📈 economics

Consumer Perception and Policy Intervention in Large-scale Implementation of Advanced Autonomous Driving: A Dynamic Evolutionary Study Based on Agent-Based Modeling

This study employs agent-based modeling to simulate the dynamic evolution of consumer trust and acceptance of advanced autonomous driving under various technological, social, and policy factors, offering data-driven insights for optimizing product strategies and regulatory interventions.

Original authors: hao zhang, Peifeng Zhu

Published 2026-06-29

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: hao zhang, Peifeng Zhu

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: The "Trust Gap" in Self-Driving Cars

Imagine the world of self-driving cars as a giant, high-tech playground. The engineers (the builders) have been working hard to build the best slides and swings (the technology). They have even gotten the government (the park managers) to give them a green light to open the park to the public.

However, there is a problem: The parents (the consumers) are still too scared to let their kids play.

Even though the rides are getting safer and the rules are clearer, people are hesitant. They worry about accidents, they don't fully understand how the cars work, and they are waiting to see if it's truly safe. This paper tries to figure out how to get those parents to finally say, "Yes, let's go!"

The Problem with Old Ways of Thinking

Before this study, researchers tried to understand this fear by asking people questions at a single moment in time (like taking a snapshot). They asked, "Do you trust self-driving cars?" and got an answer.

But the authors say this is like trying to understand a movie by looking at just one frame. It misses the action! Trust isn't a static thing; it's a moving story. It changes when a new car model launches, when a news story about an accident breaks, or when the government passes a new law.

The Solution: A "Digital Sandbox"

To solve this, the researchers built a Digital Sandbox (called an Agent-Based Model).

Think of this sandbox as a giant video game simulation where they created 500 virtual people (agents). These aren't just random numbers; they were programmed with real human traits based on a survey the researchers conducted.

Some virtual people are naturally brave.
Some are very cautious.
Some listen to their friends; others make their own decisions.

The researchers then ran this simulation forward in time, watching how these 500 people reacted to three main things:

The Tech: The cars getting better (moving from Level 2 "helper" to Level 3 "driver" to Level 4 "fully autonomous").
The Bad News: Random safety accidents happening in the simulation.
The Rules: Different government policies (stricter laws, more advertising, or money subsidies).

The Three Forces at Play

The study found that consumer trust is driven by three forces, like a three-legged stool:

The Tech Ladder: The technology doesn't improve smoothly; it jumps. Imagine a ladder where you stand on one rung for a while, then suddenly jump to the next. The study found that these "jumps" (like the government officially saying "Level 3 is legal now") act like a shockwave that changes how people feel all at once.
The Social Circle: People talk. If your neighbor says, "My car drove itself perfectly," you feel safer. If your neighbor says, "It almost hit a pedestrian," you get scared. The simulation showed that this "word-of-mouth" spreads trust (or fear) very quickly through the group.
The Personal History: How you feel today depends on what you've done before. If you've used a basic "lane-keeping" feature for years, you might trust a new self-driving car more than someone who has never used one.

The Experiment: What Works Best?

The researchers tested different "recipes" for government policy to see which one would make the most people accept the technology. They tried mixing three ingredients:

Strict Regulation: Making sure the rules are tough and safety is guaranteed.
Communication: Telling people about the tech and how safe it is.
Subsidies: Giving people money to buy the cars.

The Results:

Recipe A (No Rules, Just Hype & Cash): This got people interested at first, but when a fake accident happened in the simulation, everyone panicked and stopped buying. It was too fragile.
Recipe B (Strict Rules, No Talking): This was safe, but boring. People didn't buy the cars because they didn't know enough about them.
Recipe C (The Winning Combo): Strict Rules + Moderate Talking + Moderate Cash.

The "Goldilocks" Strategy:
The study found that the best approach is to have strict government rules to build a solid foundation of safety (like a strong fence around the playground). Once that safety is guaranteed, moderate communication helps spread the word quickly, and moderate subsidies help people take the first step.

This combination created a "tipping point" where trust grew steadily, even when accidents happened. The strict rules acted as a shock absorber, preventing trust from crashing completely when bad news hit.

The Key Takeaway

The paper concludes that you cannot just hype up self-driving cars or just give them away for free. You need a safety net first.

Think of it like teaching a child to ride a bike:

You don't just take off the training wheels and hope for the best (No Regulation).
You don't just keep the training wheels on forever and never let them ride (Strict Regulation without Communication).
The Best Way: You put on a helmet and use a stabilizer (Strict Regulation) so they feel safe. Then, you cheer them on (Communication) and maybe give them a little push (Subsidy) until they find their balance and start riding on their own.

The study proves that for self-driving cars to become a normal part of our lives, the government must first prove the technology is safe through strict rules. Once that trust is built, the rest of the adoption happens much faster.

