Understanding Consumers’ Adoption Behavior of Driverless Delivery Vehicles: Insights from the Combined Use of NCA and PLS-SEM

Abstract

The rapid development of autonomous driving technology has been a key driver for the emergence of driverless delivery vehicles. To promote wider adoption, it is essential to address consumers’ concerns about safety and reliability, leverage psychological factors, and implement supportive policies that encourage technology adoption while ensuring public safety and privacy. Therefore, it is necessary to explain and predict consumers’ behavior and intention to adopt driverless delivery vehicles. To this end, this study extends the Technology Acceptance Model (TAM) to include technological complexity and perceived trust. This study evaluates the model by applying necessary condition analysis (NCA) and partial least squares structural equation modeling (PLS-SEM) to analyze data from 579 respondents from Jiangsu Province, China. This study explores the sustainability implications of autonomous delivery vehicles, highlighting their potential to reduce environmental impact and promote a more sustainable transportation system. The outcomes indicate that perceived ease of use (PEU), attitude, perceived trust, technological complexity (TECOM), and perceived usefulness (PU) are significant determinants and necessary conditions of consumers’ intention to adopt driverless delivery vehicles. Perceived trust and TECOM had a significant and indirect influence on consumers’ intention to adopt driverless delivery vehicles via PU and PEU. Perceived trust and technological complexity had a substantial impact on consumers’ adoption intention of driverless delivery vehicles. The study recommends that managers work closely with regulators to ensure their technologies meet all local standards and regulations. It also recommends its potential to reduce carbon emissions, improve energy efficiency, and contribute to a more sustainable transportation system.

1. Introduction

Many modern technologies have been rapidly consumerized, which has enabled initiatives like driverless delivery vehicles (DDVs) [1]. China is one of the leading places for autonomous vehicle (AV) development and deployment, fueled by progressive public policy, massive investments, and creative local companies [2]. In China, the emergence of DDVs has reshaped the logistics and delivery sectors, providing several benefits, including lowered operational costs, increased efficiency, safer travel, reduced energy consumption, and decreased emissions, as well as solving some last-mile delivery challenges [1,3]. For namesakes, in Yangquan, which has a population of 1.3 million, 12 DDVs have delivered packages for six months, providing evidence of their practical and effective usage [4]. DDVs have the potential to revolutionize the delivery industry by increasing efficiency, lowering costs, and easing congestion on the roads. However, their successful adoption relies on public acceptance. Singular benefits aside, enabling early adoption is inextricably linked to addressing widespread public skepticism and concerns about viability, including safety [5]. Factors like challenging road conditions, vehicle failure, security problems, hacking and data theft, and operational issues, such as the lack of customers at delivery sites, can lead to failed delivery and an increasing cost of delivery [6,7].

The DDV development is driven by the initiatives and studies of the policy supported by the Chinese government [1]. The “14th Five-Year Plan,” published in January 2022, encourages community autonomous delivery and courier accessory facility construction, providing relevant policy support for more cities to create driverless delivery demonstration zones [8]. In recent years, the deployment of driverless delivery vehicles (DDVs) has gained significant attention. Companies like JD.com are being used to operate low-speed DDVs, with market potential expected to reach trillions by 2030 [1,5]. In the US, Europe, and China, DDVs have regulatory support for public road operations, which further helps to accelerate the adoption [5,6,9]. In addition, DDV adoption accelerated during the COVID-19 pandemic. Customers preferred contactless delivery and had high expectations for delivery performance [5,10]. For instance, JD and Meituan deployed over 300 autonomous delivery vehicles in Shanghai during the pandemic. This showed the potential of DDVs to address logistical challenges while ensuring safety and efficiency. However, privacy and data security concerns remained unaddressed, limiting broader adoption. Similarly, other cases, such as UPS’s autonomous freight delivery tests and Nuro’s driverless delivery trials, provide valuable insights into the challenges and opportunities associated with ADVs. These examples highlight the importance of understanding what drives consumer adoption of such technologies.

