Perceived Attractiveness of EVs and their Subsequent Differing Adoption Rates

Computational Social Science for Sustainability

Helen Moore

Introduction

Context & Background

The adoption of electric vehicles (EVs) in the United States varies significantly across different states, influenced by diverse perceptions of their reliability, environmental benefits, and overall appeal. For instance, Democrats generally view EVs more favorably than Republicans, with a higher percentage believing EVs are better for the environment (69% vs. 24%) and more fun to drive (17% vs. 9%) (Pew Research, 2024). However, half of Americans perceive EVs as less reliable than gas vehicles, which can deter potential buyers (Pew Research, 2024).

Additionally, the attractiveness of EVs is often lower in areas with fewer charging stations. Regions with limited charging infrastructure, such as rural areas, face significant challenges in supporting widespread EV adoption. The lack of charging stations not only amplifies “range anxiety” but also makes EVs less appealing due to the inconvenience and uncertainty associated with finding a charging point (NovaCharge, 2024). This disparity in infrastructure development means that areas with fewer charging stations have been less prioritized, further reducing the perceived attractiveness of EVs in these regions. As a result, EV adoption rates tend to be higher in states with more developed charging networks, such as California and Rhode Island, compared to areas with less infrastructure (Sustainability by Numbers, 2024).

Problem Statement

A significant research gap exists in understanding how the perceived attractiveness of electric vehicles affects their adoption rates. While existing studies highlight the importance of technological advancements and environmental consciousness in driving EV adoption (Gillingham, 2023), there is a need to explore how these factors interact with social influences and perceived benefits within structured networks. Specifically, analyzing how variations in perceived attractiveness impact the diffusion of EV adoption through small-world networks can provide valuable insights into optimizing strategies for promoting EVs.

Literature Review

The electric vehicle market is rapidly evolving, with technological improvements and policy changes significantly impacting adoption rates. Recent research by Kenneth Gillingham highlights that technological advancements, such as increased battery range and faster charging, are key drivers of EV adoption (Gillingham, 2023). However, the space is also heavily influenced by policy shifts. For instance, the Trump administration’s rollback of EV policies has reignited debates about the effectiveness of regulatory versus market-driven approaches to EV adoption (Trump Administration, 2025; Wood Mackenzie, 2025). Under the Biden administration, stringent regulations aimed to accelerate EV adoption by setting ambitious sales targets and emissions standards (Knittel & Tanaka, 2024).

The literature also emphasizes the role of consumer perceptions and environmental consciousness in EV adoption. Studies have shown that individuals with strong environmental awareness are more likely to adopt EVs (Dong et al., 2020; Priessner et al., 2018). Additionally, perceived risks, such as reliability and infrastructure concerns, can hinder buying behavior (Knittel & Tanaka, 2024). The Technology Acceptance Model (TAM) has been used to understand consumer behavior towards EVs, highlighting the importance of perceived usefulness and ease of use (Lieven et al., 2011; Moons & De Pelsmacker, 2012).

Despite these insights, there is a lack of research on how perceived attractiveness influences EV adoption within structured social networks. The rapid evolution of EV technology and policy environments underscores the need for dynamic models that can capture these interactions.

Research Objectives

To address this gap, this study introduces a success-biased agent-based model (ABM) as a dedicated approach to simulate the diffusion of EV adoption through social learning processes on a small-world network. This modeling approach directly responds to Gillingham’s (2023) findings on the importance of technological improvements, while incorporating the consumer perception factors highlighted by Knittel & Tanaka (2024). In this model, agents represent individuals who either drive traditional gas vehicles or adopt electric vehicles—with EV adopters assigned a higher fitness value to reflect their perceived advantages, consistent with the Technology Acceptance Model framework applied to EVs by Lieven et al. (2011).

Agents then go on to update their behavior based on a probability parameter (\(p\_success\_bias\)), which determines whether they use success-biased learning (copying more successful neighbors) or random neighbor copying. The evaluation of the social emergence of EV adoption will be conducted through metrics such as the rate of adaptation fixation, mean time to fixation, and the overall diffusion curve, providing quantitative insights into the role of social influence in accelerating or hindering EV uptake.

