Policy-Driven Dynamics in Sustainable Recycling: Evolutionary Dynamics on Multiple Networks with Case Insights from China

Abstract

Residents have substantial control over their daily lifestyles, and their behavior change has a considerable potential to reduce emissions. Understanding the adoption of sustainable lifestyles and recycling behaviors and how behavioral policies might shape this decision-making is critical to the transition to sustainable consumption patterns. This paper developed a multi-agent model on a multiplex complex network that integrates evolutionary game theory to simultaneously capture information transmission and behavioral interaction dynamics. In this model, residents determine decision utilities under the influence of internal and external factors and information transmission and then perform social interactions according to an evolutionary dynamics model. A real case of residents’ green express packaging purchase decisions in China was used for parameter initialization. Explorative simulations and scenario analyses were conducted to investigate the adoption patterns of sustainable lifestyles under different policy scenarios. Results indicate that the dynamic evolution of residents’ sustainable lifestyle decisions relies on social interactions and social networks. Government subsidies are effective in fostering sustainable decisions, but this effect is sensitive to the size of complex networks. Information campaigns and government subsidies have a significant marginal contribution to promoting sustainable lifestyles than green labeling schemes. Implementing subsidies and information campaigns as policy mixes can exert complementary effects and improve aggregate outcomes of policy packages.

1. Introduction

Global climate change is one of humanity’s most persistent and complicated problems, with sustainable development being one of the most pressing issues that it brings with it [1]. The increasing dependence of current consumption systems on fossil energy sources puts unprecedented pressure on the sustainable transition and further exacerbates the systemic risks of climate change. The promotion of sustainable social transitions, particularly far-reaching changes in consumption patterns, to tackle these challenges has become a global consensus and priority [2].

The transition to more sustainable consumption relies heavily on residents’ decisions to practice sustainable lifestyles. It is estimated that direct and indirect greenhouse gas emissions generated by residents’ behavior account for more than 70% of total global emissions [3]. This implies that changes in the residents’ lifestyles have a vast potential to reduce emissions through the so-called “behavioral mitigation wedge” [4]. Sustainable recycling behavior is highly relevant for sustainable lifestyles. Although residents are increasingly concerned about the environmental and climate impacts, they tend to practice unsustainable lifestyles and recycle unsustainably. The intention–behavior gap and attitude–behavior gap associated with sustainable recycling behaviors persist [5]. If the behavioral pattern of sustainable consumption is to be understood, it is important to know the evolutionary process of resident decision-making for sustainable recycling behaviors and how public policies might shape this process.

The sustainable lifestyle refers to the ecological concept that individuals develop in harmony with nature, making green consumption and green living conscious behaviors. Scholars generally agree that sustainable lifestyles include practicing sustainable consumption patterns and purchasing sustainable products in areas such as food, clothing, and transportation, and sustainable recycling behaviors [6]. A large body of literature shows great interest in influential factors and decision-making processes underlying sustainable lifestyles or recycling behaviors. Some of the drivers and barriers affecting sustainable lifestyles have been explored by scholars, including economic concern [7], social norms [8], environmental information [9], and government policies [5]. These works provide critical theoretical and empirical foundations for the research.

Decision-making research in sustainable behavior has been grounded in several foundational theoretical frameworks. The Theory of Planned Behavior [10,11] emphasizes the role of attitudes, subjective norms, and perceived behavioral control in shaping behavioral intentions. Social Cognitive Theory [12] highlights the importance of social learning and self-efficacy in behavior adoption. The Value–Belief–Norm model [13] connects personal values to environmental beliefs and moral norms that guide pro-environmental actions. These foundational models have provided important insights into individual-level decision processes.

Building upon these foundational approaches, researchers have increasingly turned to more sophisticated modeling techniques to capture the complexity of sustainable lifestyle decisions. Cheng et al. [14] built a sustainable lifestyle decision model for residents from the perspective of regret and game theory. Through agent-based simulations, they analyzed the effects of preferences, income, and interaction intensity. Yang et al. [15] constructed a network-based decision model for sustainable consumption behavior and simulated the effects of different regulatory policies on sustainable behavior adoption. Cheng et al. [16] focused on the experienced utility of policies that influence sustainable lifestyle decision-making. They developed a model to measure the experienced utility of sustainable lifestyle guiding policies. In addition, several studies have explored the consumption decision-making process for specific sustainable products and projects such as electric vehicle purchases [17], residential photovoltaics adoption [18], and sustainable food shopping [19].

While existing research on sustainable lifestyle and recycling behavior decision-making processes offers valuable insights, most studies focus primarily on information transmission effects within social networks, overlooking the crucial role of behavioral interactions among residents. Behavioral interaction refers to the dynamic process of mutual impact and interdependence of behavioral decisions among social members. In the reality of social networks, residents’ recycling decisions in their daily lives are highly interdependent on the members of their reference group. For example, when a resident practices a sustainable lifestyle and undertakes garbage classification, his neighbors may be influenced by his decisions and then change their behaviors and lifestyles in a social learning manner.

In fact, theoretical and empirical evidence highlight the important role of such behavioral interactions in sustainable lifestyles [4,20]. However, existing decision-making models for sustainable recycling behaviors are limited in their ability to explain how residents’ decisions are influenced by behavioral interactions and how sustainable recycling decisions evolve dynamically within resident groups. This leaves a critical knowledge gap. Furthermore, while some studies have examined single policy interventions, there remains insufficient understanding of how multiple policies interact and complement each other in promoting sustainable lifestyle transitions. The lack of comprehensive frameworks that can simultaneously model both information diffusion and behavioral evolution under various policy scenarios represents a significant limitation in current research.

In this regard, evolutionary game theory offers an effective approach to understanding the behavioral interactions and evolutionary mechanisms involved in sustainable lifestyle decisions. This paper, thus, introduces evolutionary game theory combined with multiplex complex network models into the sustainable lifestyle decision model to analyze the dynamic evolution of sustainable recycling behaviors within residents.

Relying solely on residents’ self-regulated behavioral changes under social influence may not be sufficient to drive transitions in group lifestyle decisions. Public policies, especially behavioral public policies aimed at promoting sustainable lifestyles or sustainable recycling, are also critical and particularly relevant to the focus of this study. A variety of policies are available, such as economic policies (e.g., subsidies incentives), information policies (e.g., information campaigns), and public goods and services. Existing research on sustainable lifestyles has focused on the potential impacts of specific policies. For example, Rajapaksa et al. [21] investigated the effect of monetary and non-monetary incentive policies on residents’ water consumption behavior. Ran et al. [19] studied the mechanisms of information policies on consumers’ sustainable food decision-making. However, for encouraging sustainable lifestyles, discussing only one type of policy may yield oversimplified and fragmented recommendations. Since no single “silver bullet” policy instrument can bring about the behavioral changes needed to address the low-carbon transition [22], it is more realistic and comprehensive to examine multiple policies within the same framework. In addition, exploring the evolution of sustainable lifestyle decisions under different scenarios can also provide insights into the dynamic impact of multiple policies.

This paper aims to improve the understanding of the underlying mechanisms and dynamic evolution of sustainable lifestyle and recycling decisions. Based on this, it is also possible to assess the dynamic effects of different policies on residents’ decision-making. Explicitly, we will address the following research questions: RQ1—What are the evolutionary mechanisms and trends of sustainable lifestyle and recycling decisions among resident groups? RQ2—How do residents’ sustainable lifestyle and recycling decisions evolve dynamically under different policy scenarios?

This paper applies the network-based evolutionary game model with the ABM approach to develop a sustainable recycling decision-making model. In the model, residents are embedded in a multiplex social network, where in one layer information transmission occurs to influence residents’ decision utility while social interaction to determine their behavioral strategies occurs in another layer. Through forward-looking simulations on various policy scenarios, the proposed model can provide dynamic insights into the directional impact of policies targeting on sustainable lifestyles.

