Environmental Awareness and Social Sustainability: Insights from an Agent-Based Model with Social Learning and Individual Heterogeneity

Abstract

Social sustainability requires both technological innovations and societal changes within energy systems, with decentralization playing a critical role. This shift emphasizes the increasing importance of individual user decision-making, posing significant management challenges. An individual’s environmental awareness has a key influence on their energy decisions. However, the relationship between individual environmental awareness and social sustainability, particularly from a systemic perspective, remains underexplored. Our study uses agent-based modeling to examine this relationship within Japan’s electricity market, focusing on social learning and consumer heterogeneity. We find that social learning leads to the formation of consumer clusters with specific electricity preferences, affecting environmental awareness differently across high- and low-carbon groups. This process reveals the nuanced role of social learning in promoting low-carbon technology adoption, which varies according to the market share of low-carbon energy. Additionally, our results suggest that initial heterogeneity in environmental awareness among consumers has a limited and varied effect on sustainable transition pathways. However, the diversity resulting from social learning significantly shapes these trajectories. These insights highlight the complex interplay between individual behaviors, societal dynamics, and technological advancements in steering the sustainable transition, providing valuable considerations for future energy system management.

1. Introduction

To achieve social sustainability, it is crucial to modify and evolve our energy systems through both technological innovation and societal change [1]. Distributed systems that integrate renewable energy sources, microgrids, and vehicle-to-grid (V2G) technology, among others, exemplify such advancements [2], highlighting decentralization as a key trend for the future of energy systems [3]. A significant challenge lies in understanding the decision-making processes of energy consumers as individual users within the broader context of energy management.

In previous studies [4,5], we employed a model centered on user decision-making to analyze the social sustainability process and the influence of policy on individual decision-making within the Japanese power sector. However, these studies have certain limitations, the most notable being the oversimplified treatment of the mechanisms underlying users’ environmental awareness. While environmental awareness is present in individuals, the relationship between this awareness and their behavioral decisions is complex and not straightforward [6]. The objective of this study is to conduct an in-depth analysis of the mechanisms through which individual environmental awareness affects the process of social sustainability.

1.1. Literature Review

Environmental awareness involves an individual’s understanding and perception of the complex interactions between humans and the environment [7]. Iizuka (2000) [8] conducted a comprehensive review of studies on environmental awareness, which primarily focused on three key aspects: the relationship with behavior, methods of evaluation, and influencing factors.

●

Connection between environmental awareness and behavior. The connection between environmental awareness and behavior is complex. Individuals with higher environmental awareness are often more influenced by price and personal preferences when practicing sustainability [6]. However, heightened awareness does not always result in eco-friendly behavior. Rodriguez (2007) [9] showed that increased awareness can sometimes boost sales of non-eco-friendly products. Yang et al. (2021) [10] investigated the link between environmental awareness and pro-environmental behavior under varying institutional influences. Kousar et al. (2022) [11] emphasized that the relationship between environmental awareness, behavior, and environmental quality improvement is not linear. On the demand side, environmental awareness can influence the availability of greener products on the supply side (Rustam et al., 2020 [12]; Zhang et al., 2015 [13]). However, eco-friendly products often come with higher costs, which may require subsidies to remain competitive in the market (Yu et al., 2016 [14]). Interestingly, non-eco-friendly products can still succeed, even as awareness grows (Liu et al., 2012 [15]). Regulations play a crucial role in harnessing environmental awareness within various macroeconomic contexts (Bohdanowicz, 2006 [16]).

●

Methods for evaluating environmental awareness. Evaluation methods include a variety of approaches, such as measuring public concern (Mei et al., 2016 [17]), attitudes, willingness-to-pay (WTP), and indicators of the human–environment relationship (Duroy, 2005 [18]). Okada et al. (2019) [19] assessed national awareness by analyzing willingness to purchase electric vehicles (EVs). Similarly, Nomura and Akai (2004) [20] employed WTP in their study on promoting green electricity. Huang et al. (2006) [21] examined household equipment in China, with a focus on the willingness to pay for eco-friendly options. Other noteworthy methods include qualitative surveys on environmental attitudes (Wong, 2003 [22]) and proportional analysis of the influence of environmental awareness (Itsubo et al., 2018 [23]).

●

Influencing factors of environmental awareness. Influencing factors vary across different contexts. Palmer et al. (1998) [24] identified a strong correlation between environmental awareness and individual experiences across eight countries. Ziadat (2010) [25] examined Jordan and found that factors such as gender, age, education, location, and socioeconomic status significantly influence environmental awareness, underscoring its heterogeneity. This diversity can also be amplified by promoters such as legislators, environmental groups, and organizations (Gadenne et al., 2009 [26]). Individual interactions and education play a crucial role as well. Uzunboylu et al. (2009) [27] highlighted the impact of mobile learning on enhancing environmental awareness, while Severo et al. (2019) [28] examined the role of social networks in fostering awareness. Mass media campaigns (Jharotia, 2018 [29]; Kushwaha, 2015 [30]) have proven effective in promoting national environmental awareness, complementing social learning. Education remains a vital tool for raising environmental awareness, especially among students (Schmidt, 2007 [31]; Simsekli, 2015 [32]), and is often embedded in national strategic plans for improving environmental consciousness (Crotty and Hall, 2014 [33]). At a national level, the interplay between environmental governance, regulation, and increased awareness is critical [34,35]. Chen et al. (2019) [36] found that heightened government environmental awareness leads to substantial benefits for environmental policies and regional environmental quality. In Nigeria, Ogunbode and Arnold (2012) [37] emphasized the importance of helping individuals recognize the connection between biodiversity loss and environmental behavior.

