Ⅰ. Introduction

1. Research Background and Objectives

While Autonomous Vehicles (AVs) promise a paradigm shift in road safety and economic efficiency, the prevailing academic discourse remains disproportionately anchored in the “car-centric” transitions of industrialized nations with mature infrastructure (Hu et al., 2026;Huang et al., 2026b;Jaydarifard et al., 2025;Kim et al., 2025;Nazari et al., 2026;Wang et al., 2025, 2026;Zhu and Hsieh, 2026). This geographic bias leaves a critical knowledge gap regarding adoption dynamics in the Global South, particularly within “Heterogeneous Disordered Traffic” (HDT) environments characterized by non-lane-based maneuvering and mixed vehicle types (Fulwadiya et al., 2025;Gu and Liu, 2026;Kenesei et al., 2025;Ullah et al., 2025). In these chaotic contexts, the Western linear model of vehicle substitution fails to explain local behaviors. Instead, for the motorcycle-dependent majority in countries like Indonesia, AVs potentially function not as a luxury convenience, but as a crucial “Safety Upgrade”-facilitating a “Modal Leapfrog” where vulnerable users bypass private car ownership to directly adopt autonomous mobility for physical protection.

Specifically, this study targets the “Urban Early Adopter” segment to decode the “Education Paradox,” where high literacy fuels both technical skepticism and adoption intent based on calculated risk. To ensure clarity, the scope of this research is defined across three dimensions: Spatially, the study focuses on urban commuters in Indonesia; Temporally, data collection was conducted between September and October 2025; and Content-wise, it is delimited to Autonomous Vehicle (AV) adoption determinants. The survey primarily targets three strategic groups: academic experts via the Eastern Asia Society for Transportation Studies (EASTS) Conference Network, the international diaspora through Indonesian Scholars networks, and urban commuters via local community networks. Recognizing that standard comfort-based incentives fail to address this “Informed Skepticism,” we employ a multi-stage “Psycho-Statistical” framework: (1) factor identification via Principal Component Analysis (PCA), (2) predictive modeling using Bootstrapped Ordinal Regression, and (3) strategic segmentation through CHAID trees to map these complex decision pathways. The ultimate objective is to formulate a data-driven Public Policy Roadmap, providing a staged strategy to transition the populace from skepticism toward active adoption by prioritizing physical safety and legal certainty over mere financial perks.

1. Research Trends

The global academic discourse on Autonomous Vehicles (AVs) currently exhibits a sharp geographic dichotomy. Literature originating from the “Global North” predominantly frames AVs as instruments for logistical efficiency, lifestyle comfort, and the reduction of driver fatigue within structured infrastructure. In these car-centric environments, adoption barriers are typically viewed through the lens of cost or technical usability, often modeled using linear frameworks like TAM (Technology Acceptance Model) or UTAUT (Unified Theory of Acceptance and Use of Technology), which assume a rational trade-off between convenience and effort (Feng and He, 2026;Gu and Liu, 2026;Huang et al., 2026a;Huang et al., 2026b;Li et al., 2025;Liang et al., 2025).

However, this “efficiency-first” paradigm falls short when applied to the “Global South,” particularly in regions characterized by Heterogeneous Disordered Traffic (HDT). As defined by Fulwadiya et al.(2025), HDT environments are marked by aggressive, non-lane-based maneuvers and a chaotic mix of vehicle types. In these contexts, recent studies in India and Vietnam suggest that the primary motivation for AV adoption shifts fundamentally from “automation” to “physical survival” (Chakraborty et al., 2025;Devi et al., 2025;Fulwadiya et al., 2025;Hung and Tin, 2025). For vulnerable road users, such as motorcyclists, AVs are not viewed as a luxury, but as a potential “Safety Upgrade”-a mechanism to escape the physical perils of two-wheeled commuting. Research indicates that in these high-risk environments, safety acts not merely as a standard variable, but as the primary determinant for adoption (Feng and He, 2026).

Consequently, a critical theoretical gap exists regarding the “Urban Early Adopter” in developing nations. Current models fail to account for the “Informed Skepticism”, where higher technical literacy in stratified societies ironically leads to lower trust in AI’s ability to navigate chaotic roads. To crystallize the theoretical divergence between the prevailing Western narratives and the reality of the Global South, <Table 1> outlines a comparative framework. While existing studies prioritize efficiency within structured environments using standard adoption models, this research shifts the focus toward ‘Safety Universalism’ within chaotic mixed-traffic contexts, necessitating a robust, non-linear methodological approach.