Academically, this study contributes to existing knowledge in three ways that directly address the identified research gaps: (1) Unlike existing models that focus primarily on information transmission, a network-based evolutionary game model is introduced into the modeling framework of sustainable lifestyle decision-making to analyze the social interactions among residents. This approach specifically addresses the critical gap in modeling behavioral interactions by simultaneously capturing both information diffusion and behavioral evolution dynamics, thereby overcoming the neglect of resident behavioral interactions in current research and enriching the modeling practice of sustainable lifestyle decisions. (2) While previous studies typically provide static analysis, the proposed model captures the dynamic nature of group behavioral decisions. By analyzing the evolutionary process of group sustainable lifestyle decisions, this study addresses the limitation of temporal understanding in existing research and expands the research on behavior change interventions from a dynamic perspective. (3) In contrast to studies that examine single policy interventions in isolation, through the scenario analysis for multiple policies, this study not only provides new insights into the directional impact of different policies on sustainable lifestyles but also fills the gap in understanding policy interactions and complementary effects, while providing theoretical tools for governments to formulate comprehensive behavioral policies.

The structure of this paper follows accordingly. Section 2 provides a literature review related to modeling and policies of sustainable lifestyles. Section 3 constructs a decision-making model for sustainable recycling and elaborates the details of the model. Section 4 reports the simulation results of various scenarios. Section 5 discusses results and analyzes policy implications. Section 6 concludes this study and suggests limitations and future research directions.

2. Literature Review

The literature related to this study can be divided into two branches: decision-making modeling for residents and policies for promoting sustainable lifestyles.

2.1. Sustainable Lifestyle Decision-Making Modeling for Residents

Sustainable lifestyle decision-making among residents has received considerable scholarly attention. Researchers have increasingly recognized such decisions as long-term dynamic social processes and have investigated their underlying mechanisms. Li et al. [23] constructed a network-based evolutionary game model for individual sustainable behavior and analyzed the key factors influencing the interaction of sustainable behavior through simulations. Gao and Tian [24] applied threshold model and contagion model to study the diffusion of green behaviors in multiplex complex networks. Li et al. [25] proposed a green behavior diffusion model to study the evolution of group green behavior in the presence of both information and behavior diffusion.

There is also literature that focuses on modeling decisions for specific scenarios and green products, such as electric vehicles [17], recycling behavior [26], and energy-saving behavior [27]. These decision-making studies have used agent-based modeling (ABM) as the main methodology. The ABM approach, also referred to as multi-agent simulation, is a bottom-up modeling approach to studying complex systems [18]. This approach allows exploring the macro-level emergent patterns of complex systems by encapsulating attributes and behavior rules for micro-level agents.

However, existing ABM studies on sustainable lifestyle decisions have focused on the effect of information transmission in social networks on decision utility and group decision-making, such as adopters spreading information about positive attitudes and, thus, influencing the decision utility of non-adopters, but more complex behavioral interaction processes such as social games and evolutionary learning have not received sufficient attention. Complex behavioral interactions among residents greatly influence individual behavioral changes and further shape overall behavioral pattern transitions. Current ABM models related to sustainable lifestyles have limited ability to capture this dynamic process. In contrast, the network-based evolutionary game approach provides an effective means of portraying the dynamics of group behavioral interactions.

Evolutionary game theory, derived from biological evolutionary theory, holds that boundedly rational individuals evolve to optimal strategies by imitating others and constantly adjusting their behavior [28,29]. With the flourishing of network science, researchers began to emphasize the vital influence of the heterogeneity of social agents and the topology of social systems on the evolution of the system. The network-based evolutionary game approach, which combines evolutionary game theory with complex network models, has emerged and shows promise for solving such heterogeneity problems [30].

More importantly, the network-based evolutionary game approach defines the interaction rules of heterogeneous agents, which is needed for the ABM approach to model complex systems. Thus, we argue that their combination is appropriate for modeling sustainable lifestyle decision-making. We introduce the network-based evolutionary game approach to sustainable lifestyle decision-making models to narrow the knowledge gap on social interactions and to contribute to modeling practices of residents’ decisions.

2.2. Policies for Promoting Sustainable Lifestyles

The motivation of scholars and practitioners to study intervention policies encouraging behavior change is the idea that most of the behaviors that people practice in daily lives lead to waste and that changing these behaviors is beneficial to sustainability [31]. The policies aimed at changing resident behavior are known as behavioral public policies or, in a broader sense, as regulatory policies in some sustainable behavior studies [32]. Many barriers to sustainable lifestyles stem from a lack of public policy, so developing effective policies is a prerequisite for changing residents toward sustainability [33].

Many policies implemented in different policy areas may influence individuals’ lifestyles and researchers have studied these policies extensively. Jin and Zhao [34] used an evolutionary game model to study the effect of China’s green labeling scheme on residents’ product purchase choices. Cheng et al. [16] constructed a policy utility dislocation model for sustainable lifestyle choices, and numerical simulations based on the model revealed significant effects of government subsidy policies. Wang et al. [35] simulated the effect of policies on resident adoption of photovoltaic products. Their results suggested that information campaign policies are effective and necessary for photovoltaic adoption.

In addition to studies in the context of China, a large number of studies have examined the effects of sustainable recycling policies from different national contexts around the globe. Mielinger et al. [36] analyzed the causes of misclassification behaviors of packaging waste in the context of Germany, and showed that more accurate information and financial incentives best motivate consumers to correctly segregate waste. Asare et al. [37] investigated incentive programs that could promote the recovery of municipal solid waste resources using Tamale, Ghana as an example, showing that incentives have a significant impact on the amount of waste, and that the use of prizes as an incentive program has a high propensity to recycle waste. Northen et al. [38] conducted a survey in the United Kingdom aimed at determining the trends in the disposal of single-use plastic products, as well as the individual recycling of plastic motivations and barriers. Saulītis et al. [39] conducted a field experiment in more than 10,000 latvian households to explore the effects of a push in the recycling system, and the results emphasized that pushes are essential to highlight systemic deficiencies and point to more effective engagement strategies. Hole et al. [40] reviewed the EU’s policies and incentives for recycling of textiles related to textile recycling, and argued that by implementing recycling labeling policies, consumer awareness can be increased. Lodhia et al. [41] analyze regulatory diversification in the Australian context, arguing that growth in waste volumes requires additional behavior change incentives.

Following the existing research on behavioral public policies, this study examines the dynamic impact of government subsidies, green labeling schemes, information campaigns, and policy mixes on the evolution of sustainable lifestyle decisions. The main reason for choosing these policies is that they are widely implemented and investigated, which means that they are more relevant and influential to the practical situation. Moreover, examining multiple policy scenarios, especially policy mixes, within the same framework can also facilitate a more nuanced insight into behavioral policy interventions, thus contributing to this literature stream.

3. Methods

3.1. Model Framework

The main stakeholders involved in the evolution of sustainable recycling decisions include residents and governments. Resident agents are embedded in a multiplex social network. The nodes in the network represent resident agents. The connected edges representing behavioral interactions between nodes indicate connections between residents who know each other in reality or have social connections. The connected edges that indicate information transmission exist only between adopters and rejectors of sustainable lifestyles and recycling. The intuition of this network setting is that sustainable lifestyle rejectors in social networks do not share information about sustainable lifestyles and products to their neighbors, while adopters do.

Drawing on previous decision-making research [42], the most natural and popular way to describe resident agents is to define a dichotomous variable indicating adoption or rejection status, which in this paper is adoption or rejection of sustainable lifestyles. In addition to this basic state, resident agents are self-adaptive and self-organizing and are assigned heterogeneous attributes and behavior parameters. The utility of residents’ decisions is influenced by willingness to pay, additional benefit, and total cost. Residents in social networks are also shaped by social interactions, which arise from information transmission on the one hand and evolutionary games of group behavior on the other.