1.2. Research Gap

Despite the extensive body of research on environmental awareness, several gaps remain. First, existing studies indicate that the relationship between increased environmental awareness and social sustainability through the adoption of environmentally friendly behaviors is complex and not fully understood. Moreover, there has been limited investigation into the role of consumers’ environmental awareness in driving social sustainability within the electricity sector—a critical area that warrants further exploration. Second, while research provides statistical insights into the heterogeneity of environmental awareness within populations, the impact of this heterogeneity on the adoption of sustainability practices is still unclear. Finally, although studies highlight the importance of policy interventions and education in raising environmental awareness, there is a lack of quantitative analysis on the relationship between the strength of social learning effects and the improvement of individual environmental awareness.

To address these research gaps, this paper aims to answer the following research questions:

• How do social learning effects and individual heterogeneity influence changes in environmental awareness?
• How do social learning effects and individual heterogeneity impact the process of sustainable development in the power sector?

1.3. Contributions

The methodology used in previous studies on environmental awareness has predominantly relied on statistical analysis. However, such methods often fall short in revealing the intricate relationships between changes in environmental awareness, shifts in individual behavior, and the progress of sustainable development. To address this, we developed a model called EMPAS (Electricity Market incorporating subsidy Policy, environmental Awareness, and demanders’ Satisfaction). EMPAS is an agent-based model (ABM) that differs from other existing models for analyzing sustainable development processes, as outlined in

Table 1. Comparison of different simulation methods in sustainable development process.

Feature	Integrated Assessment Modelling (IAM)	Dynamical Systems Model (DSM)	Agent-Based Model (ABM)
Definition	Integrates models from multiple fields to assess environmental, economic, and social impacts	Describes the evolution of a system over time using mathematical equations	Simulates interactions and behaviors among individual agents
Representation	Suitable for macro-level policy and strategic decision-making	Suitable for analyzing overall system dynamics	Suitable for micro-level individual behavior and interactions
Modeling Approach	Integrates models from multiple disciplines, often involving nonlinear equations	Modeled through a set of differential or difference equations	Modeled by defining rules and behaviors for individual agents
Data Requirements	Requires large amounts of data and expert knowledge	Relies on initial conditions and parameter values	Depends on assumptions and rules about individual behavior
Advantages	Capable of considering multiple factors, providing holistic insights	Simple models, easy to understand and analyze	High flexibility, able to simulate complex individual behaviors and interactions
Disadvantages	Complex and time-consuming to build and calibrate	May not capture complex individual behaviors and interactions	Model validation and generalization can be challenging

. The core contributions of this paper are as follows: (a) the introduction of willingness-to-pay (WTP) for green electricity as a measure of environmental awareness among users in the power sector; (b) the incorporation of mechanisms of change in users’ environmental awareness through the impacts of social learning; and (c) the quantitative analysis of how individual differences in environmental awareness affect social sustainability.

1.4. Organization

The remainder of the paper is organized as follows: In Section 2, we introduce the EMPAS model. Using this model, we simulate the impact of environmental awareness on the social sustainability process, with a focus on the roles of social learning and individual heterogeneity, as discussed in Section 3. Finally, Section 4 provides a summary of our findings and their implications.

2. Methodology and Model

2.1. Model Overview

To explore the role of individual environmental awareness in the social sustainability process in greater depth, we developed the EMPAS model. EMPAS builds on the foundational structure of a predecessor—an electricity market model designed for low-carbon transition (LCT) [4,5]. Figure 1 illustrates the methodology and model structure used in this study. Figure 1a depicts the model’s components and inclusion mechanisms. In 2016, Japan’s electricity market underwent full liberalization for all consumers [38], allowing for the free trade of electricity. EMPAS includes two main types of agents: energy demanders and energy suppliers in the electricity market. The electricity demanders represent various customer segments purchasing electricity, while the suppliers are companies selling different types of electricity. Both demanders and suppliers possess distinct attributes that define their individual characteristics. Additionally, the model includes a macro-environmental agent representing macro-level parameters such as policy, economy, and technology, which influence agent behavior and decision-making.

In the model, agents behave and make decisions based on different mechanisms. The key innovation in this study, compared to previous work, is the updated mechanism for adjusting users’ environmental awareness (highlighted in red in Figure 1a). The operation of this mechanism within EMPAS occurs across multiple time scales. As illustrated in Figure 1b, the simulation runs over multiple rounds. Each round represents a fast process, simulating activities over one year ($t=Y$), comprising two phases: the signing phase (the decision-making mechanism of the demanders and the mechanism of the competition of the suppliers take place in this phase), where the calculation step is denoted by index n with $n\in \left[0,N^{*}\right]$, and the growing phase (suppliers’ investment mechanisms occurring at the phase), where the calculation step is denoted by index m with $m\in \left[0,M^{*}\right]$. After each fast process, all agents will enter the adjustment mechanism. Multiple fast processes combine to form a slow process, generating yearly dynamics, with the time parameter defined as $t$, covering the period from 2011 to 2040.