Building upon the comparison in <Table 1>, the originality of this study lies in its focus on the ‘Safety-First Paradox’ within the Indonesian context. While the reviewed literature primarily emphasizes ‘convenience’ and ‘time-saving’ as drivers in structured environments, this research identifies that for urban early adopters in a ‘Heterogeneous Disordered Traffic’ (HDT) setting, technical literacy leads to ‘Informed Skepticism’ rather than immediate trust. Furthermore, this study departs from the homogenous group analysis common in existing research by providing a granular differentiation between motorcycle and private car users. By employing a ‘Psycho-Statistical’ approach that integrates Bootstrapping for small-sample robustness and CHAID for non-linear segmentation, this research provides a culturally nuanced Public Policy Roadmap that addresses the specific safety-seeking behaviors of the Global South.

1. Methodological Framework

To deconstruct the behavioral complexities of AV adoption within the chaotic mixed-traffic environments of the Global South, this study transcends conventional linear adoption models. Instead, we operationalize a Multi-Stage Psycho-Statistical Framework <Fig. 1>. This proprietary workflow integrates strategic purposive sampling with rigorous computational validation-specifically Bootstrapping and Cross-Validation-to ensure the robustness of inferential findings despite the sample size constraints typical of niche psychographic studies.

<Fig. 1>

Multi-Stage Psycho-Statistical Framework

2. Theoretical Framework and Strategic Sampling

Unlike traditional Technology Acceptance Model (TAM) or Unified Theory of Acceptance and Use of Technology (UTAUT) frameworks, which premise adoption on linear, rational trade-offs between perceived usefulness and ease of use in structured environments, this study requires a different theoretical lens. Standard TAM/UTAUT models fail to capture the cognitive dissonance and survival-driven motivations unique to Heterogeneous Disordered Traffic (HDT). Therefore, we explicitly distinguish our approach by operationalizing a Paradox-Based Behavioral Framework. This framework is specifically designed to capture non-linear trade-offs where perceived risks (e.g., informed skepticism) and high adoption intent can coexist simultaneously, driven by the overriding need for physical safety. The research design is anchored in a “Paradox-Based Behavioral Framework,” tailored to diagnose the specific contradictions found in developing nations: (1) “Safety Universalism,” where physical safety acts as a fundamental prerequisite; (2) “Modal Leapfrog,” proposing that motorcyclists will bypass cars to adopt AVs as a safety shelter; and (3) the “Education Paradox,” testing if high literacy correlates with informed skepticism.

Guided by this framework, the study shifted the analytical lens from the general population to the “Urban Early Adopter” segment. We employed a Stratified Purposive Sampling strategy strictly delimiting the frame to educated commuters in Indonesia’s primary economic corridor (Java-Bali). While this stringent filtering resulted in a niche dataset of N=204, the sample possesses high psychographic quality, characterized by high educational attainment (63.8% Bachelor’s/Master’s) and high traffic vulnerability (65.7% motorcycle dependence).

To rigorously validate the statistical adequacy of this sample size, a Post-hoc Power Analysis was conducted using G*Power 3.1 (Erdfelder et al., 2009). The analysis confirmed a Statistical Power (1-β) of 0.86 (assuming a Medium Effect Size at α=0.05). This metric significantly exceeds the standard scientific threshold of 0.80, confirming that the dataset possesses sufficient sensitivity to detect behavioral determinants without committing Type II errors.

3. Stated Preference Survey Development

Indonesia’s present economic and transportation environment offers a distinctive context for the implementation of Autonomous Vehicles (AVs), especially as the country addresses the challenges of the Middle Income Trap. In 2022, Indonesia’s per capita income was US$4,580, sustaining its upper-middle-income classification into 2023. The World Bank(2025) reports that Indonesia achieved a GDP growth of 5.0% in 2025; nonetheless, the transportation sector has substantial structural issues due to its heavy dependence on private mobility. BPS-Statistics Indonesia(2025) reports that the overall number of motorized vehicles in 2024 has reached 166.5 million units, with motorbikes comprising a predominant 83.77% share. This vehicle configuration generates a fragmented and crowded ‘mixed-traffic’ environment, frequently leading to stagnating productivity and elevated logistical expenses. Moreover, the physical infrastructure, consisting of 539,524 km of roadways, is inequitably distributed, with around 28.92% of roads estimated to be in a severely deteriorated state.