The government can indirectly influence resident decision-making through different policies affecting sustainable products. Different policy scenarios, including government subsidy, green labeling scheme, information campaign, and policy mix, are introduced in the model, and simulations are performed to reveal the evolution of sustainable recycling decisions. The scenario analysis of different policies conducted can provide deeper insights into the interplay of factors and the impact of various policies on the system, thus facilitating the selection and design of behavioral policies. Figure 1 illustrates the framework of the proposed model.

3.2. Basic Assumptions for Modeling

Based on the above problem description and the requirements of evolutionary game theory, the following basic assumptions are made to serve as the basis for modeling sustainable lifestyle decision-making.

(1)

All resident agents have bounded rationality and incomplete information. This assumption is grounded in behavioral economics literature, which demonstrates that individuals have cognitive limitations and cannot process all available information [43,44]. In sustainable behavior contexts, residents face complex trade-offs between environmental benefits and personal costs under uncertainty, making bounded rationality a realistic assumption. They are in a complex external environment and need to determine the better strategy through repeated games.

(2)

To quantify the utility of decisions, this paper assumes that residents adopting sustainable lifestyles and recycling behaviors always pay a certain cost for sustainable recycling while those rejecting sustainable lifestyles and recycling behaviors always do not pay this cost.

(3)

The resident population of size $N$ is embedded in a complex network $G$. There are two pure strategies for resident agents: adoption and rejection of sustainable lifestyle. The initial proportion of agents adopting sustainable lifestyle is $E_0$.

(4)

Information about residents’ strategies and payoffs is shared only among connected agents in the complex social network. All resident agents perform social interactions with their connected neighboring nodes and then update their behavior strategies synchronously.

(5)

The government influences the adoption of sustainable lifestyle through four policies: government subsidies, green labeling schemes, information campaigns, and policy mixes. The intensity of the various policies is kept constant during the discussion period.

3.3. The Network-Based Evolutionary Game Model

3.3.1. Complex Network Model

Resident decision-making in the real world does not occur in a social vacuum but in relation to other heterogeneous residents. The relationships between residents structure the topology of social networks and form groups of internally interrelated residents. From the perspective of complex networks, social connections among residents are essential because they shape the channels through which information and behavior diffuse. In such a resident network, the strategies and behaviors of each node may influence other nodes in the network, especially the connected nodes.

The choice of scale-free networks for modeling social connections is theoretically and empirically justified. Scale-free networks exhibit power–law degree distributions where a few nodes have many connections while most nodes have few connections, reflecting the reality of social networks where some individuals are highly connected while others have limited social ties [45]. Empirical studies have demonstrated that real-world social networks [46]. In addition, previous studies on the evolution of sustainable behaviors have argued that scale-free networks are suitable for describing social ties between residents [47,48]. Therefore, this paper establishes scale-free networks as the carrier network model for resident agents. This network is denoted as $G=\left(V,E\right)$, where $V$ refers to the set of resident nodes and $E$ refers to the set of undirected edges in the network. Behavioral interactions occur between nodes in the network with connected edges. Then, the scale-free network model is established with the following algorithm [45]:

(1)

Growth: starting with a small network, a new node is introduced each round and connected to the existing $m$ nodes.

(2)

Preferential attachment: denoting the degree of node $i$ by $d_i$, the probability that a newly introduced node is connected to node $i$ is calculated by Equation (1).

$$p\left(i\right)={\displaystyle \frac{d_i}{{\sum }_{n=1}^N{d}_n}}$$

(1)

3.3.2. Evolutionary Game Model

Since this paper assumes that sustainable lifestyle adopters always recycle sustainably and pay the costs, sustainable recycling decisions can be analyzed in two aspects: on the one hand, the decision utility analysis according to the willingness to pay and benefits and costs shaping residents’ pay decisions, and on the other hand, the interaction of resident behavior in the social game process. Thus, the total payoff to residents for adoption or rejection of sustainable recycling is determined by the utility of the adoption behavior itself and the payoff of the social game. This paper first analyzes and models the utility of residents’ adoption decisions and then presents the game payoffs of social interaction by residents with different strategies.

(1)

Utility of adoption of sustainable lifestyles

The adoption of sustainable lifestyle and recycling behavior among residents is a transition from unsustainable to sustainable lifestyle. In other words, the adoption of sustainable lifestyle can be seen as a change in the status quo. Therefore, the decision utility of adopting sustainable lifestyle is the focus of modeling. Referring to the utility functions constructed by [15] for consumption decision-making, the utility of resident $i$ adopting sustainable lifestyle and purchasing sustainable products at time $t$ can be expressed as Equation (2):

$$U_{i,t}={WTP}_{i,t}+{BEN}_{i,t}-{TCO}_{i,t}$$

(2)

where ${WTP}_{i,t}$ is the resident’s willingness to pay, ${BEN}_{i,t}$ is the additional benefit of adopting sustainable consumption behavior, and ${TCO}_{i,t}$ is the cost paid for sustainable lifestyle. Here, the first-term willingness to pay refers to the highest value residents are willing to pay for the adoption of sustainable lifestyles and recycling, including the basic fee and the additional fee determined by the actual perceived value. The second-term additional benefits include the environmental benefits and government subsidies gained by purchasing sustainable products. The last term refers to the total cost to residents for changing the status quo to adopt sustainable lifestyles, including fixed cost, green cost, and information search cost. Furthermore, the three components of the utility function of resident adoption behavior, i.e., willingness to pay, additional benefit, and total cost, can be calculated by Equations (3)–(5).

$${WTP}_{i,t}=f_b+\left(V_a-V_b\right)\times f_a$$

(3)

$${BEN}_{i,t}=k_e\times g\times a+S\times g$$

(4)

$${TCO}_{i,t}=C+k_c\times g+{\displaystyle \frac{\eta \times g^2}{2}}$$

(5)

Here $f_b$ and $f_a$ represent the basic fee and the additional fee influenced by perceived value that residents are willing to pay for the product. $V_a$ and $V_b$ are the actual perceived value of the sustainable product and the basic perceived value of all products, respectively. $g$ refers to the green degree of the product, $k_e$ is an environmental coefficient, $a$ is the environmental awareness of residents, and $S$ is the subsidy rate when the government implements the subsidy policy. In the total cost ${TCO}_i$, $C$ is the fixed cost, $k_c$ is the unit cost for green degree, and $\eta$ is the information search cost for the resident.

The linear additive structure of our utility function (Equation (2)) is well-established in green behavior decision-making research [48,49,50]. This specification assumes that different utility components contribute independently to overall decision utility, which is theoretically justified when the components represent distinct aspects of the decision problem: economic willingness to pay, environmental benefits, and adoption costs. Linear additive utility functions have been widely applied in studies of sustainable consumption behavior, environmental product choice, and green lifestyle adoption, where researchers typically decompose decision utility into separable economic, environmental, and social components.

Regarding the quadratic specification of information search cost in Equation (5), this functional form reflects the theoretical concept of increasing marginal search costs that is particularly relevant in green behavior contexts. The quadratic cost structure captures the reality that initial environmental information is relatively accessible through general media and basic product labeling, but detailed sustainability information requires progressively more effort to obtain from specialized sources, technical reports, or certification databases. This convex cost pattern is commonly observed in green consumer behavior research [14,15,49,51], where consumers face escalating cognitive costs when seeking comprehensive environmental impact information. The utility function formulation is consistent with established frameworks in green behavior decision-making literature, where researchers typically model sustainable consumption decisions as trade-offs between immediate economic costs, perceived environmental benefits, and information acquisition efforts. This approach has been successfully applied in studies of green product adoption, sustainable lifestyle transitions, and environmental behavior change, supporting the theoretical foundation of our utility specification.

