and for focused compensation payments in SoNARe. Apparently, the mode of providing incentives to a population influences the success of an intervention irrespective of the particular kind of incentive.
For the case of neighbourhood support simulations point out that passive behaviour of hedonistic lifestyles inhibits full mobilisation under the modelled information campaigns.
However, there is some empirical evidence from an on‐going survey that especially these lifestyles react sensitively to material incentives that reward contributing to neighbourhood support. As a conclusion, selective incentives which parallel information campaigns could be a promising approach to achieve comprehensive mobilisation. To this end further modelling exercises are required.
6.3.2 Empirical grounding
HAPPenInGS‐N and SoNARe share a common theoretical embedding of individual decision‐
making in the HAPPenInGS theory which implies a strong similarity of the models in terms of the structures and processes they represent. However, the respective ABSS setups differ in the way they use either empirical data or assumptions to initialise structures and processes.
In HAPPenInGS‐N, the agent population is set up from large‐scale empirical data. This includes the initialisation of agent preference profiles, agents’ spatial distribution, and social networks. In addition, a regional climate projection is used as external driver. However, empirical research which could be used to set up the production function of neighbourhood
support was not available to the project. Therefore, the production function of the public good is largely based on assumptions derived from theory. Consequently, the induced dilemma structure is to an extent artificial and idealised in nature.
In contrast, due to the lack of available data, agent initialisation in SoNARe is largely based on assumptions. Nonetheless, the assumptions are non‐arbitrary as they are at least qualitatively related to empirical observations. In contrast, the environmental context of agent decision‐making is represented as a coupled hydro‐agricultural simulation model of the situation characteristics in the target region. The analysis of the induced “real‐world”
social dilemma (cf. section 5.4.2) showed that incentives are not as “cleanly” in line with the game‐theoretic conception of the incentive structure underlying HAPPenInGS‐N. To this end, the case of land reclamation illustrates well our theoretical point that social dilemmas emerge from the complex dynamics governing the temporal and spatial development of the environmental context of individual decision‐making and acting.
In particular for HAPPenInGS‐N a step‐wise refinement of the empirical grounding is a promising direction of future work. During the simulated time span external social drivers might gain importance. In the simulations agents “ageing” might become a relevant mechanism in order reflect demographic change in the region. This would feed back to the public good dilemma because the need for neighbourhood support should in turn increase.
Likewise, the lifestyle distribution in the population changes over time which is not yet represented in the simulations. Finally, empirical founding can be further improved by including results of on‐going surveys of the target area. This yet unexploited methodical link between model and survey data is further discussed in the following section.
In summary, both case‐specific ABSS exercises have strengths and limitations in terms of their empirical soundness. Likewise, direct validation of simulation results is rarely feasible due to the lack of available datasets. To this end, we have to rely on domain experts to confirm that simulation results reasonably reflect the situation characteristics in the target area. All these points have to be transparently documented and clearly stated when simulation results and in particular policy implications are presented.
6.3.3 The potential of the methodical approach of HAPPenInGS
The three agent‐based simulation models presented in this thesis are representatives of a particular paradigm within the multitude of existing approaches in ABSS. The thesis framed this school of modelling under the heading of psychologically sound middle‐range models.
The distinctive property of middle‐range agent‐based models is that they capture the observed individual and structural characteristics of a problem domain such that the model remains applicable to similarly structured cases. Such middle‐range models obtain psychological soundness by grounding the model representation of individual decision‐
making adequately in psychological theory. In this respect, the models reported in this thesis are built around the HAPPenInGS theory. This particular setup has some notable methodical implications.
First of all, the grounding of agent decision‐making in psychological theory ensures that all relevant drivers of behaviour and their interaction are consistently represented and therefore ensure the validity of a model’s structure. Secondly, HAPPenInGS is sufficiently generic to remain transferable to simulation exercises for other empirical contexts involving a public good dilemma.
Finally, the fact that HAPPenInGS is embedded in the TPB and in theory on social value orientations links the class of HAPPenInGS models to a rich toolbox of instruments from empirical psychology: In principle, survey data from questionnaires for the TPB can be used for model initialisation or validation (Schwarz & Ernst, 2009). Likewise, empirical methods for measuring social value orientations can contribute to the empirical grounding of simulation models. In particular the latter implication of the methodical approach was not explored in this thesis and indicates promising follow‐up research.