Figure 1c illustrates the flow of the simulation, encompassing both the fast process and the slow process development, as well as the integration of the two. In the following sections, Section 2.2 provides a detailed description of the properties of the agents in the model. Section 2.3 explains the mechanisms within the model, and Section 2.4 covers the calibration and validation of the model parameters.

2.2. Agent and Global Environment

2.2.1. Electricity Demander

In EMPAS, energy demanders represent one of the two key agent types, functioning as participants in the electricity market who both purchase and consume energy. These demanders are classified into three distinct subtypes: high-voltage demanders, which cater to large-scale factories and buildings; mid-voltage demanders, which serve medium-sized factories and buildings; and low-voltage demanders, which correspond to households and small businesses. The demanders are positioned in a discrete 32 × 32 grid, a size and connectivity referenced from [39]. This grid structure appropriately reflects the diversity of users in the electricity market trading process while ensuring model convergence and maintaining an acceptable level of computational complexity. While this study focuses on the current grid size to address the research questions, we acknowledge that larger and more realistic networks represent a key direction for future simulations in this field. Each grid is labeled by an index ($j\in [\mathrm{1,1024}$]), which identifies the specific location of each demander. The social learning effect among demanders is influenced by interactions occurring between neighboring grids. Each demander agent is characterized by a range of attributes, including the following:

• Types of demanders denoted as $DT\left(j\right)$. $DT\left(j,n,t\right)=-1$ represents high-carbon energy demanders, $DT\left(j,n,t\right)=1$ represents low-carbon energy, and $DT\left(j,n,t\right)=0$ represents nuclear demanders, respectively.
• Capacity of demand denoted as $DC\left(j,t\right)$. Sum of energy demands of all the demanders equals the total annual electricity demand in Japan.
• Environment sensitivity denoted as $ES\left(j,t\right)$. This sensitivity refers to the extent to which demanders care about environmental protection, hence having its influence on their environmental awareness.
• Willingness-to-pay for low-carbon energy denoted as $WTP(i,j,n,t)$. This is a quantitative measure of environmental awareness of demanders. Index $i$ is for energy type.
• Eco-fighter tag denoted as $EFI\left(j\right)$. This binary tag defines if the demander is an environmentalist.

2.2.2. Electricity Supplier

Electricity suppliers represent the second type of agent in the model, embodying entities responsible for selling and distributing energy within the electricity market. There are three distinct suppliers, labeled $i\in \left[\mathrm{1,3}\right]$; these suppliers use different energy sources: high-carbon (such as coal, natural gas, and oil), low-carbon (including wind, biomass, geothermal, and solar), and nuclear energy. They are able to move freely within the discrete space of demanders, facilitating the delivery and unrestricted sale of electricity. The key attributes of the suppliers include the following:

• Capacity of a supplier denoted as $SC\left(i,n,t\right)$. This indicates the available installed capacity required for supplier’s power generation.
• Willingness-to-invest denoted as $IW(i,m,t)$. This means suppliers’ willingness-to-invest while gaining profits in the growing phase of the fast process.
• Profit, represented as $P(i,m)$, indicates the earnings generated by energy suppliers through the sale of electricity.
• Cost of electricity generation denoted as C(i, t).
• Environmental value denoted as $EI\left(i\right)$. Low-carbon and nuclear energy are considered environmentally friendly and hence have a higher value.

2.2.3. Global Environment

Attributes of the global environmental agent are as follows:

• Advocacy and legal intensity of policies denoted as $ALI$, which indicate policy intensity to incentivize demand-side use of low-carbon energy.
• Annual rate of change of GDP denoted as $GDP\left(t\right)$.
• Average and variance of willingness-to-pay of demanders denoted as ${\mu }_{WTP}\left(t\right)$ and ${\sigma }_{WTP}\left(t\right)$, respectively. These parameters are used to indicate the average level and the heterogeneity of demanders’ environmental awareness.

2.3. Mechanisms

2.3.1. Market Transactions

Market transactions occur within a discrete space, where electricity suppliers interact with energy demanders. These interactions can result in the formation of power sales contracts during the signing phase, influenced by factors such as energy prices, energy types, and the environmental awareness of the demanders. Once the agreement is in place, suppliers start generating profits, while demanders begin consuming electricity during the subsequent growth phase.