Regarding the actual perceived utility in Equation (3), the actual perceived utility of residents practicing sustainable lifestyles is usually higher than purchasing general products. Residents may perceive higher environmental value in sustainable products and share this value in social networks in the form of information transmissions, which leads to higher actual perceived utility. For example, residents may increase the actual perceived utility of photovoltaics due to the adoption of residential photovoltaics by their neighbors. According to [17,35], the actual perceived utility $V_{a,i}$ of resident $i$ can be expressed as

$$V_{a,i}=\left\{\begin{array}{c}{\displaystyle \frac{{\sum }_{j}c\left(i,j\right)\times V_{a,j}\times s_j}{{\sum }_{j}c\left(i,j\right)\times s_j}}, if {\sum }_{j}c\left(i,j\right)\times s_j>0 \\ V_b, if {\sum }_{j}c\left(i,j\right)\times s_j=0\end{array}\right.$$

(6)

where $c\left(i,j\right)=0 \mathrm{o}\mathrm{r} 1$ denotes the relationship between resident $i$ and resident $j$, and $c\left(i,j\right)=1$ indicates the existence of connected edges between them. Another binary variable $s_j=1 \mathrm{o}\mathrm{r} 0$ denotes the adoption or rejection of sustainable behavior by resident $j$. This equation captures the impact of information transmission in social networks on the actual perceived utility of potential adopters. Thus, the utility of resident $i$ adopting sustainable lifestyle can be alternatively given as Equation (7).

$$U_{i,t}=f_b+\left(V_a-V_b\right)\times f_a+k_{e}ga+Sg-\left(C+k_{c}g+{\displaystyle \frac{\eta g^2}{2}}\right)$$

(7)

The parameter choices in our utility function formulation are grounded in established theoretical frameworks. The cost–benefit structure reflects the fundamental trade-off identified in environmental economics literature, where sustainable behaviors involve immediate costs for long-term environmental benefits [52]. The social interaction component builds upon social influence theory, which demonstrates that individual decisions are significantly affected by peer behavior and social norms [53].

(2)

Game payoffs and game matrix

The adoption or rejection of sustainable behavior is a daily environmental decision dilemma and group game scenario [23,24]. While adopting sustainable behaviors can maximize the benefits to society as a whole, the individual payoff structure creates a tension between private and social interests. In our model context, rejecting sustainable behaviors may result in higher short-term individual benefits under specific conditions: (1) when the immediate costs of sustainable behavior adoption (such as higher prices for green products, time investment in environmental practices, or inconvenience) exceed the immediate private benefits for individual actors; (2) when positive environmental externalities are not internalized through market mechanisms or policy interventions; and (3) when there is insufficient social pressure or reputational consequences to offset the private cost advantages of non-sustainable choices.

However, this individual advantage of rejection is conditional and context-dependent. The relative payoff advantage of rejecting sustainable behavior diminishes when policy interventions alter the cost structure, when social norms create reputational costs for non-sustainable behavior, or when environmental degradation reaches levels that directly impact individual welfare. Moreover, the collective outcome becomes increasingly problematic as more residents reject sustainable behaviors, creating negative externalities that ultimately reduce societal well-being and may even harm the long-term interests of the rejectors themselves, thus creating a green game dilemma [54].

Our model captures this tension by incorporating both the immediate individual cost–benefit calculations and the social interaction effects that can modify these payoffs through reputation factors, social norms, and policy interventions, thus providing a more nuanced representation of the sustainable behavior adoption dilemma.

In addition, there are complex interactions between residents’ behaviors. Residents may change their behaviors driven by benefit and conformity or motivated by compliance with social norms under the influence of their neighbors’ behavior. The game theory is an effective way to analyze such behavioral decisions and interactions among multiple agents. Therefore, the adoption of sustainable lifestyle by residents can be seen as a long-term and repeated game process, which is suitable to be modeled by the evolutionary game framework. We introduce the evolutionary game framework to capture the impact of behavioral interactions between residents on decision total payoffs.

In this paper, we assume that social interactions require costs. An interaction cost is reduced by the positive externality of sustainable behavior when both a resident and another neighbor resident in the network adopt the sustainable behavior. In contrast, a game between adopters and rejectors leads to higher interaction costs for adopters and additional reputation losses for rejectors [55].

If both players in the game adopt sustainable recycling behavior, their game payoff is denoted as $W^r$. If one player chooses the adoption strategy and the opponent chooses the rejection strategy, the game payoff of the adopter is denoted as $W^s$, and the payoff of the rejector is denoted as $W^t$. The payoff of both players’ rejection of sustainable recycling behavior is denoted as $W^p$. Then, the total payoff of the resident decision-making and the social interaction is denoted as follows:

$$W_{i,t}^r=U_{i,t}-bI$$

(8)

$$W_{i,t}^s=U_{i,t}+p-I$$

(9)

$$W_{i,t}^t=U^{\prime }-p$$

(10)

$$W_{i,t}^p=0$$

(11)

where $I$ is the social interaction cost, $b$ is the cost-sharing factor when both players to the game are adopters, $p$ is the reputation factor, and $U^{\prime }$ is the basic utility of rejecting sustainable lifestyle in the game. The game matrix of social interactions between residents is denoted as follows:

$$M_e=\left(\begin{array}{cc}{(W}_{i,t}^r,W_{j,t}^r)& {(W}_{i,t}^s,W_{j,t}^t)\\ {(W}_{i,t}^t,W_{j,t}^s)& {(W}_{i,t}^p,W_{j,t}^p)\end{array}\right)$$

(12)

3.3.3. Evolutionary Mechanism

In each round of the game, each resident agent plays with all neighbor residents in the network according to the above game matrix and then accumulates the payoffs. After the total payoffs are obtained, each agent randomly selects a neighbor agent as a candidate for social learning to update the strategy. This process allows agents with bounded rationality to learn from other agents in social interactions. Without loss of generality, the Fermi function is used to calculate the probability of agent $i$ imitating the strategy of agent $j$ as follows [47,56]:

$$P\left(s_i\leftarrow s_j\right)={\displaystyle \frac{1}{1+\exp \left[-\frac{W_{j,t}-W_{i,t}}{\tau }\right]}}$$

(13)

where $W_{i,t}$ and $W_{j,t}$ are the total payoffs of resident $i$ and resident $j$ at time $t$. $\tau$ is the noise factor during evolutionary process, associated with irrational factors in social learning. This evolutionary rule implies that resident $i$ will readily imitate the strategy of $j$ if its total payoff is lower than that of resident $j$. Conversely, even if the total payoff of resident $i$ is higher than that of $j$, it will still imitate $j$’s strategy with a smaller probability, because the social learning is not completely rational.

The adoption of the Fermi function for modeling human imitation behavior is grounded in evolutionary game theory literature, where it has been widely used to model probabilistic strategy updating in social learning contexts [48,57,58]. The Fermi function acknowledges that individuals do not always choose the seemingly optimal strategy, which aligns with behavioral observations of bounded rationality in decision-making. The noise parameter, $\tau$, serves as a mechanism to incorporate various factors that influence imitation behavior beyond pure payoff considerations. This parameter can reflect the influence of social preferences, cultural norms, and other contextual factors that affect strategy adoption decisions. While more complex hybrid learning rules could potentially capture additional behavioral nuances, the Fermi function provides a well-established and tractable approach for modeling social learning dynamics in evolutionary game contexts.

Since this paper focuses on the adoption of sustainable lifestyle at the group level, the adoption rate, $E_t$, is used to measure the evolution of the adoption strategy at time, $t$. This rate is calculated by the proportion of agents adopting sustainable lifestyle ($N_c$) to all agents in the network, as shown in Equation (14).

$$E_t={\displaystyle \frac{N_c}{N}}$$

(14)

In the subsequent simulations we are concerned with the variation in the adoption rate, $E_t$, over time, $t$, which can reflect the evolution of sustainable lifestyle adoption in the social network.