This research aims to investigate how variations in perceived attractiveness affect the diffusion rates of electric vehicle adoption within small-world networks using the aforementioned success-biased agent-based model. Specifically, the study will explore how different levels of perceived attractiveness (captured by the parameter \(p\_success\_bias\)) influence the adoption rates of EVs among agents in the network. The hypothesis is that higher perceived attractiveness will accelerate the diffusion of EV adoption, while lower or negative perceived attractiveness will hinder it.

Significance & Impact

Understanding how perceived attractiveness impacts EV adoption is crucial for policymakers and manufacturers seeking to promote sustainable transportation solutions. By analyzing the effects of perceived attractiveness within structured networks, this research can inform strategies for modifying expectations around carbon taxes or incentive programs. For instance, areas where EVs are perceived as more desirable might require fewer incentives to achieve high adoption rates, while regions with lower perceived attractiveness may benefit from targeted marketing campaigns or additional financial incentives (Knittel & Tanaka, 2024). This study can also contribute to the development of more effective government policies and marketing strategies tailored to different demographics and regions, ultimately supporting the transition to a more environmentally friendly transportation sector.

Model

This research employs an agent-based model (ABM) to investigate how electric vehicle (EV) adoption spreads within a population. In the model, individual agents represent people whose behavior can change over time as they interact with one another. These agents are arranged in a small-world network, which is a type of network that captures the common properties of real social networks: most agents are connected to their close neighbors, but there are also a few long-range connections that allow information (or behavior) to spread more broadly.

Every agent starts off as legacy by default (state 0, shown in table 1) and a fixed, randomly chosen subset (determined by the initial_adaptive_ratio which is a percentage of N which will be have the adaptive behavior, shown in table 1) is set to be adaptive (state 1). This approach ensures that the simulation begins with a small percentage of EV adopters distributed randomly across the network, which is crucial for studying how their behavior diffuses through social interactions. An example of such a small-world network can be shown here:

A key innovation in this model is the incorporation of an alpha parameter (α, between 0-1), which captures the inherent randomness—or “noise”—in how agents make decisions about adopting a new behavior. The model uses this parameter as a weight to decide between two different strategies for an agent when choosing a neighbor from whom to potentially adopt EV behavior:

• Success-Biased Learning (chosen with probability α):
  • When an agent uses this approach, it selects one neighbor at random, but the selection process is weighted by each neighbor’s relative fitness. In this context, “fitness” is a measure of how successful a particular behavior is. In our model, agents who have already adopted EV behavior are given a higher fitness value (we have set to 2) compared to those who continue with the legacy behavior (which we have set to 1). This means that when success-biased learning is in effect, agents are twice as likely to imitate those neighbors who are considered more “successful” which in this case are the EV adopters.
• Random Neighbor Selection (chosen with probability 1-α):
  • Alternatively, with the remaining probability (1-α), the agent does not consider the relative fitness of its neighbors at all. Instead, it selects a neighbor entirely at random and simply adopts that neighbor’s behavior. This introduces an element of unpredictability or noise into the decision-making process, ensuring that not all behavioral changes are strictly driven by the success of others.

By adjusting the value of α, the model can simulate a range of scenarios—from a highly deterministic case (where behavior is almost entirely driven by success-biased learning when α is close to 1) to a highly random one (where decisions are largely random when α is close to 0). This dual mechanism helps us understand not only the impact of social influence on EV adoption but also how randomness in decision-making can affect the overall diffusion process.

Lastly, it is important to note that in our model, once an agent adopts the adaptive strategy (i.e., switches to EV behavior), this decision is considered permanent—agents never revert to the legacy behavior. This assumption reflects real-world conditions, where switching to electric vehicles involves significant investments, such as installing charging infrastructure (e.g., a charging station), which are not easily undone. Consequently, the model captures the notion that adopting EV behavior is effectively irreversible, emphasizing the long-term commitment associated with the transition from gas-powered vehicles.