This research examines urban early adopters in Indonesia. Data collection occurred from September to October 2025 with a total of 204 participants, focusing exclusively on the determinants of Autonomous Vehicle (AV) adoption. The survey primarily focuses on three key groups: academic specialists through the Eastern Asia Society for Transportation Studies (EASTS) Conference Network, the overseas diaspora via Indonesian scholars’ networks, and urban commuters through local community networks.

The data collection instrument was a structured Stated Preference (SP) survey comprising four logical modules designed to capture the tension between functional demand and psychological barriers:

As detailed in <Table 2>, the respondents represent a specific ‘Urban Early Adopter’ segment, primarily centered in the Java-Bali agglomeration (90.2%). The sample is dominated by the established middle class (54.9% earning Rp 2.5-9 million/month) with high educational attainment, where 63.8% hold at least a Bachelor’s degree. This demographic is critical because it represents a ‘Vulnerable Road User’ profile; despite their middle-class status, there is a heavy reliance on motorcycles (65.7%). Most alarmingly, 52.5% of respondents have experienced traffic accidents, with ‘Motorcycle with Injury’ being the most significant type (23.5%). This high exposure to physical risk explains why ‘Increased Safety’ (50.5%) is identified as a primary desired benefit, significantly outweighing concerns like ‘High Costs’ (42.6%). Consequently, the market profile reflects a ‘Safety-First’ demand, where reliability and experience are prioritized over minimizing travel fares (8.3%).

4. Psycho-Statistical Workflow

To transform raw psychometric data into actionable policy insights, the analysis followed a three-stage “Statistical Fortification” protocol using IBM SPSS Statistics 27.

Phase 1: Diagnostic Validation & Bias Control

To mitigate potential Common Method Bias (CMB) inherent in self-reported surveys, a Harman’s Single Factor Test was executed. The first unrotated factor accounted for only 19.74% of the variance–well below the 50% critical threshold–confirming that the data reflects genuine behavioral variance rather than systematic error. Concurrently, Principal Component Analysis (PCA) was utilized to condense perceptual items into orthogonal latent constructs.

Phase 2: Robust Determinant Modeling (Bootstrapping)

Latent factors were entered into an Ordinal Logistic Regression model to identify adoption drivers. To ensure the model possessed sufficient sensitivity to detect significant effects within the niche sample, a Post-hoc Power Analysis was cross-referenced, confirming a power level of 1-β=0.86, effectively minimizing Type II error risks. Furthermore, recognizing the non-normal distribution, the model was fortified using a Bootstrapping procedure (5,000 resamples) to generate Bias-Corrected 95% Confidence Intervals (BCa CI). This dual-validation ensures that identified determinants are statistically stable and not artifacts of specific outliers.

Phase 3: Non-Linear Segmentation (CHAID)

Transcending linear assumptions, the final stage deployed the CHAID (Chi-squared Automatic Interaction Detection) algorithm to map the market’s hierarchical decision structure. This method segmented respondents into distinct psychographic nodes (e.g., “Safety Skeptics” vs. “True Believers”) based on non-linear interactions. To ensure generalizability, the decision tree was validated via 10-Fold Cross-Validation. These validated psychographic nodes are then directly translated into Public Policy Roadmap, offering a sequential implementation plan for regulatory intervention.

Ⅳ. Results and Discussions

1. Diagnostic Profiling and Data Integrity

Prior to inferential modeling, data integrity was rigorously vetted. Harman’s Single Factor Test indicated that the first factor explained only 19.74% of the variance, confirming the absence of pervasive Common Method Bias. Crucially, despite the niche sample size (N=204), the Post-hoc Power Analysis <Fig. 2> confirmed a statistical power (1-β) of 0.86, exceeding the standard threshold of 0.80. This visual validation ensures the dataset possesses high sensitivity to detect behavioral determinants without committing Type II errors.