Signing Phase

At the start of the simulation ($t=0$), demander $j$ is initially in a state of no contract, as illustrated in Figure 2a. Once the signing phase of the first fast process begins, demander $j$ interacts with supplier $i$, and the two parties enter into a contract, as shown in Figure 2b. The fast process continues until $n=N^{*}$, during which demander $j$ may encounter multiple suppliers, potentially leading to a change in contracts. This process simulates the decision-making of the demander and the competition among suppliers. It is important to note that during each signing phase of the fast process, demanders may alter their contracts due to updates in agent attributes and influences from the global environment. The specific conditions under which demanders may change their contracts are as follows:

• As shown in Equation (1), if supplier i is matched with demander $j$ at step n of the simulated fast process in year t, the supplier’s supply capacity ($SC$) must exceed the demander’s demand capacity ($DC$).

  $$SC\left(i,n,t\right)\ge DC\left(j,t\right)$$

  (1)
• As shown in Equation (2), the comprehensive price ($CP$) offered by the new supplier $i$ at step $n$ of the simulated fast process in year t must be lower than the price offered by the original supplier $i^{\prime }$ at step $n-1$.

  $$CP\left(i,j,n,t\right)<CP\left(i^{\prime },j,n-1,t\right)$$

  (2)

Here, the comprehensive price is composed of both the price of the energy supplied and the demander’s perception of the energy, as expressed in Equation (3).

$$CP\left(i,j,n,t\right)=C(i,t)+WTP(i,j,n,t)$$

(3)

$C\left(i,t\right)$ represents the actual price, which is composed of both fixed costs ($FC$) and variable costs ($VC$). Fixed costs include expenses such as power transmission, transformation, labor, and other operational overheads, while variable costs cover expenditures like construction and raw materials.

$$C\left(i,t\right)=VC\left(i,t\right)+FC\left(i\right)$$

(4)

Demanders’ willingness-to-pay for low-carbon energy, denoted as $WTP(i,j,n,t)$, is calculated as a virtual price metric that reflects the environmental awareness of the demander.

$$WTP\left(i,j,n,t\right)=EI\left(i\right)·ES\left(j,t\right)$$

(5)

The parameter ranges within the described mechanism are defined as follows: $t\in [2011,2040]$ represents the simulation years; $i \in [1, 3]$ corresponds to different categories of suppliers; $j \in [1, 1024]$represents the various demanders; and $n \in [0, N^{*}]$ indicates the calculation step for the signing phase. Parameter $N^{*}$ serves as a calibration factor, with its value taken from [5] to ensure that both parties in the transaction have full contact. When $n=N^{*}$, each demander $j$ will choose energy from supplier $i$ at a specific step $n^{\prime }$, finalizing the contract at that step. This step minimizes the comprehensive price, determined by ${arg min}_{\left\{{i(n}^{\prime }\left)\right|0\le n^{\prime }\le N^{*}\right\}}CP\left(i,j,t\right)$. Once this step is completed, the contracting phase concludes, as illustrated in Figure 2c.

Growing Phase

During the growth phase, suppliers generate profits from electricity sales, which boosts their willingness-to-invest (as shown in Equation (6)). This, in turn, influences the expansion of their installed capacity in the subsequent year. The construction of this mechanism is based on the framework outlined in [40].

$$IW\left(i,m,t\right)=IW\left(i,m-1,t\right)+EG\left(t\right)\cdot \Theta$$

(6)

$$\mathrm{E}\mathrm{G}\left(\mathrm{t}\right)={\displaystyle \frac{100+\mathrm{G}\mathrm{D}\mathrm{P}\left(\mathrm{t}\right)}{100}}$$

(7)

$\Theta$ represents the translation of willingness-to-invest into installed capacity growth. The growth in willingness-to-invest is also influenced by macroeconomic conditions, where $EG$ represents economic growth, as shown in Equation (7). $\Theta$ serves as a calibration parameter, with its value referenced from [5].

At the conclusion of the growth phase, market transactions are completed, signaling the end of the fast process. Following these transactions, all agents will update their attributes based on the outcomes (refer to Section 2.3.2 for further details). These adjustments will influence the evaluation of the comprehensive price in the next round of the fast process.

2.3.2. Agent Adjustment

Environmental Awareness Adjustment

Building on prior research (Section 1), changes in environmental awareness among demanders are driven by three primary factors: policy (including education, advocacy, diffusion, and incentives such as carbon taxes), macroeconomics (which influences individual incomes), and the effects of social learning. Within the model, the parameter $ALI$ represents the intensity of policy factors. To simulate its impact, we introduce a parameter PW to control the extent of its influence. $EG\left(t\right)$, as defined in Equation (7), represents the macroeconomic component. Importantly, during economic downturns and periods of reduced income, individuals may show less enthusiasm for financial commitments to environmental protection—a phenomenon supported by Sundt and Rehdanz (2015) [41] and Gao et al. (2020) [42]. In the model, parameter EW is incorporated to adjust the significance of the macroeconomic factor.