Public acceptance of autonomous driving vehicles (DDVs) is influenced by a range of factors, including demographic, psychological, and socioeconomic aspects [1,3,11]. To understand this phenomenon, researchers have often employed theoretical frameworks such as the Unified Theory of Acceptance and Use of Technology (UTAUT) and the Technology Acceptance Model (TAM). However, the literature has not adequately compared these frameworks in the context of DDV adoption. UTAUT is a comprehensive framework that integrates multiple theories to explain the acceptance and use of technology. It considers factors such as performance expectancy, effort expectancy, social influence, and facilitating conditions [12]. UTAUT has been widely used to study the adoption of various technologies, including driverless buses in China [12]. However, it may not fully capture the unique psychological and trust-related aspects of autonomous vehicle adoption. TAM focuses primarily on two key constructs: perceived usefulness and perceived ease of use [13,14,15]. While TAM has been successful in explaining technology adoption in many contexts, it may overlook other critical factors such as technological complexity and perceived trust, which are particularly relevant for DDVs [13]. Despite the strengths of UTAUT and TAM, there are significant gaps in the literature regarding the adoption of DDVs in China. Previous studies using UTAUT have provided valuable insights but have not addressed the impact of technological complexity and perceived trust [12]. Technological complexity refers to the difficulty in understanding and using the innovation [13,16]. Perceived trust is the level of confidence individuals have in autonomous vehicles [5,16]. These factors are crucial for understanding public acceptance but have been largely overlooked. Moreover, while TAM has been widely used to study technology adoption, it often focuses on perceived usefulness and ease of use, neglecting other important factors [13,14,15]. Integrating technological complexity and perceived trust into TAM can provide a more comprehensive understanding of the factors driving citizens’ intention to accept DDVs. To address these gaps, this study integrates technological complexity and perceived trust into the TAM framework. We use Partial Least Squares Structural Equation Modeling (PLS-SEM) [17] and Necessary Condition Analysis (NCA) [18] to explore and examine the influence of these factors on consumers’ adoption intention in China. PLS-SEM is widely used in business and marketing to assess predictive models and supports the predictive power of TAM [19]. Unlike Covariance-Based Structural Equation Modeling (CB-SEM), which focuses on model fit, PLS-SEM evaluates the relationships between measured and predicted constructs [17]. NCA identifies necessary conditions for adoption, determining what must be present for the desired outcome [18].

It makes a useful contribution to knowing how consumers behave in relation to DDVs in countries such as China. Separate evaluations of the future scope of the proposed research model can be made through NCA and PLS-SEM, which handle more practical rationale as it implies the difference between the importance of the elements [20,21,22]. Through the incorporation of TAM, this research adds to the literature that focuses on consumers’ adoption intentions for DDVs by suggesting an innovative paradigm to examine other significant constructs and how they contribute to consumer behavior. The research scrutinizes the mediating effect of TAM constructs on the link between technological factors (like technological complexity and perceived trust) and consumers’ adoption intentions toward DDVs. The increasing attention of scholars to autonomous vehicles and their challenges and opportunities [1,23,24] only makes it more relevant to understand factors affecting the adoption of DDVs. Additionally, this research bridges a gap in the existing studies by applying TAM within the context of DDVs and highlighting the role of PU, PEU, and attitude in driving consumers’ continued involvement with DDVs. These findings provide actionable insights into speeding up the acceptance and incorporation of DDVs by all stakeholders in the logistics and transport industry. The rapid advancement of autonomous delivery vehicles (ADVs) holds the promise of transforming the logistics and delivery industry through enhanced efficiency and sustainability. However, the widespread adoption of this technology hinges on understanding consumer perceptions and intentions. This study aims to explore the factors influencing consumer adoption of ADVs, with a particular focus on how these vehicles can contribute to technological innovation (SDG 9), resilient infrastructure, reduced urban traffic congestion and pollution (SDG 11), and climate change mitigation through lower greenhouse gas emissions (SDG 13). By examining these aspects, our research seeks to provide actionable insights for policymakers and industry stakeholders to foster a more sustainable and efficient delivery ecosystem.

2. Hypothesis Development

2.1. Technology Acceptance Model (TAM)

The TAM is a major theoretical framework for adopting technology and information systems. It was first introduced by Davis in 1984; it intends to explain what motivates users to use and accept new technologies [25,26]. The model believes that perceived ease of use (PEU) and perceived usefulness (PU) are the key components leading to a user accepting technology [27]. This model has evolved and adapted with time, and new constructs have been added to the original model, including attitude towards using the system, task-technology fit, and cognitive fit [28,29]. The TAM is centered around constructs such as PU, PEU, and attitude toward using the technology [30,31]. This model is well-known due to its predictive power regarding user acceptance and usage behavior.