<Fig. 2>

Post-hoc Power Analysis using G*Power (N=204)

Dominated by the productive workforce within the Java-Bali economic corridor (80.8% aged 25-59), the sample profile confirms the study’s specific focus on the ‘Urban Early Adopters’ segment. Specifically, 63.8% possess ‘Intellectual Capital’ (Bachelor’s/Master’s degrees), yet 65.7% remain motorcycle-dependent. Notably, 52.5% report a history of road trauma, validating the premise that this demographic represents ‘Vulnerable Road Users’ seeking safety, rather than sheltered car users seeking luxury.

2. Empirical Validation of Behavioral Paradoxes

To validate the structural assumptions of the framework, comparative analyses were conducted to rigorously test the study’s two core hypotheses. First, regarding the Modal Leapfrog Effect, an Independent Samples T-Test confirmed a distinct behavioral gap where vulnerable motorcyclists exhibited a significantly higher willingness to switch to AVs compared to private car users (Cohen’s d=0.34). This statistical evidence confirms that the motivation to “upgrade safety” creates a stronger pull for motorcyclists than for car users who are already within a comfort zone.

Concurrently, to test the Education Paradox, a One-Way ANOVA validated the counter-intuitive hypothesis that higher literacy fosters skepticism; respondents with Postgraduate degrees expressed the highest level of “Legal & Ethical Anxiety” (Mean = 0.57), nearly double that of the Non-Bachelor group. This proves that the urban early adopter possesses a “Calculated Risk” mindset–they are highly conscious of systemic risks yet rationally willing to adopt if safety is guaranteed.

3. Dimension Reduction Analysis

To structure the complex perception variables, Principal Component Analysis (PCA) was conducted, extracting five orthogonal components explaining 63.16% of the variance. As detailed in <Table 3>, the analysis yielded two critical psychographic insights. First, Perceived Safety Advantage (Factor 4) emerged as a standalone dimension distinct from Traffic Efficiency (Factor 5), statistically confirming that for Indonesian users, “Safety” is a fundamental value proposition separate from mere convenience. Second, the Pragmatic Utility dimension (Factor 3) intriguingly grouped “Convenience” benefits with “High Costs.” Conceptually, this structure indicates that in a developing economy context, urban commuters perceive advanced autonomous features (e.g., automated parking and productivity) not as standard utilities, but as premium luxury services. Therefore, convenience and high financial cost are mentally coupled as a single ‘premium utility’ trade-off. These five orthogonal factor scores were saved as independent predictors for the subsequent regression modeling.

4. Bootstrapped Ordinal Regression Analysis

To determine the distinct impact of psychographic and demographic predictors on the willingness to switch to AVs, an Ordinal Logistic Regression was conducted. Given the limited sample size (N=204) and the skewed distribution of the private car demographic, a Bootstrapping procedure with 5,000 resamples (BCa 95% Confidence Intervals) was employed to ensure the robustness of the parameter estimates. The model demonstrated valid goodness-of-fit. While the standard Test of Parallel Lines yielded a significant result (p = .000), SPSS diagnostics indicated that the validity of this test is uncertain due to the presence of empty cells (zero frequencies) caused by the continuous nature of the factor scores. Consequently, following standard robust practices, we relied on the Bootstrapped Confidence Intervals (5,000 resamples) to ensure the stability and validity of our parameter estimates.

The bootstrapped parameter estimates <Table 4> reveal three critical findings. First, consistent with the “Safety Universalism” premise, Perceived Safety Advantage emerged as the most dominant predictor (β= 1.196, p < .001, Exp(β) = 3.307), with a robust confidence interval [0.765,1.703] that does not cross zero. Its impact magnitude is nearly double that of Traffic Efficiency (β=0.660) or Convenience (β=0.586), confirming that for Indonesian commuters, the value proposition is rooted in survival rather than mere comfort.

Second, the analysis validates the “Modal Leapfrog” hypothesis. Private car users exhibited strong negative resistance to switching (β=-1.327, p=.012), whereas motorcyclists showed significantly less resistance (β=-0.428). This statistical gap confirms that “sheltered” car users are highly accustomed to the safety of their current mode, while “vulnerable” motorcyclists are pushed by safety needs to leapfrog directly to AVs.