Social learning reflects the influence of societal dynamics, including how individuals are shaped by their immediate surroundings. In the model, social learning refers to the mechanism by which the energy choices of neighboring demanders affect an individual’s environmental awareness. When nearby neighbors choose low-carbon energy, an individual’s environmental awareness is enhanced through social learning and imitation—a phenomenon well-documented in the literature on social networks, as discussed by Will et al. (2020) [43]. Conversely, demanders who favor high-carbon energy also influence their neighbors, but with an opposing effect. This relationship is captured by $LCPS(j,t)$, and the parameter SW determines the weight of the social learning influence. Each demander considers the impact of eight or four surrounding neighbors, as shown in Figure 3. This setup allows for an analysis of varying intensities of social learning. When social learning is deactivated, SW is set to zero. In a scenario of weak social learning, $LCPS(j,t)$ follows Equation (8), while a strong social learning scenario utilizes Equation (9).

$$LCPS\left(j,t\right)={\displaystyle \frac{DT\left(j-1\right)+DT\left(j+1\right)+DT\left(j-D\right)+DT\left(j+D\right)}{4}}$$

(8)

$$LCPS\left(j,t\right)={\displaystyle \frac{DT\left(j-1\right)+DT\left(j+1\right)+DT\left(j-D\right)+DT\left(j+D\right)+DT\left(j-D-1\right)+DT\left(j-D+1\right)+DT\left(j+D-1\right)+DT\left(j+D+1\right)}{8}}$$

(9)

Here, $D$ signifies the numerical gap between vertically adjacent demanders—set at 32 in this study. The value of $D$ indicates that the network structure in EMPAS is a regular network, as referenced in [39]. In this type of network, each agent is connected to only a few neighboring agents, forming a ring or lattice structure, which is commonly used to model spatial near-neighbor interactions, such as geographic relationships. This reflects the dependence of renewable energy applications on geographical features. Modifying the value of $D$ would alter the overall network structure and, consequently, the role of social learning. Possible network configurations could include fully connected networks, stochastic networks, or scale-free networks. However, network restructuring is beyond the scope of this paper and will be the focus of our future research.

Demanders’ environmental awareness, denoted as $WTP(i,j,n,t)$, is described in Equation (5). Its adjustment is influenced by changes in the demander’s environmental sensitivity $ES(j,t)$, as illustrated by Equation (10). This adjustment mechanism builds on our previous study [5], where we distinguished the impacts of policy ($ALI$), macroeconomic conditions ($EG$), and social learning ($LCPS$), and assigned corresponding weight coefficients to each: PW, EW, and SW. These coefficients satisfy $PW+EW+SW=1$, with each coefficient ranging between 0 and 1. A coefficient of 0 indicates that the corresponding factor has no influence. Additionally, we adjusted the mechanism for environmentalists; their environmental awareness remains constant and is unaffected by this adjustment mechanism. In line with Williams’ 2013 study [44], environmentalists make up approximately 20% of the population.

$$ES\left(j,t\right)=ES\left(j,t-1\right)-PW·ALI\left(t-1\right)-EW·EG\left(t-1\right)-SW·LCPS\left(j,t-1\right)$$

(10)

Adjustment of Other Attributes

Beyond environmental awareness, various attributes also undergo adjustments, including the demander’s electricity consumption refer from [45], supplier’s available installed capacity, macroeconomic circumstances, and energy costs. These adjustments transpire following the fast process. As delineated in Figure 1b,c, the simulation of numerous fast processes yields an overarching slow progression. Once the ultimate fast process concludes, all simulations reach their culmination, and the model furnishes its outcomes.

2.4. Validation and Calibration

Net Logo, an open-source software environment, was used to execute EMPAS (Tisue S, Wilensky U. 1999 [46]). To accommodate the inherent stochasticity of agent-based modeling, in each round of simulation the experiment was iterated 100 times, and the outcomes were calculated as median (Lee J S et al. 2015 [47]). The model was fine-tuned by utilizing a random search methodology (Solomatine D P 1999 [48]). The parameters, once calibrated, were subsequently employed to simulate the low-carbon transition progression within Japan’s electricity sector, spanning the years 2010 to 2020. The congruence of the outcomes with the empirical results attests to the model’s validation [4,5].

In terms of calibrating the mechanism of environmental awareness change, reference was drawn from Gao et al. (2020) [42]. This study presents the 2015 WTP values for residents across diverse regions in Japan and computes the WTP projections for 2030 under disparate macroeconomic conditions. A comparison between these findings and the EMPAS simulations is presented in

Table 2. Comparison of demanders’ environmental awareness (WTP) change with literature results.

Gao et al. (2020) [42]	2015 Baseline	2030 Economy Stable	2030 Economy Growth	EMPAS	2015 Baseline	2030 Economy Stable	2030 Economy Growth
WTP range $(\mathrm{J}\mathrm{P}\mathrm{Y}/\mathrm{K}\mathrm{w}\mathrm{h}$)	1.19~5.07	2.76~9.01	3.80~11.21	$\mu$	1.89	3.12	4.85
				$\sigma$	2.25	2.28	2.64

.

$\mu$ signifies the mean of the demander’s $WTP(i,j,n,t)$, as Equation (11).

$\sigma$ denotes the standard deviation of the aforementioned mean, in accordance with Equation (12).

$J$ denotes the total count of demander agents.

$$\mu ={\displaystyle \frac{\sum _{j}WTP(i,j,n,t)}{J}}$$

(11)

$$\sigma =\sqrt{{\displaystyle \frac{\sum _j{\left(WTP\left(i,j,n,t\right)-\mu \right)}^2}{J}}}$$

(12)

3. Results and Discussion

3.1. Experiments Setup

To address the research questions, a series of model experiments have been devised, as outlined in

Table 3. Experiments setting.