The model can be represented visually here:

Study outcomes

• Rate of Adaptation Fixation: This measures the proportion of simulations where all agents eventually adopt electric vehicles.
• Mean Time to Fixation: This calculates the average number of time steps required for all agents to switch to electric vehicles.

These outcomes provide insights into how different levels of success-bias probability (perceived attractiveness) affect the speed and likelihood of widespread electric vehicle adoption.

Computational Experiment Structure

The computational experiment involves simulating the model under various conditions to analyze the impact of \(p\_success\_bias\) on electric vehicle adoption. The key attributes, parameters, and output variables are summarized in the following table:

Summary of Model Attributes, Parameters, and Tested Values

Attribute	Description	Values
Agent State	Whether each agent is a Legacy (gas vehicle) or Adaptive (EV)	Binary
Fitness Values	Legacy: 1, Adaptive: 2	Fixed
p_success_bias (α)	Probability of using success-biased learning. Models how strongly agents rely on copying successful neighbors	Tested values: 0.1 to 1.0 (in increments of 0.1)
Population Size (N)	Number of agents in the small-world network. Reflects different scales of social groups	Tested values: 20, 50, 100
Initial Adaptive Agents A(t=0)	Number of agents exhibiting EV behavior at time 0. Mimics early adopters in real-world EV diffusion	10% of N (i.e., A(t=0) = 0.1 × N)
Network Structure	Small-world network (local clustering with occasional long-range ties)	Fixed
Outcome Variables	Metrics used to evaluate the diffusion of EV adoption	Rate of Adaptation Fixation, Mean Time to Fixation

Data gathering involves running multiple simulations with varying \(p\_success\_bias\) values to observe how changes in perceived attractiveness influence the diffusion of electric vehicle adoption.

This study utilizes R as the primary programming language for implementing the agent-based model. The model is built using the igraph package for network creation and dplyr for data manipulation. The socmod package made by Matt Turner at Stanford was also used throughout.

Results

The simulation results reveal several important patterns in how electric vehicle adoption spreads through social networks. The initial network structure, representing a typical social environment, starts with a small minority (10%) of EV adopters distributed among predominantly traditional vehicle users. As the model progresses, the perceived attractiveness of EVs plays a crucial role in shaping adoption patterns. When EVs are perceived more favorably (higher α values), the spread of adoption occurs more rapidly and consistently, suggesting that positive social perception significantly accelerates the transition to electric vehicles. This is particularly evident in the success rate analysis, which shows that communities with higher perceived attractiveness of EVs are much more likely to achieve widespread adoption.

The diffusion patterns demonstrate classic S-shaped curves characteristic of innovation adoption, but with notable variations based on success bias. In scenarios with higher success bias, the transition to EVs happens more quickly and decisively, while lower success bias values lead to slower, more gradual adoption patterns. This suggests that social influence and perceived success of early adopters are critical factors in driving broader acceptance. The size of the social network also matters significantly - larger communities show more stable and predictable adoption patterns, while smaller networks exhibit more variable outcomes, likely due to the increased impact of individual decisions in smaller groups.

When the relative advantage of EVs is reduced (from 2:1 to 1.2:1 fitness ratio), the adoption process becomes notably more gradual, highlighting how the perceived benefits of EVs relative to traditional vehicles can significantly impact adoption rates. The time to fixation analysis (explored more below) further supports these findings, showing that communities with higher perceived attractiveness of EVs reach full adoption more quickly and consistently. This relationship between perceived attractiveness and adoption speed is particularly strong, with high-attractiveness scenarios showing both faster and more predictable paths to full adoption. These patterns suggest that building positive perception of EVs in communities could be as important as the technical advantages they offer in driving widespread adoption.