Third, regarding the “Education Paradox,” the results reveal a “Calculated Risk” mindset among the early adopters. While Postgraduates expressed the highest concerns in descriptive analysis, the findings conceptually demonstrate that they simultaneously possess the highest willingness to adopt (Reference Category vs. Non-Bachelor β=-1.063). It should be noted that this paradox is interpreted conceptually in parallel, where high academic literacy independently correlates with both skepticism and adoption intent, rather than through a direct statistical interaction or mediation analysis.

This implies they are highly conscious of risks yet rationally decide to adopt because they value the safety benefits significantly more. Intriguingly, Technological Risk showed a positive association (β=0.675), suggesting that “Concern” acts as a proxy for “Technological Engagement”-respondents knowledgeable enough to worry about cybersecurity are the same demographic eager to adopt.

A critical nuance in validating the ‘Modal Leapfrog’ hypothesis is the explicit distinction between voluntary technological preference and what must be termed ‘forced’ or compensatory adoption. For the motorcycle-dependent demographic, the high willingness to switch to AVs does not stem from an inherent fascination with autonomous technology. Rather, given the alarmingly high rate of prior traffic trauma (52.5%), their choice represents a pragmatic escape compelled by the chronic physical dangers of Indonesia’s Heterogeneous Disordered Traffic (HDT) environment.

However, addressing the endogeneity of the ‘motorcycle user’ variable reveals a deeper socio-economic reality. Relying on a two-wheeler is dictated by economic constraints and the tactical necessity to minimize travel time in severe urban congestion. While vulnerable road users desire the protective enclosure of an AV, transitioning to a private four-wheeled AV would paradoxically result in a significant increase in travel time and is financially impossible for the broader populace. Therefore, this ‘safety-seeking imperative’ represents a latent desire rather than an immediate effective demand. This desire can potentially materialize if AVs are deployed as Shared Autonomous Services (S-AVs) operating on dedicated lanes, directly offsetting both the prohibitive financial barrier and the travel-time penalty.

Furthermore, the positive correlation between Technical Risk (F2) and adoption intent (β=0.675) presents a ‘Technical Literacy Paradox.’ While seemingly counterintuitive, this suggests that our respondents-largely ‘Urban Early Adopters’ with high education-do not view technical risk as an absolute deterrent, but rather as a systemic challenge to be mitigated through regulation and technological maturity. However, addressing the reviewer’s critical insight, we must acknowledge that this finding may be subject to a ‘statistical illusion’ driven by sample bias. Because our survey heavily sampled highly literate segments (e.g., academic networks and international scholars), these results primarily reflect the distinct ‘techno-optimist’ perspective of an urban elite. For this specific demographic, high technical comprehension fosters a ‘calculated risk’ mindset. In contrast, for the general Indonesian populace with lower technical literacy, high perceived technical risk would likely act as a traditional, insurmountable barrier to adoption. Therefore, while this paradox is valid for the early-adopter segment modeled in this study, future research with a more socio-economically diverse sample is strictly necessary to determine if this behavior holds across the broader market.

5. CHAID Based Market Segmentation

To validate the robustness of the CHAID model given the sample size (N=204), the model’s stability was rigorously tested using 10-fold cross-validation. The analysis achieved an overall classification accuracy of 58.8%, which is notably impressive for a three-category behavioral model and significantly exceeds the proportional chance criterion. The model demonstrates particularly strong predictive precision for the ‘Netral’ (70.0%) and ‘Agree’ (64.0%) categories. To ensure statistical reliability and prevent over-fitting, strict stopping rules were implemented, requiring a minimum of 50 cases for parent nodes and 20 cases for child nodes. Consequently, every terminal node in the final tree maintains a substantial size of at least 41 cases, ensuring that the identified segments are statistically stable. With a cross-validation risk estimate of 0.466 (SE: 0.035), the model provides a sound and credible foundation for deriving targeted public policy roadmaps.

Transcending linear regression, the CHAID analysis (Fig. 3) mapped the hierarchical structure of the market, revealing a two-tiered decision process driven by Physical Safety as the primary root and Socio-Legal Concern as the secondary filter.