Parameters	Exp1	Exp2	Exp3	Exp4	Exp5
Social learning effect	Off	Weak	Strong	Strong	Strong
$\mathsf{\sigma }$(2021)	2.2	2.2	2.2	2.6	3.0

. Exp1–Exp3 are geared towards scrutinizing the influence of social learning, while Exp3–Exp6 delve into the ramifications of demanders’ heterogeneity. For each experiment, three evaluative metrics have been identified:

The rate of average $WTP(i,j,n,t)$, represented by $\mu$ in accordance with Equation (11).

The standard deviation of $WTP(i,j,n,t)$, denoted as $\sigma$, as detailed in Equation (12).

The propagation of low-carbon energy, encapsulated by $Rl$, as articulated in Equation (13).

$$Rl={\displaystyle \frac{\sum _{j}DT\left(j,n,t\right)}{J}}, DT\left(j,n,t\right)>0$$

(13)

Several assumptions underpin these experiments:

• A consistent decline in the cost of low-carbon energy is presumed, owing to steady technological advancements.
• The utilization of nuclear energy remains stringently regulated due to the aftermath of the Fukushima nuclear power plant incident based on Japanese government’s energy plan [49].
• The stability of international high-carbon energy prices, encompassing commodities like oil, coal, and natural gas, is presupposed.
• The occurrence of extreme weather phenomena, which could potentially disrupt existing renewable energy sources, is excluded from consideration. The model operates on the premise of their uninterrupted function under stable conditions.
• The intensity of policies aimed at fostering national environmental awareness remains invariant ($ALI$ remains constant).
• Given Japan’s economic development, macroeconomic stability is assumed to be maintained, regardless of its impact on individual environmental awareness, meaning $EW=0$.
• Policy and social learning are assumed to have an equal influence on users’ environmental awareness, with $PW=SW=0.5$.

3.2. Social Learning

Figure 4 illustrates the dispersion pattern of low-carbon energy demanders over different years under the influence of strong social learning effects (Exp3). It is evident that social learning prompts demanders to exhibit a clustering tendency in their energy choices, leading to the formation of numerous energy clusters in scenarios characterized by robust social learning. To better quantify this clustering phenomenon, we introduce a novel metric called the non-dimensional boundary area (NBA), which is defined as follows:

$$NBA={\displaystyle \frac{\sum _{j^{*}}DT\left(j^{*},n,t\right)}{\sum _{j}DT\left(j,n,t\right)}}$$

(14)

In this equation, $j^{*}$ represents low-carbon demanders located at the boundaries of clusters, where $DT\left(j^{*},n,t\right)>0$. Additionally, the condition $DT\left(j^{*}-1\right)+DT\left(j^{*}+1\right)+DT\left(j^{*}-D\right)+DT\left(j^{*}+D\right)<4$ must be satisfied. A larger $NBA$ value indicates greater decentralization among low-carbon energy demanders. Conversely, a smaller $NBA$ value suggests a higher concentration of demanders choosing low-carbon energy, leading to the formation of low-carbon energy clusters.

As shown in Figure 5, the simulation results for the $NBA$ and the number of low-carbon energy demanders are presented across different levels of social learning intensity. The findings highlight that as the strength of the social learning effect increases, the overall number of low-carbon energy demanders tends to decrease. Notably, the scenario with the strongest social learning effect exhibits the lowest $NBA$ value, indicating a significant concentration of demanders choosing low-carbon energy and the formation of energy clusters. On the other hand, scenarios without social learning effects display a more dispersed distribution of energy preferences.

3.2.1. Social Learning’s Impact on Demanders’ Environmental Awareness

Figure 6 illustrates the significant impact of social learning on the development of individual demanders’ environmental awareness. In scenarios without social learning effects (Exp1), demanders are willing to pay an average of 2.32 JPY/kWh in 2021 to support low-carbon energy. By 2040, this value increases to 4.12 JPY/kWh, reflecting a substantial rise of approximately 78%. Introducing weak social learning effects (Exp2), where demanders are influenced by their four immediate neighbors—those above, below, to the left, and to the right—the average willingness-to-pay could reach 3.33 JPY/kWh by 2040, indicating a growth of approximately 51%. When social learning effects are intensified further (Exp3), with each individual being influenced by eight neighboring individuals, the average willingness-to-pay decreases to 1.85 JPY/kWh by 2040, marking a significant reduction of approximately 20%. The model’s calculations clearly highlight the profound influence of social learning on the expansion of individual environmental awareness.

The standard deviation in the heterogeneity of individual environmental awareness is shown in Figure 7. In both Exp1 and Exp2, the standard deviation of individual willingness-to-pay remains relatively stable throughout the simulation. However, a significant increase in heterogeneity is observed in the Exp3 scenario, which is characterized by a strong social learning effect (as also depicted in Appendix A, Figure A1). This phenomenon can be attributed to the clustering effect that arises from shifts in energy preferences under the influence of strong social learning. Within low-carbon clusters, individuals demonstrate a rapid increase in willingness-to-pay, while in high-carbon clusters, willingness-to-pay tends to decline. This contrast contributes to the overall rise in heterogeneity.