PU is the extent to which an individual thinks that exploiting a certain system will make them better at their job. This construct is critical as it directly relates to the intention to use the technology [29]. Specifically, PU can be connected to the delivery efficiency and reliability of driverless vehicles [4,32,33]. The majority of empirical studies have reported PU as a significant predictor of technology usage and intention to use technology [34,35,36]. PU is driven by the flexibility, reliability, accuracy, and completeness of the system [37,38]. Furthermore, PU was suggested to encourage user acceptance through task-technology fit and cognitive fit [29].

PEU is defined as the extent to which an individual believes using a specific system would be free of effort. However, a user’s belief that the technology is easy to use has a great influence on the intention to use it [39]. PEU, in the context of driverless delivery, could refer, for example, to how easy it is to interact with the control systems of the vehicles or how simple the integration of such vehicles into logistics networks would be [40]. According to the authors of [26], PEU is presumed to have a direct effect on PU and indirect impacts on intentions through favorable or unfavorable attitudes toward the technology. PEU is impacted by a number of factors, including interface design, user training, and system feedback mechanisms [37,41]. Ease of use is also critical to mitigate user anxiety and/or increase user satisfaction with the technology [42].

Another important construct in the TAM is the Technology use attitude. It reflects a person’s general attitude towards exploiting a specific technology [43]. According to research, attitudes toward technology are regarded as a mediating variable between PU/PEU and behavioral intentions [44,45,46]. These attitudes involve both affective and epistemic aspects and behavioral components [47]. Positive attitudes toward technology are positively related to use intention [48,49]. Previous studies indicated that capturing a comprehensive underlying mechanism that evaluates attitudes could help to correctly measure their impact on an acceptance model [50,51].

The TAM was extended and modified by various researchers after its initial introduction. TAM2 and TAM3 included new constructs, like facilitating conditions and social influence, to account for differences in user acceptance in various contexts [28,52]. The UTAUT, with behavior theories such as the TPB and DOI, provided an even clearer explanation for these constructs [39,53]. In the context of operations and supply chain management, attitudes and soft skills play a crucial role in influencing the adoption and effective utilization of technology, as they can shape how individuals perceive and interact with new technological tools [54].

The TAM is a well-established structure for apprehending technology adoption. As the driverless vehicle is complex and a new technological innovation is being introduced to the market. This study extends the TAM framework by integrating technological complexity and perceived trust to offer a comprehensive understanding of driverless delivery vehicles’ acceptance and use (see Figure 1). Recent studies have shown that autonomous delivery vehicles can significantly reduce greenhouse gas emissions compared to traditional delivery methods, contributing to a more sustainable urban environment. The proposed hypotheses will inform future work and enable stakeholders to formulate strategies to effectively diffuse this nascent technology. Drawing upon the reviewed literature, the following hypotheses can be formulated for future empirical testing:

Hypothesis 1 (H1).

PU positively impacts consumers’ adoption intention of driverless delivery vehicles.

Hypothesis 1a (H1a).

PU positively impacts consumers’ attitude.

Hypothesis 2 (H2).

Attitude positively impacts consumers’ adoption intention of driverless delivery vehicles.

Hypothesis 3 (H3).

PEU positively impacts consumers’ adoption intention of driverless delivery vehicles.

Hypothesis 3a (H3a).

PEU positively impacts consumers’ attitude towards driverless delivery vehicles.

Hypothesis 3b (H3b).

PEU positively impacts perceived usefulness.