Technical Summary: Consumer Perception and Policy Intervention in Large-scale Implementation of Advanced Autonomous Driving

Problem Statement
Despite rapid technological advancements and robust policy support in China's autonomous driving industry, consumer perception remains cautious, creating a "cognitive bottleneck" that hinders the transition from pilot demonstrations to mass commercial application. Existing literature often treats autonomous driving as a homogeneous technology, focusing on idealized Level 5 (L5) scenarios while neglecting the critical transitional phases (L2 to L4) where human-machine role boundaries shift. Furthermore, traditional research relies heavily on cross-sectional surveys and static models (e.g., TAM, TPB), which fail to capture the non-linear feedback loops, dynamic trust evolution, and emergent macro-level patterns triggered by safety accidents, technological milestones, and policy interventions. The core problem addressed is the lack of a forward-looking, dynamic framework to understand how consumer trust and acceptance evolve under the complex interplay of technological progression, public opinion shocks, and policy strategies.

Methodology
This study employs an Agent-Based Modeling (ABM) approach to construct a "computational laboratory" that simulates the dynamic evolution of consumer cognition. The methodology proceeds in three stages:

Empirical Calibration: A questionnaire survey ( $N=184$ ) based on the Technology Acceptance Model (TAM) was conducted to calibrate initial parameters. Maximum Likelihood Estimation (MLE) confirmed that initial consumer trust ( $\mu=0.38$ ) and risk perception ( $\mu=0.62$ ) follow normal distributions, reflecting individual heterogeneity.
Model Construction: A discrete-time simulation (500 ticks, where 1 tick = 1 month) was built using the Python Mesa framework. The model integrates three driving forces:
- Technological Force: A stepped evolution mechanism where technology levels (L2+ $\to$ L3 $\to$ L4) jump at specific policy milestones (e.g., L3 market access at tick 121), acting as exogenous trust signals.
- Social Force: Consumers are embedded in a Watts-Strogatz small-world network ( $N=500$ ). Agents update their trust based on peer effects (social influence) and word-of-mouth diffusion.
- Individual Force: Agents possess heterogeneous psychological traits (risk aversion, social sensitivity) and prior experiences (L2 usage).
Policy Simulation: The model tests the impact of three policy instruments—Regulatory Intensity ($reg$), Communication Intensity ($comm$), and Subsidy Intensity ($sub$)—ranging from 0 to 1. Safety accidents are modeled as a Bernoulli process ( $p=0.01$ ) to simulate negative shocks. The study runs full factorial experiments to identify optimal policy combinations under budget constraints.

Key Results
The simulation reveals distinct evolutionary trajectories based on policy combinations and technological progression modes:

Optimal Policy Portfolio: Under a fixed budget, the combination of strict regulation ($reg=1.0$), moderate communication ($comm=0.5$), and moderate subsidies ($sub=0.5$) yields the highest comprehensive score. This strategy achieves a final adoption rate near 100% with a rapid diffusion speed (crossing the 50% tipping point at tick 166).
Role of Regulation: Strict regulation is identified as the foundational determinant of consumer trust. Without it, trust remains trapped in a low range (0.4–0.5) and is highly vulnerable to accident shocks. Regulation establishes a "safety bottom line" that stabilizes the trust baseline.
Role of Communication and Subsidies: Communication accelerates the diffusion of positive cognition but has limited marginal returns without the safety guarantees provided by regulation. Subsidies effectively lower economic barriers for early adopters but cannot compensate for a lack of institutional trust.
Technological Evolution Mode: A stepped evolution (aligned with policy milestones like L3 access) significantly outperforms a linear evolution. The "cognitive leap" triggered by clear regulatory milestones acts as a strong positive signal, breaking cognitive stagnation more effectively than continuous, incremental technical improvements.
Resilience: The optimal policy portfolio demonstrates high resilience; while safety accidents cause temporary trust dips, the combination of regulation and communication allows for rapid recovery, unlike scenarios relying solely on subsidies or publicity.

Significance and Contributions
The paper claims to offer both theoretical and practical contributions to the field of autonomous driving adoption:

Theoretical Advancement: The study moves beyond static acceptance models by integrating cognitive transition, trust repair, and social influence theories into a unified dynamic evolutionary framework. It explicitly addresses the "path-dependent" nature of trust, demonstrating that trust established at lower automation levels (L2) does not automatically transfer to higher levels (L3/L4) without recalibration.
Methodological Innovation: By introducing ABM as a "computational laboratory," the study provides a novel tool for deriving macro-emergent patterns from micro-level interactions. This approach overcomes the limitations of cross-sectional data in capturing non-linear feedback and emergent phenomena in complex socio-technical systems.
Policy Implications: The findings provide actionable guidance for policymakers and manufacturers. Specifically, the study argues that regulatory clarity is the prerequisite for effective communication. Policies should prioritize establishing a robust safety regulatory framework first, followed by targeted communication and subsidies to accelerate market penetration. This "balanced" approach is presented as the most efficient path to overcoming the cognitive bottleneck and achieving sustainable large-scale implementation.

The authors conclude that their simulation model serves as a "digital sandbox" for evaluating long-term policy effects, offering a low-cost, forward-looking platform for decision-makers to navigate the complex transition from technological feasibility to market acceptance.