As the social learning effect intensifies, the curve representing the growth of environmental awareness shifts from a linear to a nonlinear pattern. This change is exemplified by a parabolic trajectory in Exp3, which initially declines and then rises. This trajectory reflects the population’s herd mentality under the strong influence of social learning. In the early stages of the sustainable development process, average environmental awareness decreases, primarily due to the prevalent use of high-carbon energy. As a result, the social learning effect has a negative impact. However, as the market share of low-carbon energy expands significantly in the later stages, average environmental awareness begins to rise, accompanied by a positive social learning effect. This phenomenon highlights the dual impact of social learning on demanders’ environmental awareness.

3.2.2. Social Learning’s on Sustainable Development

Figure 8 presents the findings related to the market penetration of low-carbon energy. Across the various scenarios with different levels of social learning effects, the development trajectory of the low-carbon energy market follows a similar pattern. Initially, up to 2030, low-carbon energy faces a noticeable cost disparity compared to high-carbon energy. During this period, demanders’ environmental awareness has limited influence, leading to similar trends in energy choices across different scenarios. However, a shift occurs after 2030, as the cost of low-carbon energy declines and demanders’ environmental awareness begins to play a more crucial role.

Exp1, which lacks a social learning effect, demonstrates the fastest growth in the average willingness-to-pay among demanders. As a result, a larger proportion of demanders choose low-carbon energy. In contrast, when social learning begins to influence decisions, the growth rate of the low-carbon energy market share starts to slow. Interestingly, the outcomes of sustainable development in this study differ in some ways from previous research, which highlights the significant role of social and mass media in real-world contexts (Severo et al., 2019 [28]; Kushwaha, 2015 [30]). As outlined in Section 3.2.1, the mechanism of social learning has a dual nature. In scenarios where high-carbon energy sources are dominant, the strong influence of social learning can negatively impact demanders’ environmental awareness, thus affecting the progress of sustainable development initiatives.

These findings have significant implications for policy formulation. The effectiveness of incentives aimed at increasing consumer awareness may be limited when the cost of low-carbon energy remains relatively high. The optimal time for promoting awareness and utilizing social and mass media interventions appears to be around 2030, which coincides with the period when the costs of high-carbon energy are expected to approach parity with those of low-carbon alternatives. In particular, within the context of strong social learning effects, the positive impact of social learning only becomes evident once the market penetration of low-carbon energy reaches a certain threshold. Until that point, energy-focused strategies should prioritize the rapid advancement of technologies and the provision of subsidies to companies committed to reducing the cost of low-carbon energy.

3.3. Initial Heterogeneity

Figure A2 in Appendix A visually illustrates the evolution of willingness-to-pay for demanders under both high and low initial heterogeneity scenarios. Notably, in both cases, the WTP of demanders shows an overall rightward shift as the simulation progresses, indicating increased environmental awareness. Simultaneously, individual heterogeneity also increases. In scenarios with low initial heterogeneity, the distribution of WTP among individuals becomes more concentrated, while in scenarios with high initial heterogeneity the distribution becomes more dispersed.

3.3.1. Initial Heterogeneity’s Impact on Demanders’ Environmental Awareness

Figure 9 shows that increasing initial heterogeneity, which indicates a greater diversity in initial WTP values among demanders, has a limited impact on the overall process of evolving environmental awareness. The growth trajectory of environmental awareness remains relatively steady across all scenarios. Differences in the final WTP values achieved by 2040 are due to a combination of market share dynamics and the interplay of social learning effects.

Furthermore, the simulation reveals a gradual increase in heterogeneity throughout the experiment, as shown in Figure 10. These findings highlight that, even with the presence of social learning effects, environmental awareness does not uniformly converge among demanders without human intervention—a trend also observed in Figure A2 in Appendix A. Individual differences persist, leading to the formation of diverse groups with varying levels of willingness-to-pay.

3.3.2. Initial Heterogeneity’s Impact on Sustainable Development

It is crucial to emphasize that the initial level of heterogeneity significantly impacts the progression of the sustainable development process. As shown in Figure 11, the dynamic evolution of the low-carbon energy market share undergoes a distinct shift in response to variations in initial heterogeneity. Using $\sigma \left(2021\right)=2.6$ as a critical threshold, when heterogeneity is below 2.6, the sustainable development follows a nonlinear trajectory, characterized by a period of deceleration followed by acceleration, reaching a turning point around 2030. Conversely, when heterogeneity exceeds 2.6, the sustainable development process exhibits a deceleration-acceleration-linear pattern, with turning points around 2030 and 2035.

Regarding the ultimate market share of low-carbon energy in 2040, the scenario with the critical threshold of $\sigma \left(2021\right)=2.6$ achieves a market share of approximately 67%, slightly surpassing the 65% observed in the low initial heterogeneity scenario. However, as heterogeneity increases beyond this threshold, the final market share of low-carbon energy begins to decline. In the scenario with the highest level of heterogeneity, the ultimate market share of low-carbon energy drops to just 62%.