2.2. Perceived Trust

Trust serves as an important factor in the adoption of all Internet of Things (IoT) technologies. Still, it is particularly relevant in new or uncertain interaction contexts, such as those involving autonomous vehicles [55,56]. According to the TAM, PU plays an important role in determining technology acceptance and subsequent usage [51]. The availability of technology and its optimal performance are among the many factors that are paramount to trust [55,57]. One of the biggest constructs from a study of autonomous vehicles decoded trust as the influencer of PU [58]. PEU is defined as the extent to which a person perceives that using a specific system would not require effort [51]. In this context, trust is important for creating security and reliability, which softens use. Users have been shown to find systems easy to use when they can trust that the system will work correctly without causing inordinate hardship or going down. For instance, previous work has shown that trust is directly related to PEU, which decreases the effort to learn and use new technologies [59,60]. Indeed, PEU has been shown to have a direct impact on PU [61], and both are essential predictors of consumer AV-related behavior [5,62]. Moreover, trust also plays a mediated role in alleviating perceptions of risk and amplifying perceptions of benefit [63]. This explains the indirect contribution of trust towards perceived ease of use of innovation. Additionally, trust has been identified as a critical factor in reducing uncertainty and enhancing user confidence in autonomous systems, which is essential for widespread adoption [64]. This explains the indirect contribution of trust towards perceived ease of use of innovation. We propose the following:

Hypothesis 4 (H4).

Perceived trust positively impacts consumers’ adoption intention of driverless delivery vehicles.

Hypothesis 4a (H4a).

Perceived trust positively impacts PU.

Hypothesis 4b (H4b).

Perceived trust positively impacts PEU.

2.3. Technological Complexity

Technological complexity refers to a perception by members of a social system that technology is too complex to learn and use [65]. The production and implementation of automated driving vehicles (ADVs) depend significantly on technological complexity. This complexity is due to the integration of large AI systems, sensor technology, and software algorithms needed to function autonomously [66,67,68]. This complexity has implications for the operational features of ADVs as well as their reliability and safety, which are major concerns for end users and regulators [69]. A possible escalation of complexity for driverless delivery vehicles could trigger users to reorient due to the need to learn new behaviors with even more sophisticated actions, and suffer losses of functioning for the devices [65,70]. When technology is complex, it might be harder for users to learn to use it effectively, which can result in frustration that lowers their willingness to use it [71]. This becomes particularly vital as novel technologies such as autonomous vehicles disrupt transportation, and access must be intuitive for users if they are to embrace their utility [72,73]. Moreover, technological complexity has been shown to directly impact user adoption by influencing both perceived ease of use (PEU) and perceived usefulness (PU), which are key determinants of technology acceptance [51]. Our data is consistent with the notion that complexity can be a predictor of PU and PEU, each of which impacts technology adoption [61,74]. Also, effective management of organizational complexities is critical for successful technology adoption [75]. Higher complexity may have an adverse effect on PEU, as it can be perceived that consumers do not perceive connected and autonomous vehicles as easy or intuitive to operate [71]. As complexity increases, consumers may have a harder time seeing how the technology applies to them directly, resulting in a lower perceived usefulness. This concurs with results from technology acceptance studies, where complexity was shown to limit user involvement and satisfaction [76]. Therefore, we hypothesize the following:

Hypothesis 5 (H5).

Technological complexity negatively impacts consumers’ adoption intention of driverless delivery vehicles.

Hypothesis 5a (H5a).

Technological complexity negatively impacts perceived usefulness.

Hypothesis 5b (H5b).

Technological complexity negatively impacts PEU.

3. Materials and Methods

3.1. Context and Data Collection

We finalized a detailed questionnaire after significant preparatory work before the actual survey. The first version of the survey was informed by a framework from previous social analysis research conducted by [77], where they scaled multiple measurement items related to the latent constructs. The survey content consisted of three sections: decision variables, intention to adopt, and demographic data. Participants were residents of Jiangsu Province, China, aged 18 or above, familiar with the autonomous vehicles and delivery issues involved. We conducted an online questionnaire based on a cross-sectional design issued through Sojump. This professional online survey platform provides advanced features for data collection, survey design, customized charts, and data analysis [78]. The link to the survey was disseminated across multiple media platforms, and recipients were urged to share the link to increase participation.