4. Results

4.1. Parameter Initialization

A case study of green express packaging purchase decisions by residents of Wuhan, China, was used for parameter initialization and simulation experiment of the proposed model. Since the Wuhan Municipal Government issued the “Waste-free City Construction Implementation Plan”, Wuhan’s express industry is carrying out the construction of “green express networks” and is replacing non-green express packaging with green express packaging. At express sites, residents can choose recyclable green express packaging boxes or bags (such as plastic bags and disposable woven bags), or non-recyclable non-green packaging boxes or bags (such as biodegradable bags and tape-free cartons).

The decision-making process of choosing these two types of express packaging with well represent the sustainable lifestyle choices of residents. Therefore, we used field surveys to collect data related to residents’ green express packaging decisions for parameter initialization. The main data were obtained from questionnaire surveys of residents and interviews with express site staff, and other data were obtained from government policy documents and existing decision-related literature.

A total of 330 questionnaires were obtained, of which 247 were valid (an effective questionnaire rate of 74.8%). Invalid questionnaires including incomplete responses and contradictory answers were excluded, meeting the previous standards for questionnaire distribution by Yang et al. [59]. The measurement items and samples in the case and the demographic characteristics are given in the Appendix A. Our parameter settings focus on the measurement of attributes related to the sustainable lifestyle of the residents, so the average value of the measured parameters is used as the initial value for the simulation. The parameters, their implications in the case, and the data sources are shown in

Table 1. Parameter, implication in the case, initial value, and data source.

Parameter	Implication in the Case	Initial Value	Data Source
$f_b$	Basic fee for green express packaging	6	Interviews
$f_a$	Additional fee influenced by perceived value	3.2	Questionnaire
$V_b$	Basic perceived value of green and non-green express packaging	3	Interviews
$C$	Fixed cost of choosing green express packaging	3	Interviews
$k_c$	Unit cost for green degree of green express packaging products	3.2	Questionnaire
$k_e$	Environmental coefficient	0.8	Questionnaire
$a$	Environmental awareness of residents	0.5	Questionnaire
$g$	Green degree of products	0.9	Questionnaire
$\eta$	Information search cost of residents	0.9	Questionnaire
$b$	Cost-sharing factor of residents in social interactions	0.7	[59]
$I$	Social interaction cost of residents	3.9	Questionnaire
$p$	Reputation factor	3.4	Questionnaire
$U^{\prime }$	Basic utility of rejecting purchasing non-green express packaging	2.8	Questionnaire
$\tau$	Noise factor during evolutionary process	0.5	[60]
$S$	Government subsidy rate	0	[61]
$E_0$	Proportion of residents adopting sustainable lifestyle at the beginning of the evolution	0.1	Assumed
$N$	Size of the resident group	200, 500, 1000, 1500	Assumed

Note: Questionnaire-derived parameters represent rounded mean values. Detailed descriptive statistics are provided in the text.

.

For the other parameters, interviews with express site staff revealed that the price of a recyclable box with dimensions of 40 cm × 30 cm × 20 cm is 6 CNY, while the price of a disposable box of the same size is 3 CNY. Therefore, the basic fee for purchasing green express packaging $f_b=6$ and the basic perceived value of express packaging $V_b$ and the fixed cost $C$ were set to 3. Referring to the study by [15] on evolutionary game in express packaging recycling, they estimated the cost-sharing factor of individuals in the social game to be 0.7, thus $b=0.7$. The noise factor during the evolutionary process is set to $\tau =0.5$ [60]. Finally, the size of the resident group $N$ and the proportion of residents initially purchasing green express packaging are assumed to be 500 and 0.1, respectively.

Regarding data processing and parameter derivation, we provide the following clarification of our methodology. The parameter values shown in Table 1 were obtained from multiple sources including questionnaire surveys, interviews with express site staff, literature references, and reasonable assumptions based on the research context.

For parameters derived from our questionnaire data (N = 247 valid responses), the measurement and processing approach varied by parameter design. Some variables (environmental coefficient, environmental awareness, green degree of products, information search cost) were designed as continuous variables with values between 0 and 1, where higher values represent higher degrees of agreement. Other variables (additional fee influenced by perceived value, social interaction cost, reputation factor, basic utility measures, unit cost) were measured using 5-point categorical scales where 1 indicates complete inconsistency and 5 indicates complete agreement, as detailed in

Table A1. Variable measurement items.

Variables	Measurement Items	Assignment Criteria
Environmental coefficient	I get a good sense of pleasure from purchasing green product	Continuous variables with values between 0 and 1. Higher values represent higher degree of agreement.
Environmental awareness of residents	Green labels on express packaging can reverse my idea of not choosing green express packaging
Green degree of products	Green express packaging can improve environmental pollution and alleviate resource shortages.
Information search cost coefficient	It takes time and effort to understand how green express packaging can improve environmental pollution and alleviate resource shortages.
Additional fee influenced by perceived value	I care about the perceived value generated by purchasing green express packaging.	Categorical variables, where 1 indicates complete inconsistency, 2 indicates relative inconsistency, 3 indicates basic agreement, 4 indicates relative agreement, and 5 indicates complete agreement.
Social interaction cost	It takes time to discuss sustainable lifestyles with neighbors.
	It takes effort to discuss sustainable lifestyles with neighbors.
Reputation factor	I care about the impact of non-green behavior on my reputations in social situations.
Basic utility of rejecting purchasing non-green express packaging	Rejecting the purchase of green express packaging can save time.
Basic utility of rejecting purchasing non-green express packaging Unit cost for green degree of green express packaging products	Rejecting the purchase of green express packaging can save effort.
	Rejecting the purchase of green express packaging can make me feel relieved from a burden.
	I am willing to pay more for a greener product.

. Mean values were calculated for each parameter type across all valid responses and then rounded to one decimal place for model implementation consistency.

To ensure statistical transparency, we provide descriptive statistics for the questionnaire-derived parameters before rounding, which are displayed in the Appendix A. These statistics demonstrate reasonable variance in responses and confirm that our parameter initialization reflects genuine heterogeneity in the survey data.

Parameters obtained through interviews with express site staff (such as basic fees for green express packaging, basic perceived values, and fixed costs) were used as reported without additional processing. Other parameters were obtained from literature sources or set as reasonable assumptions for simulation purposes, as indicated in the data source column of Table 1. This multi-source approach ensures that our parameter initialization reflects both empirical data and established research findings while maintaining model tractability.

For policy-related parameters (government subsidies, green labeling intensity, and information campaign intensity), we employ a systematic approach to explore their effects across different intensity levels. Rather than setting fixed parameter values, these policy parameters are treated as relative intensity. In our simulations, we examine several discrete intensity levels for each policy instrument to capture the full spectrum of policy implementation scenarios. This approach allows us to investigate how varying policy intensities affect sustainable lifestyle adoption and to identify intensity combinations for policy mixes.

The case study is simulated in the modeling environment of Python 3.7. After parameter initialization is completed, the model can generate the initial scale-free network and set the initial strategy for the agent. And for each scenario, 250 independent simulations were conducted to prevent additional disturbances from the initial strategy distribution and finally to obtain a stable average adoption pattern. The proposed model, thus, serves as a laboratory for simulating the effects of changes in exogenous policy shocks on the evolution of sustainable lifestyle decisions in the system. With this model, we are able to address “What-if” questions, such as “what happens if the intensity of a given policy changes?”.

4.2. Baseline Scenario of the Evolution of Sustainable Lifestyle Decisions

Scenario analysis is a subsequent step in ABM, designed to provide insights into more specific research questions using the proposed model. We first set the initial parameters according to Table 1 and executed simulations in the network with $N$ of 200, 500, and 1000 to obtain the average pattern of adopter and rejecter behavioral evolution. This scenario was served as the baseline scenario. The simulation results of the baseline scenario, including the time evolution and spatial evolution of the adoption of sustainable recycling behavior, are shown in Figure 2.