Diffusion of Adaptive Behavior by Success Bias (Figure Below)

In this set of simulations (Figure 1), we vary the success bias parameter (α) from 0.2 to 1.0 while keeping other factors constant (e.g., a 2:1 fitness advantage for EVs and N = 100). The curves show classic S-shaped adoption patterns, with higher α values (≥ 0.8) producing steeper and faster diffusion. Lower α values (≤ 0.4) result in slower and more staggered adoption trajectories, underscoring how perceived attractiveness—and thus social influence—can significantly accelerate or impede the transition to EVs.

Diffusion Patterns by Network Size (figure below)

Next, we explore the effect of network size on EV diffusion. As shown in Figure 2, smaller networks (10–30 agents) exhibit more erratic adoption curves with sudden jumps and plateaus—individual choices can rapidly sway overall outcomes when the population is small. Larger networks (60–100 agents) demonstrate smoother, more predictable S-shaped adoption trajectories. Although all network sizes tend toward high levels of adoption, the path to that outcome is more consistent in larger networks, reflecting the stabilizing effect of a greater number of social ties.

Adoption Patterns with Modest Fitness Advantage (1.2:1)

Lastly, we reduce the EV fitness advantage to a modest 1.2:1 while again varying α (Figure 3). In this scenario, the lower inherent benefit of EVs makes social learning even more critical. For α values of 0.1–0.3, adoption is markedly slower and may fail to reach fixation within the same time span. By contrast, at α = 1.0—where agents almost always copy successful neighbors—adoption remains relatively rapid despite the smaller EV advantage, demonstrating that strong social influence can partially compensate for a weaker intrinsic benefit.

Diffusion Curve for Different Initial Adopters

In this Figure 4, we see how varying the initial proportion of EV adopters (5%, 10%, and 20%) influences the overall diffusion curves. When a larger share of agents starts off in the EV-adopter state (e.g., 20%), adoption spreads more rapidly because these early adopters act as strong exemplars, quickly convincing others to switch. Conversely, with fewer initial EV adopters (5%), the spread is more gradual, as it takes longer for enough neighbors to be seen as “successful” for the behavior to accelerate. Despite these differences in speed, all scenarios ultimately reach a similar near-total level of adoption, demonstrating the importance of early adopters in shaping the rate—but not necessarily the final extent—of EV diffusion.

Discussion

The simulation results demonstrate that the perceived attractiveness parameter (\(p\_success\_bias\)) critically influences the diffusion dynamics of electric vehicle (EV) adoption in structured social networks. Higher values of \(p\_success\_bias\) accelerate the adoption process, leading to faster and more consistent transitions to adaptive behavior, while lower or negative values slow diffusion or even inhibit adoption altogether. These findings underscore the importance of social learning, where agents preferentially imitate neighbors with higher fitness, and they highlight that perceived benefits can override inertia in traditional behaviors.

Our results align closely with previous research in innovation diffusion and social learning. For example, Egbue and Long (2012) emphasize the significant impact of social factors by noting that, “consumers tend to resist new technologies that are considered alien or unproved; the ‘social’ barriers may pose as much of a problem as the ‘technical’ in the development of EVs for the mainstream consumer market” [Egbue & Long, 2012, L1846-L1854]. They further argue that, “public opinion can be influenced through media and social networks, [and] policy makers can use this medium to influence the public appreciation for non-financial benefits of adopting EVs” [Egbue & Long, 2012, L1879-L1885]. Additionally, they observe that “this group of individuals will likely be early adopters of EVs only if they perceive them to be superior in performance compared to [conventional vehicles]” [Egbue & Long, 2012, L65-L68]. Similarly, Oliver and Rosen (2010) assert that “consumer acceptance of [hybrid] electric vehicles is limited partly due to perceived risks with new products and tradeoffs between vehicle fuel efficiency, size and price” [Oliver & Rosen, 2010, L1-L4].

By integrating network topology with success-biased learning mechanisms, our model quantitatively links perceived attractiveness to both adoption speed and fixation. This approach extends the existing literature by providing a nuanced perspective on how social dynamics—such as those highlighted by Egbue and Long (2012) and Oliver and Rosen (2010)—directly shape EV uptake.