Individuals were recruited online, and the sample size was determined to sufficiently test the hypotheses of the present study. To maintain participant confidentiality, no identifying information or personal characteristics were kept. As suggested by [79], the factor analysis criteria were configured as at least 3 times the number of samples surveyed and utilized as research variables. Data were collected from 24 November 2023 to 12 January 2024, using an online questionnaire completed by 580 out of an original sample size of 600. One questionnaire had to be discarded due to suspicious response patterns (see ref. [17]) leading to a final sample size of 579 respondents. According to a study of the inverse square root approach, the size meets the threshold of statistically significant results when the path coefficients are equal to 0.11 or higher with 80% power and a significance level of 0.05 [80]. A sample size greater than 200 is sufficient to minimize the bias and verify the adequacy of the SEM model [81]. Demographic information of the participants was summarized using SPSS 30.0.0.0. The sample composition reveals a diverse group of 579 participants with distinct demographic characteristics in

Table 1. Sample composition.

Variable	‘Subcategory’	Frequency	Percent
Gender	Male	345.0	59.6
	Female	234.0	40.4
Age	Under 25	101.0	17.4
	26–35	178.0	30.7
	36–45	112.0	19.3
	46–55	89.0	15.4
	Above 55	99.0	17.1
Marital status	Single	241.0	41.6
	Married	338.0	58.4
Employment status	Student	128.0	22.1
	Self-employed	111.0	19.2
	Government worker	63.0	10.9
	Private company worker	228.0	39.4
	Unemployed	49.0	8.5
Level of education	High school and below	47.0	8.1
	Junior high school	94.0	16.2
	Vocational/Technical Education	83.0	14.3
	Bachelor’s	264.0	45.6
	Master’s	87.0	15.0
	Ph.D.	4.0	0.7
Income level	Below CNY 2500	95.0	16.4
	CNY 2501–5000	62.0	10.7
	CNY 5001–7500	163.0	28.2
	CNY 7501–10,000	187.0	32.3
	Above CNY 10,000	72.0	12.4
Type of community	Urban area	463.0	80.0
	Rural area	116.0	20.0
The total of each variable		579 (100%)

. The majority of respondents are male (59.6%) and fall within the age range of 26–35 years (30.7%), suggesting a focus on younger, working-age individuals. Marital status is predominantly married (58.4%), while employment status shows a significant representation of private company workers (39.4%) and students (22.1%). In terms of education, the sample is relatively well-educated, with 45.6% holding a bachelor’s degree and 15.0% holding a master’s degree. Income levels are primarily concentrated between CNY 5001 and CNY 10,000 (60.5%), indicating a middle-income profile. Geographically, the sample is predominantly from urban areas (80.0%), with only 20.0% from rural regions. This demographic breakdown provides a comprehensive representation of different segments, highlighting the potential influence of gender, age, marital status, employment, education, income, and location on consumer adoption behavior towards driverless delivery vehicles.

3.2. Scale Validation Process

To ensure the reliability and validity of our measurement scales, we conducted a rigorous validation process. For each construct in our survey, such as perceived trust and technological complexity, we tested content validity through expert evaluation. A panel of experts was invited to review the survey items. These experts were selected based on their academic qualifications and extensive experience in the field of autonomous vehicles and related technologies. Specifically, the panel included 4 experts (Professors) with backgrounds in relevant fields, such as transportation engineering, human–computer interaction, and consumer behavior. These experts were selected based on their demonstrated expertise in their respective fields, including published research and professional experience related to autonomous vehicles and user acceptance.

The experts were provided with the survey items and construct definitions to evaluate whether the items adequately represented the constructs. They were asked to assess each item’s relevance, clarity, and comprehensiveness. Feedback was collected through a structured questionnaire and follow-up interviews. The results of this evaluation were used to refine the survey items. For instance, items with low content validity indices (CVIs) were revised or removed to enhance the overall validity of the scales.

To assess the underlying structure of our measurement items, we conducted an exploratory factor analysis (EFA). The EFA was performed using SPSS, with principal component analysis as the factor extraction method and varimax rotation. The criteria for factor retention included eigenvalues greater than 1 and an examination of the scree plot. The resulting factor structure revealed [number of factors] distinct dimensions for each construct, which aligned well with our theoretical framework. Factor loadings for each item were examined, and all the loading items were above 7.0. The reliability of each factor was assessed using Cronbach’s alpha, with values ranging from 0.921 to 0.933, indicating acceptable internal consistency.