The evolutionary trajectory of the adopter proportion shows that the adoption rate of sustainable lifestyles in different sizes of networks shows a similar evolutionary trend, that is, it increases rapidly until reaching evolutionary equilibrium. The influence of different network sizes on the evolutionary trend is shown by the smoother adoption rate curve with larger network sizes. The spatial evolution diagram in Figure 2 can be used to explain the microscopic mechanisms of sustainable lifestyle decisions evolution. It can be observed that about 10% of the residents in the network adopt sustainable recycling behavior when $t=0$. As the game repeats, the rejectors are constantly influenced by the neighboring adopters in the network and imitate their strategies, which leads to 50% adopters at $t=50$. At this point, the number of blue nodes in the network is about the same as the number of red nodes. After 100 iterations ($t=100$), 60% of the residents in the network have adopted sustainable lifestyle.

From the perspective of a network-based evolutionary game, the impact of exogenous policy shocks on the evolution of resident behavior is realized through complex social interactions among residents. On the one hand, the topology of complex networks constitutes a channel for behavioral adoption, and sustainable lifestyle strategy diffuses only among connected residents in the network. On the other hand, residents do not make decisions directly based on the strength of behavioral policies but rather through comparisons with their neighbors’ payoffs and strategies and under the influence of social interactions. Thus, the microscopic mechanism by which policies influence the evolution of residents’ sustainable lifestyle decisions can be thought of as providing payoffs to adopters, thus motivating rejecters to change their behavior under social interaction.

Based on the baseline scenario, we further simulate the impact of single policies, including government subsidies, green labeling schemes, and information campaigns on the evolution of sustainable lifestyle adoption. In the three subsequent scenarios, only one of the parameters was varied in each simulation, so the variation in the adoption rate curve can be considered to be caused by the change in a specific policy-related parameter.

4.3. Scenario 1: Government Subsidy Scenario

Government subsidies are a common public policy to promote the purchase of sustainable recycling-related products, such as government subsidies for environmentally friendly boxes. Subsidies can relatively increase the economic benefits to residents of practicing sustainable lifestyles, thus enabling residents to make more sustainable choices among alternatives. In Scenario 1, the value of government subsidy rate $S$ was adjusted to 0.3, 0.5, 0.7, 0.9 to obtain the evolutionary trajectory of sustainable lifestyle adoption rate under different $S$ values. The results are shown in Figure 3.

It can be seen that the evolution of the adoption rate under the different intensity of government subsidy policies shows a similar trend. In the early stage of the evolution (before about $t=10$), the adoption rate increases rapidly. As the iteration time increases, the increase in adoption rate slows down until it approaches the evolutionary equilibrium. In addition, government subsidies steadily promote the adoption of sustainable consumption behavior in terms of both evolutionary trends and evolutionary outcomes. For the former, the adoption rate rises faster with higher government subsidies. For the latter, the final adoption rate reaches approximately 0.33, 0.50, 0.63, and 0.82 when $S$ is 0.3, 0.5, 0.7, and 0.9.

The above results suggest that the implementation of government subsidy policies can indeed promote sustainable lifestyle adoption. Unlike information policies, the implementation of government subsidy policies in different regions may imply noticeable differences in effectiveness and cost. In other words, the effectiveness of implementing a subsidy policy of equal intensity should vary across social networks of different sizes. Therefore, it is natural to explore the sensitivity of network size to the effectiveness of government subsidies. We set the network size, i.e., the size of the resident population, to 200, 500, 1000, and 1500 for the simulation. Figure 4 plots the final adoption rate $E_{100}$ as a function of the subsidy rate $S$ under different network sizes $N$.

As can be observed in Figure 4, the effect of government subsidies on promoting sustainable behavior is suppressed as the network size increases. For example, the final adoption rate in a 1500-node size network is almost always 10% lower than that in a 200-node size network at all intensities of government subsidies. Our results here suggest that the effect of subsidy policy is sensitive to the size of complex social networks and that the effectiveness of subsidy policy decreases as the size of the social network increases.

4.4. Scenario 2: Green Labeling Scheme Scenario

The green labeling scheme, also known as the sustainability labeling scheme, is a government-implemented certification method for the environmental performance of products and is an important policy to promote sustainable behaviors [34]. For example, Type $\mathsf{{\rm I}}$ labels (ISO 14024) mandate that products meet environmental standards and are resident friendly throughout the entire process from production to distribution [62]. Other examples include China’s Environmental Labeling Scheme, which imposes more stringent environmental requirements for product design, production, packaging, and transportation, and the EU’s Fairtrade label [63]. Such policies facilitate manufacturers’ green messaging to residents, which helps residents understand the environmental efforts of sustainable products and attracts them to practice sustainable behaviors [64].

The green labeling scheme policy influences resident decisions by reducing the information search cost of practicing sustainable behaviors. In Scenario 2, this paper used different information search cost coefficient $\eta$ to represent different policy intensities of green labeling schemes and simulated the evolutionary trend of sustainable lifestyle adoption for $\eta =0.3, 0.5, 0.7,$ and $0.9$, with the results illustrated in Figure 5.

The implementation of green labeling schemes to reduce the information search cost positively promotes the adoption of sustainable behaviors, as reflected by the evolutionary outcomes of adoption rates for $\eta$ of 0.9, 0.7, 0.5, and 0.3 reaching 0.38, 0.43, 0.51, and 0.59, respectively. However, a comparison of this scenario with Scenario 1 reveals that the green label is less effective in promoting sustainable lifestyle decisions than the subsidy. This implies that although the presence of more eye-catching green labels on express packaging or other green products is beneficial to residents’ choice of them, this may be insufficient in terms of policy effectiveness for policymakers.

4.5. Scenario 3: Information Campaign Scenario

Information campaign policy refers to a type of information policy that communicates information related to sustainable product to residents through channels such as social media, posters and magazines, and the Internet. Governments can publicize the additional value and environmental benefits of sustainable products to encourage residents to make the “right choices” and practice more sustainable behaviors. For example, residents may be influenced to consume sustainable food by the news and magazines they read [19].

Scenario 3 was developed to inspect the impact of information campaign policies on the adoption of sustainable lifestyle. Information campaigns influence behavioral strategies by raising residents’ environmental awareness, so changing the value of residents’ environmental awareness, $a$, can represent the impact of different intensity of information campaign policies on residents’ decisions. Figure 6 shows the evolution of the adoption rate, $E_t$, when resident environmental awareness, $a$, is 0.3, 0.5, 0.7, and 0.9, respectively.

The curves of the adoption rates in Figure 6 indicate that the information campaign policy noticeably fosters the adoption of sustainable behaviors in both evolutionary trends and outcomes. Similarly to the effectiveness of government subsidies, an increase in the intensity of information campaigns not only promotes rapid adoption of sustainable lifestyle during the early evolutionary stages but also leads to a marked increase in the final adoption rate. The reason for this stable promotion effect is that information policies raise residents’ environmental awareness and thus increase the additional benefits of practicing sustainable recycling behaviors. This allows rejecters to change their strategies and become adopters in dynamic decision processes of “change-adjustment-convergence” in repeated games.

4.6. Scenario 4: Policy Mix Scenario

The transition to sustainable consumption patterns involves complex social, economic, and behavioral changes. It is often believed that no single “silver bullet” policy can cope with this complexity. The role of policy mixes, i.e., complex arrangements of multiple policies targeting the same sustainability issue, in sustainable transitions has received increasing attention. The core idea of policy mix is that interaction occurs when different policies are linked to the same policy objective. For example, the simultaneous implementation of economic incentives and regulatory instruments may improve the policy effectiveness of regulatory instruments [22]. Moreover, in real economic systems, behavioral public policies are also often carried out in the form of policy mixes.