<Fig. 3>

CHAID Decision Tree for AV Adoption

At the extremes, the model identifies two distinct clusters based on physical trust. It should be noted that the split thresholds (e.g., -0.689 and 0.992) represent standardized factor scores (Z-scores) derived from the preceding PCA, where a mean of 0 indicates the sample average. The “Safety Skeptics” (Node 1), characterized by low safety perception (≤-0.689), exhibit a disproportionately high rejection rate of 23.7%, confirming that without a baseline threshold of physical trust, adoption is unlikely regardless of other benefits. Conversely, the “True Believers” (Node 3), who have crossed the high-safety threshold (>0.992), represent the ideal early adopters with a substantial 75.6% willingness to switch and negligible resistance (2.4%), proving that high safety assurance is the single strongest driver of adoption.

Crucially, the analysis uncovers a “hidden” divergence in the middle segment. Among respondents with moderate safety perception, Socio-Legal Concern acts as the decisive filter. The “Anxious Pragmatists” (Node 5) are held back by high institutional distrust (liability/privacy concerns), resulting in a dominant “Neutral” stance (55.0%). However, those with low concerns emerge as “Unburdened Adopters” (Node 4), showing a strong willingness to switch (59.1%) with zero rejection. This finding is pivotal for policy: it suggests that resolving legal friction (Node 5 → Node 4) is just as effective as maximizing safety features for capturing the middle-class market.

The nomenclature for each identified cluster is derived from the convergence of the model’s split thresholds and established innovation adoption theories. Specifically, the labels ‘Safety Skeptics’ and ‘True Believers’ are grounded in the Z-score extremes of physical trust, reflecting the ‘Risk Aversion’ and ‘Early Adopter’ archetypes within Rogers’ Diffusion of Innovation theory. The ‘Anxious Pragmatists’ and ‘Unburdened Adopters’ are defined based on the socio-legal ‘filter’ effect, where institutional trust acts as the primary differentiator for the pragmatic middle-segment. This grounded labeling ensures that the policy segments are not merely descriptive but reflect both statistical boundaries and established behavioral archetypes.

Beyond the statistical groupings, these findings necessitate a critical reflection on the socio-economic realities of the Indonesian context. While the model identifies a high adoption intent among certain segments, it is essential to distinguish between voluntary preference and a ‘forced’ safety-seeking behavior. For many motorcyclists, the inclination toward AVs may represent a pragmatic escape from the physical perils of the HDT environment, despite the potential trade-off in travel-time maneuverability. Furthermore, the reported intent should be interpreted as a ‘desired transition’ rather than immediate market demand, as economic constraints remain a significant barrier for the broader populace. The positive correlation between Technical Risk and adoption intent among these ‘Urban Early Adopters’ also highlights a ‘Technical Literacy Paradox,’ where risk is perceived as a regulatory challenge to be managed rather than a deterrent. By acknowledging these nuances, the CHAID segments provide a more realistic foundation for a public policy roadmap that accounts for both techno-optimism and economic pragmatism.

6. Public Policy Roadmap

Translating the derived psychographic nodes into actionable metrics, a quantitative “What-If” simulation was conducted. It is important to note that these simulations represent conditional structural projections based on the current model’s hierarchy, rather than asserting direct causal effects of regulatory reform.

First, the “Safety Leap” simulation quantifies the efficacy of institutionalizing national AV safety standards via high-level executive decrees (e.g., Presidential or Ministerial regulations). To navigate the stochastic nature of Heterogeneous Disordered Traffic (HDT) environments, this framework mandates strict spatial segregation or complete prohibition of motorcycles on AV-designated expressways. This critical policy intervention significantly narrows the Operational Design Domain (ODD) for AV perception systems. Concurrently, to address domestic capital constraints, the roadmap leverages Official Development Assistance (ODA) mechanisms such as KOICA or JICA to finance and deploy the prerequisite Intelligent Transport Systems (ITS) infrastructure. Given that the regression model identifies Perceived Safety as the dominant predictor (β=1.196), a policy intervention that increases the population’s safety perception by one standard deviation is projected to increase the odds of adoption by 3.3 times (330%). In terms of segmentation, this shift would effectively migrate users from the “Safety Skeptics” cluster (Node 1)-where active rejection is high at 23.7%-directly toward the “True Believers” profile (Node 3), where willingness to switch reaches 75.6%.