To confirm the factor structure identified in the EFA and assess the fit of the measurement model, we conducted a confirmatory factor analysis (CFA). The CFA was performed using SmartPLS software 4.1.0.0, with maximum likelihood estimation. Model fit was evaluated using several indices, including the standardized root mean square residual (SRMR), normed fit index (NFI), and goodness of fit index (GFI) in

Table 2. Study model fit.

Items	Current Research Model	Benchmark Value
SRMR	0.023	<0.08
NFI	0.942	>0.80
GFI	0.918	>0.80
R2	0.907

. The results indicated a good fit of the model, with SRMR = 0.023 [82], NFI = 0.942 [83], and GFI = 0.918 [84]. Additionally, the R-squared (R²) value of 0.907 indicates that the model explains 90.7% of the variance in the dependent variable, highlighting its strong predictive power [85]. Factor loadings and cross-loadings were examined to ensure each item loaded significantly on its intended factor. The reliability and validity of the constructs were further assessed using composite reliability and average variance extracted (AVE). The composite reliability values ranged from 0.941 to 0.949, and the AVE values were above the threshold of 0.5, indicating convergent validity. Discriminant validity was confirmed by ensuring that the square root of the AVE for each construct was greater than the inter-construct correlations.

3.3. Data Analysis and Scales

The multi-item scales utilized in this research were modified from existing instruments, with minor alterations made to measure the latent constructs of the model. The constructs were identified as reflective, which aligns with the original scales utilized in the research. The measurement items included Technological Complexity, which was measured by exploiting five items based on ref. [65]. PEU: Measured exploiting 5 items from ref. [37]. Attitudes: Five items from ref. [46]. PU: Measured by five items from ref. [29]. TST: Measured exploiting 5 items from ref. [58]. Intention to adopt: The ten items were drawn from refs. [43,86] (see Appendix A). Participants were asked to rate each item on a five-point Likert scale from 1 (Strongly disagree) to 5 (Strongly agree).

Table 3. The descriptive statistics.

Construct	Median	Max	Min	Skewness	Standard Deviation	Excess Kurtosis
PU	0.399	1.104	−2.439	−1.225	1.000	−0.095
TST	0.404	1.121	−2.297	−1.197	1.000	−0.174
PEU	0.394	1.096	−2.423	−1.202	1.000	−0.181
DDVI	0.397	1.105	−2.435	−1.228	1.000	−0.113
ATT	0.412	1.130	−2.462	−1.238	1.000	−0.111
TECOM	0.319	1.509	−2.258	−1.2	1.000	−0.166

exhibits the descriptive statistics of the study. Furthermore, the skewness and kurtosis values for all items fell between −2 and 2, indicating that all items conformed to a normal distribution [17].

Data analysis relied on the implementation of the PLS-SEM and predictive model comparisons [17] methods, as well as the Necessary Condition Analysis (NCA) [19]. PLS-SEM analysis: PLS-SEM was preferred since it provides merit in assessing the predictive and explanatory power of the proposed model [87]. The PLS-SEM algorithm was run with a maximum number of iterations set as 3000, a stop criterion set as 10⁻⁷, and a path weighting scheme. To test each relationship’s significance in the structural model and confidence intervals, bootstrapping of 10,000 subsamples was employed via percentile bootstrapping [87]. The PLS-SEM model assessment applies additive or sufficiency logic principles, which enables the identification of what level of predictors equates to the higher level of the result [88]. The ‘NCA’ acronym can also refer to Necessary Condition Analysis. This approach draws on necessity logic to investigate whether the antecedents that were investigated act as NCA or constraints for the occurrence of an outcome of interest [21,88]. NCA is particularly useful for identifying necessary conditions that must be met for consumer adoption to occur. By focusing on necessity logic, we can determine which factors are essential precursors to adoption, even if they are not sufficient on their own. This approach helps us identify critical barriers and enablers that need to be addressed. It was performed separately for each study, using unstandardized latent variable scores from the PLS-SEM results. The combined use of NCA and PLS-SEM provides a more nuanced understanding of the adoption process. While NCA identifies the necessary conditions that must be met for adoption to occur, PLS-SEM helps us understand how these conditions interact with other factors to drive adoption. This dual approach allows us to address both the necessity and sufficiency of our constructs, providing a more comprehensive picture of the adoption process.