To gain a more comprehensive understanding of the above policy scenarios, in Scenario 4, we conducted simulation experiments for each two-policy mix of the three policies. The subplots in Figure 7 visualize the impact of policy mixes of (a) government subsidy and green labeling scheme, (b) government subsidy and information campaign, and (c) green labeling scheme and information campaign on the final adoption rate $E_{100}$, respectively.

From the general trend, policy mix scenarios that combine these policies appear to perform better than isolated policy scenarios. As can be seen from the location of the blue areas in each of the subplots in Figure 7, all three policy mixes demonstrate more pronounced effectiveness compared to the single-policy scenarios when both policy intensities are high in the mix. This implies that when both policies in the policy mix have a certain intensity, these policy levers work in concert and together increase the final adoption rate.

In terms of specific sub-scenarios, government subsidies appear to be necessary for large-scale resident adoption of sustainable strategies, as shown by the near-impossibility of achieving a 0.8 final adoption rate in the sub-scenario where subsidies are not implemented. In addition, a comparison of Figure 7a,b reveals that the marginal promotion effect of information campaigns on residents’ sustainable behaviors is more significant than that of the green labeling scheme, which is also consistent with the difference in the dynamic evolutionary trend of adoption rates observed in Scenarios 2 and 3.

The policy effects of jointly implementing government subsidies and information campaigns show the most significant policy complementary effects. It can be observed from Figure 7b that as the intensity of both government subsidies and information campaigns increases, there is a monotonic increase in the final adoption rate. In addition, the shapes of the left and right sides of the surface plot suggest that the impact of these two types of policies on the final adoption rate is similar. The results here, combined with those in Figure 7a,c, can be summarized that combination of government subsidy policies and information campaign policies can promote the adoption of sustainable lifestyles more stably and positively compared to green labeling schemes.

4.7. Sensitivity Analysis

To evaluate the robustness of our simulation results and address concerns about parameter dependency, we conducted comprehensive sensitivity analyses examining both structural assumptions and key parameter variations. The sensitivity analysis serves two primary purposes: first, to validate the stability of our main findings under different model configurations, and second, to identify the relative importance of different parameters in driving sustainable behavior adoption.

Figure 8a presents the sensitivity analysis results for different network topologies. The comparison between scale-free, small-world, and random networks reveals that our model exhibits remarkable stability across different network structures. The final adoption rates remain consistently high across all three network types, with scale-free networks achieving 0.616, small-world networks reaching 0.653, and random networks attaining 0.600. The narrow range of variation (0.053 difference between highest and lowest) demonstrates that our policy effectiveness findings are robust to network topology assumptions. Figure 8b examines the sensitivity to different strategy update mechanisms, comparing asynchronous and synchronous updating approaches. The results demonstrate strong robustness, with asynchronous updating yielding a final adoption rate of 0.619 and synchronous updating achieving 0.669. The minimal difference (0.050) between these two fundamentally different updating mechanisms indicates that our model’s core dynamics are driven by the underlying behavioral and policy mechanisms rather than by specific computational implementations.

Figure 9 provides detailed sensitivity analysis for five key parameters across a ±10% variation range. The analysis reveals differential sensitivity patterns that offer important insights into model behavior. The basic utility parameter ($U^{\prime }$) exhibits the steepest sensitivity, with final adoption rates declining from approximately 0.78 to 0.53 as the parameter increases by 10%. Such sensitivity is theoretically justified, as $U^{\prime }$ represents the fundamental utility of rejecting sustainable behavior, directly influencing the cost–benefit calculation that drives individual decisions. The cost-sharing factor ($b$) shows moderate negative sensitivity. The environmental coefficient ($k_e$) demonstrates positive sensitivity with relatively gentle slopes, indicating that while environmental preferences matter, the model is not excessively dependent on precise calibration of these values. The green degree parameter ($g$) shows similar moderate positive sensitivity, confirming that product characteristics influence adoption but do not create unstable model behavior. Importantly, all parameters maintain adoption rates within reasonable bounds (0.51–0.78) across the ±10% variation range, indicating that the model does not exhibit excessive sensitivity to parameter calibration errors.

In summary, the overall stability of outcomes within reasonable ranges suggests that our model provides reliable guidance for policy design even when exact parameter values may be uncertain in practice.

5. Discussion

Residents will undoubtedly shape the path towards sustainable development in complex and meaningful ways. They are influential stakeholders in public policy and have substantial control over their daily lifestyles and consumption behavior. Therefore, understanding the decision-making mechanisms of sustainable lifestyles and, thus, tapping into residents’ potential to reduce emissions is central to meeting sustainable development goals and net-zero reduction targets. To this end, we construct a network-based evolutionary game model for sustainable lifestyle decision-making and model multiple policy scenarios. A case study of a green express packaging purchase decision in Wuhan is used for parameter initialization of the model. The proposed model can be a practical support for policymakers and provide directional answers to explorative “What-if” policy scenarios.

Going back to the research questions presented in the Introduction, we first conducted simulations in the baseline scenario and presented results on the spatiotemporal evolution of sustainable lifestyle decisions. Doing so provides insights into the evolutionary mechanisms and trends of sustainable behavior adoption. Trends in spatiotemporal evolution reveal that social interactions are critical in the adoption of sustainable lifestyle. Specifically, residents do not make decisions directly based on changed payoffs but make adoption decisions indirectly through social interactions with neighboring residents in the social network. Thus, the micro-mechanism of behavioral public policy on sustainable lifestyle adoption can be considered as increasing the payoffs of adopters to attract rejecters to switch their behavior. From a network-based evolutionary game perspective, we revealed the underlying mechanisms of sustainable lifestyle adoption, which provides theoretical support for further scenario analysis.

In Scenario 1, we simulated the dynamic evolution of sustainable lifestyle adoption by residents under the government subsidy scenario. The results show that government subsidies positively contribute to the adoption of sustainable lifestyle. Our simulations on network size also suggest that the incentive effect of government subsidies is sensitive to network size, which is reflected in the weaker incentive effect with larger network size. This finding may, to some extent, explain the regional heterogeneity of policy effects in reality, that is, the same policy is implemented in different regions with different outcomes. For instance, similar incentives for photovoltaic installations have achieved the opposite outcome of photovoltaic diffusion in Germany and Switzerland [65]. Additionally, our results remind policymakers of the need to consider social network conditions and to tailor policy intensity to regional realities when implementing local subsidy policies to intervene in resident behavior.

The results on the facilitative effect of government subsidies on sustainable behavior are consistent with the study by [15] They conducted policy simulations in a resident network model and found that the increase in government subsidies effectively promoted green behavior, which supports the validity of our finding. Our results additionally reveal the important role of network size on incentive effects. In terms of incentive effects, previous empirical studies on behavioral interventions have been controversial regarding the effectiveness of incentives in promoting behavior change. While much of the scientific evidence emphasizes the effectiveness of incentives [31], other studies have argued that incentives do not significantly promote sustainable behavior change [21,66]. Our results suggest that differences in the size of complex social networks may account for this variation in incentive effectiveness across domains.

Scenario 2 provides insight into the dynamic evolution of sustainable lifestyle adoption in the green labeling scheme scenario. We find that increasing the intensity of this policy promotes the sustainable decisions of resident group. However, the simulation of Scenario 4 further reveals that the promotion effect of the green label is not significant compared to the other two policies. In terms of validating the effectiveness of the green labeling scheme, our results are similar to a great deal of prior theoretical and empirical studies on labeling effects that have highlighted the positive impact of labels on changing consumption patterns (e.g., [63,64]). More importantly, our study also reveals the relative effectiveness of different policies by modeling multiple policies within the same framework. Regarding the promotion of green products, the lack of relative effectiveness of green labeling schemes, despite being positive, implies that they are more suitable to be implemented as part of a policy mix.