Second, the “Legal Unlock” simulation estimates the conversion rate achievable by resolving regulatory uncertainties. The CHAID analysis reveals a critical behavioral divergence in the middle-class segment based solely on legal concerns. By enacting clear liability and data privacy laws, the model predicts a shift of the “Anxious Pragmatists” (Node 5) to the “Unburdened Adopters” (Node 4) profile. Based on the model’s structure, addressing this institutional barrier is conditionally projected to be associated with a doubled adoption rate within this segment (from 28.3% to 59.1%), suggesting that regulatory certainty could serve as a cost-effective driver to convert the hesitant majority without immediately requiring further technological upgrades.

Finally, addressing the ‘High Cost’ barrier identified in the Pragmatic Utility dimension (Factor 3), the ‘Pragmatic & Infrastructure Dividend’ simulation targets the remaining passive believers to achieve market saturation. Despite high trust, approximately 22% of the ‘True Believers’ (Node 3) remain neutral due to transactional barriers and concerns over Indonesia’s chaotic traffic. To bypass the prohibitive initial capital expenditure of private AV ownership for the middle class, the strategy prioritizes a transition toward Shared Autonomous Services (S-AVs) operating on dedicated lanes. Furthermore, to stimulate early private adoption, the roadmap incorporates targeted fiscal instruments, including subsidized green financing schemes and luxury tax exemptions (e.g., PPnBM reductions) for Level 4 autonomous vehicles. Coupled with Environment-Adaptive AI standards to navigate the HDT environment, these combined financial and infrastructural interventions are projected to push the segment’s adoption rate to a near-saturation level of approximately 86.6%. These three sequential phases-Trust, Legal, and Pragmatic Feasibility-are synthesized into the Phased Strategic Roadmap presented in <Table 5>.

Ⅴ. Conclusion

This study fundamentally challenges the prevailing Western-centric narratives of Autonomous Vehicle (AV) adoption by providing empirical evidence from the chaotic, mixed-traffic context of the Global South. Utilizing a “Paradox-Based Behavioral Framework,” the research validates the existence of the “Modal Leapfrog” effect. However, it clarifies that for vulnerable, motorcycle-dependent populations, transitioning to AVs represents a ‘forced’ pragmatic escape driven by the endogeneity of their current mode, rather than a purely voluntary technological preference. AVs function not as a luxury convenience but as a critical “Safety Upgrade” to bypass car ownership for physical protection. Furthermore, the analysis resolves the “Education Paradox,” revealing that the Indonesian urban early adopters operate on a “Calculated Risk” mindset; higher academic literacy correlates with increased technical skepticism, yet simultaneously fuels higher adoption intent. The study critically acknowledges that this reflects the techno-optimist bias of an urban elite, which may act as a strict barrier for the broader populace.

Methodologically, the application of a “Psycho-Statistical” framework integrating Bootstrapped Ordinal Logistic Regression and CHAID Decision Trees successfully mapped robust adoption determinants despite strategic sampling constraints (N=204). The regression results identify Perceived Safety Advantage (β=1.196) as the most dominant predictor, with an impact magnitude nearly double that of traffic efficiency. Crucially, the non-linear segmentation reveals that market resistance is hierarchical: while “Safety Skeptics” (Node 1) are deterred by physical distrust, the substantial “Anxious Pragmatist” segment (Node 5) is stalled solely by institutional ambiguity regarding liability, proving that legal certainty is as vital as technological reliability in converting latent demand.

Consequently, this study proposes a sequential Public Policy Roadmap moving from a “Trust Foundation” to “Market Acceleration.” Governments in emerging economies must prioritize infrastructure financing through Official Development Assistance (ODA) frameworks to simplify the Operational Design Domain (ODD)-specifically via the regulatory prohibition or physical segregation of motorcycles-alongside enacting clear Liability Laws to unlock the skeptical and anxious tiers. Once trust is established, market saturation requires overcoming prohibitive financial barriers through Shared Autonomous Services (S-AVs) and targeted fiscal instruments (e.g., tax exemptions). While this research is limited to the urban early adopter demographic in Java-Bali, the findings offer a conditional structural framework that may be applicable to other Global South contexts: the transition to autonomous mobility is not merely about creating smart cities, but about providing a “safer shelter” for the vulnerable majority to ensure their survival on the road.