In this study, we employ Necessary Condition Analysis (NCA) and Partial Least Squares Structural Equation Modeling (PLS-SEM) to identify and validate the necessary conditions for consumer adoption of driverless delivery vehicles. For instance, the successful deployment of Level 4 autonomous vehicles by Cainiao Group in Hangzhou highlights the importance of technological reliability and efficiency as necessary conditions. To analyze the necessity of the antecedents for the outcome of interest, the data fit the Continuous Ceiling Regression with Free Disposal Hull (CR-FDH) model [89]. This model was chosen due to its ability to effectively handle the non-parametric nature of the data and provide a robust estimation of the necessity effects. The CR-FDH model is particularly suited for identifying the upper bounds of the relationships between the antecedents and the outcome, thereby highlighting the conditions under which the result can be achieved.

In addition to the CR-FDH model, we employed the Ceiling Envelopment with Free Disposal Hull (CE-FDH) model. These models are designed to envelop the data points from above, providing a clear visualization of the ceiling effects and identifying the necessary conditions for the outcome. The CE-FDH models were used to generate ceiling line plots, which visually represent the maximum achievable outcome for each level of the antecedents. The individual contribution of each antecedent was evaluated using effect sizes, which quantify the magnitude of the necessity effects. To ensure the statistical significance of these effects, we conducted an approximate permutation test with 10,000 iterations at a significance level of 0.05 [90]. This rigorous statistical approach allowed us to validate the necessity of each antecedent in achieving the outcome.

The results of the NCA agreed with the antecedents described in the NCA ceiling line plots. Specifically, the ceiling line plots for participants’ attitude (Figure 2), perceived ease of use (PEU) (Figure 3), perceived usefulness (PU) (Figure 4), perceived trust (Figure 5), and technological complexity (Figure 6) all demonstrated significant necessity effects. These findings indicate that these antecedents are necessary conditions for achieving the desired outcome. The entire analysis was executed using SmartPLS 4 software, providing a comprehensive platform for conducting PLS-SEM and NCA analyses. The software’s advanced features allowed for seamless integration of the two methods, ensuring that the results were statistically robust and theoretically meaningful.

The NCA ceiling line chart labeled Figure 2 illustrates the relationship between consumers’ attitude towards driverless delivery vehicles (DDVs) and their intention to adopt these vehicles. Utilizing Necessary Condition Analysis (NCA), the chart employs two analytical methods, Ceiling Regression with Free Disposal Hull (CR-FDH) and Ceiling Envelopment with Free Disposal Hull (CE-FDH),to identify the minimum threshold of attitude that must be met for consumers to consider adopting DDVs. The chart’s horizontal axis represents the level of attitude, while the vertical axis signifies the corresponding utility or satisfaction level. The blue dots denote individual observations from the study’s sample. The gray area indicates the results from the CR-FDH method, which is more conservative and focuses on the upper boundary of the data points. In contrast, the yellow area represents the CE-FDH method, which is more inclusive and covers a broader range of data points. The chart demonstrates that for attitude to be a necessary condition for DDV adoption, the satisfaction level must exceed a certain threshold, with the yellow area indicating the necessary condition as determined by the CE-FDH method. This visualization is instrumental in understanding the minimum level of positive attitude required to influence consumer adoption intentions for driverless delivery vehicles, thereby aiding in developing targeted strategies to foster the acceptance and utilization of this emerging technology.

The NCA ceiling line chart in Figure 3 is designed to analyze the relationship between perceived ease of use (PEU) and the intention to adopt driverless delivery vehicles (DDVs). This chart uses two methods of Necessary Condition Analysis (NCA), specifically Ceiling Regression with Free Disposal Hull (CR-FDH) and Ceiling Envelopment with Free Disposal Hull (CE-FDH), to establish the minimum acceptable levels of PEU that are necessary for the adoption of DDVs. The horizontal axis represents the degree of perceived ease of use, while the vertical axis shows the utility or satisfaction level associated with DDV adoption. The blue dots are individual data points from the study’s participants. The gray area indicates the results from the CR-FDH method, which is more conservative and focuses on the upper boundary of the data points that meet the necessary conditions. The yellow area represents the CE-FDH method, which is more inclusive and covers a broader range of data points that satisfy the necessary conditions. The chart reveals that the satisfaction level must surpass a certain threshold for PEU to be considered an essential condition for DDV adoption. Data points falling within the yellow area are above the ceiling line set by the CE-FDH method, indicating that PEU is a necessary condition for adoption. Similarly, data points within the gray area meet the necessary condition but are subject to a more stringent criterion per the CR-FDH method. This visualization is crucial for understanding the minimum level of perceived ease of use required to influence consumer adoption intentions for driverless delivery vehicles. It aids in developing strategies to enhance the user-friendliness of DDVs, thereby promoting their acceptance and use among consumers.