In Scenario 3, we investigate the impact of information campaign policies on the evolution of sustainable lifestyle decisions. The results show that increased information campaigns do promote the adoption rate in terms of both evolutionary trends and evolutionary outcomes, and this effect is similar to that of government subsidies. The results are supported by the findings of [14], who constructed a multi-agent model for residents’ sustainable lifestyle decisions. Simulation results based on the model suggest that government publicity activities are highly effective in promoting sustainable lifestyles after a period of time. Actually, despite widespread pro-environmental behavior change-related campaigns by international organizations and civil societies, surveys show that current resident sustainability education is low, and knowledge and understanding of sustainability issues are inadequate [67]. Furthermore, given that subsidy incentives are often costly and more likely to be limited by the area of implementation, promoting adoption through information campaigns may be a breakthrough point for governments’ behavior change interventions.

Simulations on the policy mix in Scenario 4 not only validate findings in single-policy scenarios but also reveal the complementary effects of government subsidy policies implemented together with information campaign policies. In other words, the simultaneous implementation of these two policies enhances the aggregate outcomes of a policy mix and facilitates achieving the desirable performance. This finding clearly demonstrates that, as many policy studies have suggested, the combination of economic incentives and information campaigns can significantly improve the performance of policy mixes [19,65]. In this regard, an example of green building retrofitting is that if only subsidy incentives are implemented without information campaigns, residents may not know how to access government financial support, thus inhibiting the incentive effect of the policy [68].

While our simulation results demonstrate the theoretical effectiveness of simultaneous policy implementation, translating these findings into practical deployment requires consideration of several operational factors. First, our findings regarding the superior complementary effects of government subsidies and information campaigns suggest that policymakers should prioritize the coordinated implementation of these two policy types. The challenge lies in ensuring that both policies reach the same target population simultaneously and that residents can clearly understand the connection between available information and financial incentives.

Second, successful policy mix implementation requires robust institutional coordination mechanisms. Government subsidies typically involve financial and regulatory agencies, while information campaigns often fall under environmental or public communication departments. Establishing clear communication channels between relevant agencies and developing unified messaging strategies can help ensure that policy combinations achieve their intended complementary effects rather than creating confusion or administrative burden for residents.

Third, our results highlighting the network size sensitivity of policy effectiveness suggest that policymakers should consider local social network characteristics when designing implementation strategies. In smaller, more tightly connected communities, lower policy intensities may achieve desired outcomes, while larger networks may require higher policy intensities to reach similar adoption levels. This implies the need for flexible implementation frameworks that can be adapted to different community contexts and network structures.

Finally, resource allocation strategies should reflect the synergistic nature of these policy combinations. Rather than optimizing individual policies in isolation, policymakers should consider the combined effectiveness when determining resource distribution between different policy instruments. Our findings suggest that balanced investment in both subsidies and information campaigns may yield superior outcomes compared to heavily investing in a single policy type.

6. Conclusions and Policy Implications

Residents’ sustainable lifestyle and recycling decisions, influenced by complex social interactions alongside internal and external factors, significantly impact the sustainable transition. Understanding the underlying mechanisms and dynamics of these decisions, as well as exploring how multiple policies contribute to sustainable lifestyle adoption, represents a critical research priority. We address this need by developing a network-based evolutionary game model for sustainable lifestyle decisions in multiplex social networks. Four policy scenarios, including government subsidies, green labeling schemes, information campaigns, and policy mixes, are developed to investigate the dynamic and directional effects of policies on sustainable lifestyle adoption. The following conclusions can be drawn from the forward-looking simulation of policy scenarios.

(1)

Social interactions of residents, especially behavioral interactions, are crucial for sustainable lifestyle and recycling decisions. The impact of exogenous shocks on resident group decision-making is realized through residents’ connections in social networks and repeated social interactions.

(2)

The differences in the group evolutionary trends of residents’ sustainable lifestyle decisions under different policy scenarios are mainly reflected in the stabilization of adoption rates at different equilibrium levels after a rapid increase.

(3)

Government subsidies positively promote the adoption of sustainable recycling behaviors. This facilitation effect is sensitive to the size of social networks, as reflected by the evolutionary outcome of adoption rates decreasing with increasing network size. Policymakers need to design subsidy incentives tailored to the realities of social networks.

(4)

Information campaigns are effective and robust in promoting the adoption of sustainable lifestyle. They have a more pronounced marginal promotion effect than green labeling schemes and have a cost advantage over subsidy incentives.

(5)

Implementing government subsidies and information campaigns as policy mixes can exert complementary effects and enhance the aggregate outcomes of the government policy package. Green labeling schemes are also suitable to be implemented as part of policy mixes.

Based on these findings, several practical recommendations emerge for policymakers. Information campaigns should be prioritized as primary policy instruments due to their robust effectiveness across different network contexts and significant cost advantages over subsidy programs. When implementing subsidy programs, policymakers should adopt network-sensitive approaches, applying higher intensities in larger networks and modest levels in smaller, tightly knit communities where social influence amplifies policy effects. Policy combinations, particularly subsidies with information campaigns, should be pursued through coordinated inter-agency implementation to maximize complementary effects. Green labeling schemes should be primarily utilized as components within broader policy packages rather than standalone interventions.

This study makes three distinct theoretical and practical contributions to sustainable lifestyle behavior research. Theoretically, we advance the field by introducing a novel methodological framework that integrates multiplex network structures with evolutionary game theory, enabling simultaneous modeling of information diffusion and behavioral interaction dynamics that existing studies have not adequately captured. This integration reveals how social interactions fundamentally shape sustainable behavior adoption through network-mediated evolutionary processes rather than individual rational choice alone. Practically, our research provides the comprehensive comparative analysis of multiple policy instruments within a unified modeling framework, demonstrating the superior effectiveness of information campaigns and the network-sensitive nature of subsidy policies, while revealing specific complementary effects in policy combinations. The novel combination of the network-based evolutionary game model and the ABM method allows for dynamic insights into sustainable behavior decision-making processes that overcome current limitations in modeling behavioral interactions among residents. Although the simulation results of the proposed model are not the ultimate reality of policy implementation, modeling and simulation of residents and policies targeting residents can still inform policymakers and help them navigate an uncertain future by providing directional guidance on policy effectiveness and design principles.

Although the present study is conducive to a more nuanced understanding of the decision-making process and policy options related to sustainable behavior, several important limitations must be acknowledged when translating these findings to real-world policy contexts. First, our model assumes homogeneous information processing and uniform responsiveness across all residents, while real-world populations exhibit significant heterogeneity in educational backgrounds, income levels, and environmental attitudes that may affect policy effectiveness. Second, the simulation framework simplifies administrative and institutional complexities inherent in policy implementation, such as bureaucratic delays, eligibility restrictions, and budget constraints that can reduce theoretical effectiveness. Third, our model assumes stable social networks and policy environments, while actual implementation occurs in dynamic contexts where network structures evolve and external events influence behavioral priorities. Additionally, there are still some limitations due to the complexity of the decision-making problem. The current model simplifies the socioeconomic relationships in the adoption of sustainable behavior and inevitably abstracts away some details of decisions, which are more complex in the real world. Our focus on sustainable recycling behavior may also limit generalizability to other sustainable lifestyle domains that involve different degrees of social visibility, economic costs, and habit formation requirements.

Future research could extend this model in a multidisciplinary context, for example, by combining psychology and behavioral economics to construct utility functions to make the decision model more realistic. In addition, given the attitude–behavior gap in sustainable behavior adoption, extending the current framework by analyzing the attitude attributes of residents and the interplay between attitudes and behaviors would be a direction for improvement. Moreover, incorporating dynamic network structures and heterogeneous agent characteristics could enhance the model’s realism and policy relevance. We believe that further research in these directions would be beneficial to better understand and facilitate the transition to sustainability. Therefore, adaptive management approaches and pilot testing programs are essential when translating these theoretical insights into practical policy implementations.