The sustainable development trajectory highlights that in scenarios with high initial heterogeneity, the growth rate peaks when WTP becomes significant, typically around 2030. This phenomenon occurs due to the presence of a substantial number of individuals with heightened environmental awareness within the system, as shown in the distribution of demanders’ WTP in Figure A2 (lower graph). However, after 2035, the negative consequences of the high initial heterogeneity scenario become evident, as depicted once again in Figure A2 (lower graph). Due to the pronounced initial heterogeneity, a significant number of demanders with low WTP remain in the market. Despite a moderate increase in the WTP of this group as the simulation progresses (indicated by an overall rightward shift in the distribution), it is not enough to motivate them to choose low-carbon energy sources.

In contrast, low heterogeneity scenarios (Figure A2, upper graph) demonstrate concentrated environmental awareness among demanders within the system. The proportion of individuals with either very high or very low WTP is not significant, leading to a relatively lower market share of low-carbon energy around 2030. However, as time progresses and individual environmental awareness increases (as shown by the overall rightward shift in the distribution), individuals with more focused sensitivity collectively adjust their energy choices. This results in accelerated growth of the low-carbon energy market share after 2035, eventually surpassing the high-heterogeneity scenario post-2035. Consequently, the overall sustainable development process exhibits a nonlinear progression, characterized by an initial deceleration followed by acceleration.

While the initial heterogeneity of demanders’ environmental awareness does not significantly impact the actual growth of environmental awareness, it has a profound influence on the trajectory of sustainable development. A market characterized by high heterogeneity in environmental awareness among demanders might facilitate sustainability promotion during the initial stage but can lead to challenges in later phases. Conversely, in scenarios with low heterogeneity, the pace of sustainability promotion might be slower in the early stages, but it becomes more effective in the later stages. Therefore, an optimal sustainable development strategy involves maintaining an appropriate level of heterogeneity among demanders. This approach ensures that individuals with high environmental awareness adopt low-carbon energy early on, while behavioral changes among those with lower environmental awareness are driven by the effects of social learning. This strategy addresses both the early and later stages of sustainable development. The implications of these findings for policy are significant: reducing income disparities between regions to mitigate environmental awareness heterogeneity is crucial for promoting sustainable development. This can be achieved by implementing equitable economic policies focused on improving income distribution, along with providing education and training opportunities.

4. Conclusions

The impact of individual environmental awareness on sustainable development is a critical consideration for energy system managers, particularly in the context of an increasingly decentralized landscape. This factor, combined with the dynamics of social learning and the variability in individual behaviors, presents a complex challenge. Despite the importance of this issue, there is a significant gap in existing research regarding effective simulation tools that accurately assess the influence of environmental awareness on decision-making processes in the electricity sector from a holistic system perspective. To address this gap, our study develops an agent-based model and conducts simulation experiments focused on the Japanese electricity market. We quantify demanders’ environmental awareness using the willingness-to-pay (WTP) parameter, integrating actual statistical data to enhance the realism of our simulations. Our analysis explores the nuanced interaction between environmental awareness and sustainable development, considering the effects of social learning and diversity among consumers.

Firstly, within the sustainable development framework of the electricity sector, the introduction of social learning creates a clustering effect. In scenarios characterized by strong social learning, numerous clusters of demanders emerge, utilizing both high-carbon and low-carbon energy sources. Social learning has a dual effect on demanders’ environmental awareness. Within low-carbon clusters, the environmental awareness of demanders accelerates and influences the awareness of peers in neighboring clusters, fostering gradual expansion. Conversely, within high-carbon clusters, demanders’ environmental awareness diminishes, affecting their surroundings. Under strong social learning effects, a system predominantly reliant on high-carbon energy experiences a decline in average environmental awareness, leading to slower sustainable development (negative social learning effect). However, when the proportion of low-carbon energy demanders reaches a critical threshold, the social learning effect shifts to a positive influence. The average environmental awareness of demanders increases, accelerating the pace of sustainable development. This insight offers valuable guidance for real-world sustainable development, highlighting the importance of selecting the optimal timing for promoting public awareness of environmental preservation, which is closely linked to the market size of the low-carbon energy sector at each stage.

Secondly, the initial heterogeneity of demanders’ environmental awareness does not significantly impact the overall growth trajectory of average environmental awareness; however, it is amplified by the clustering effect driven by strong social learning. Despite this, initial heterogeneity does play a notable role in sustainable development. The initial diversity in demanders’ environmental awareness shapes the path of sustainable development. Leading up to 2030, varying levels of initial heterogeneity result in similar developmental patterns. Beyond 2030, differences emerge, with high-heterogeneity scenarios initially gaining momentum and securing a substantial market share for low-carbon energy. However, around 2035, low-heterogeneity scenarios begin to accelerate, eventually surpassing the progress rate of high-heterogeneity scenarios. By 2040, low-heterogeneity scenarios ultimately achieve a greater market share of low-carbon energy than their high-heterogeneity counterparts. Therefore, the optimal sustainable development strategy involves maintaining an appropriate level of heterogeneity among demanders. This strategy ensures that individuals with higher environmental awareness adopt low-carbon energy early on, while behavioral changes among those with lower environmental awareness are driven by the effects of social learning. This approach considers both early and later stages of sustainable development.