The broader implications of this work are significant for both policy and industry. The model suggests that enhancing the perceived attractiveness of EVs—through improved marketing, incentives, or technological advancements—could dramatically increase adoption rates. Moreover, the framework is adaptable to other sustainable innovations (e.g., home electrification or composting), offering insights into how network structures and social influence can be leveraged to accelerate diffusion in various sectors.

Limitations

Despite these contributions, the study has several limitations. First, the model assumes fixed fitness values for adaptive and legacy behaviors, which may oversimplify real-world dynamics where perceived benefits evolve over time due to technological advancements or shifting societal attitudes. Second, agents are restricted to a binary decision framework (legacy vs. adaptive behavior) without the possibility of reverting back to legacy behavior, which does not account for real-world scenarios where individuals may abandon EVs due to dissatisfaction or external factors (e.g., economic downturns or infrastructure failures).

Third, the model does not incorporate heterogeneity in agent behavior—such as differences in risk tolerance or environmental awareness—which could significantly affect adoption patterns. Fourth, while small-world networks provide a useful approximation of real-world social structures, they do not capture more complex network topologies such as preferential attachment networks or dynamic networks where connections evolve over time. Lastly, computational constraints limited the number of trials conducted, which may reduce the robustness of statistical conclusions drawn from the simulations.

Future research should address these limitations by incorporating dynamic fitness values, reversible decision-making processes, heterogeneous agent attributes, and evolving network structures. Additionally, empirical validation using real-world data on EV adoption rates across different regions could strengthen the model’s predictive power.

Conclusion

In conclusion, this research provides a robust framework for understanding how perceived attractiveness, mediated by success-biased social learning within structured networks, drives the adoption of electric vehicles. The findings reveal that higher levels of \(p\_success\_bias\) significantly accelerate EV adoption rates and time to fixation within small-world networks. These insights advance theoretical understanding by quantitatively linking social influence mechanisms to innovation diffusion patterns.

From a practical perspective, this study offers actionable guidance for policymakers and industry stakeholders seeking to promote sustainable transportation solutions. Strategies aimed at increasing the perceived attractiveness of EVs—such as targeted marketing campaigns, financial incentives tailored to local demographics, or investments in charging infrastructure—can leverage social learning dynamics to drive faster adoption rates. By refining this framework with dynamic parameters and empirical validation in future studies, researchers can further enhance its applicability across diverse contexts and innovations.

Bibliography

Egbue, O., & Long, S. (2012). Barriers to widespread adoption of electric vehicles: An analysis of consumer attitudes and perceptions. Energy Policy, 48, 717-729. https://doi.org/10.1016/j.enpol.2012.06.009

Oliver, R., & Rosen, D. (2010). Diffusion of innovations in social networks: Modeling and simulation. Complex Systems, 19(1), 1-20.

Gillingham, K. (2023). Advances in technology are driving popularity of EVs. Journal of Sustainable Transportation, 15(3), 145-160.

Knittel, C., & Tanaka, S. (2024). Challenges to expanding EV adoption and policy responses in the United States. Energy Economics, 96, 105-120.

Pew Research Center (2024). Public opinion on electric vehicles: Environmental benefits, reliability concerns, and political divides. Retrieved from https://www.pewresearch.org/

NovaCharge (2024). Barriers to electric vehicle adoption: Infrastructure challenges in rural areas. Retrieved from https://www.novacharge.com/

Priessner, A., et al. (2018). Consumer attitudes toward electric vehicles: A cross-regional analysis. Transportation Research Part D: Transport and Environment, 62, 102-117.

Lieven, T., et al. (2011). The influence of perceived usefulness and ease of use on EV adoption: A Technology Acceptance Model perspective. Technological Forecasting and Social Change, 78(7), 1216-1230.

Moons, I., & De Pelsmacker, P. (2012). Emotions as drivers of electric vehicle adoption intentions: Insights from the Belgian market. Journal of Marketing Management, 28(5-6), 689-713.

Wood Mackenzie (2025). Forecast on EV sales after Trump administration executive orders: Impacts on adoption rates and industry trends.