Figure 4 presents an NCA ceiling line chart that examines the relationship between perceived usefulness (PU) and the intention to adopt driverless delivery vehicles (DDVs). The chart is utilized to determine the necessary conditions for adoption based on the perceived usefulness of the technology. It employs two analytical approaches within the NCA framework: Ceiling Regression with Free Disposal Hull (CR-FDH) and Ceiling Envelopment with Free Disposal Hull (CE-FDH). The horizontal axis of the chart represents the perceived usefulness scores, while the vertical axis denotes the utility or satisfaction level associated with the adoption of DDVs. Each blue dot corresponds to an individual observation from the study’s dataset. The gray area illustrates the results from the CR-FDH method, which is more conservative and focuses on the upper boundary of the data points considered necessary for adoption. The yellow area, on the other hand, represents the CE-FDH method, which is more inclusive and encompasses a broader range of data points that meet the necessary conditions for adoption. The chart reveals that for PU to be a necessary condition for the adoption of DDVs, the satisfaction level must exceed a certain threshold. Observations within the yellow area are above the ceiling line established by the CE-FDH method, signifying that a certain level of perceived usefulness is necessary for consumers to consider adopting DDVs. Similarly, observations within the gray area also meet the necessary condition but are subject to a stricter criterion defined by the CR-FDH method. This visualization is essential for identifying the minimum level of perceived usefulness required to influence consumer adoption intentions for driverless delivery vehicles. It provides valuable insights for stakeholders to enhance the perceived usefulness of DDVs, which can ultimately promote their acceptance and integration into consumer lifestyles.

Figure 5 presents an NCA ceiling line chart focusing on the relationship between perceived trust (TST) and the intention to adopt driverless delivery vehicles (DDVs). This chart is instrumental in identifying the necessary conditions for adopting DDVs based on the level of trust consumers place in the technology. The horizontal axis of the chart represents the perceived trust scores, while the vertical axis indicates the utility or satisfaction level associated with adopting DDVs. The blue dots on the chart correspond to individual observations from the study’s dataset. The gray area, which represents the CR-FDH method, is more conservative and highlights the upper boundary of data points that meet the necessary conditions for adoption. In contrast, the yellow area, representing the CE-FDH method, is more inclusive and covers a broader range of data points that satisfy the necessary conditions. The chart reveals that for TST to be considered a necessary condition for adopting DDVs, the satisfaction level must exceed a certain threshold. Data points within the yellow area are above the ceiling line set by the CE-FDH method, indicating that a certain level of perceived trust is necessary for consumers to consider adopting DDVs. Observations within the gray area also meet the necessary condition but are subject to a stricter criterion defined by the CR-FDH method. This visualization is crucial for understanding the minimum level of perceived trust required to influence consumer adoption intentions for driverless delivery vehicles. It provides valuable insights for stakeholders to enhance the trustworthiness of DDVs, which can ultimately promote their acceptance and integration into consumer lifestyles. By ensuring that the technology is perceived as trustworthy, companies can foster a more favorable environment for the adoption of DDVs.

Figure 6 displays an NCA ceiling line chart that analyzes the impact of technological complexity (TECOM) on the intention to adopt driverless delivery vehicles (DDVs). The chart uses Necessary Condition Analysis (NCA) to determine the threshold level of technological complexity that consumers must perceive before they consider adopting DDVs. Two methods are applied within the NCA framework: Ceiling Regression with Free Disposal Hull (CR-FDH) and Ceiling Envelopment with Free Disposal Hull (CE-FDH). The horizontal axis of the chart represents the scores for technological complexity, while the vertical axis shows the utility or satisfaction level associated with the adoption of DDVs. Each blue dot corresponds to an individual observation from the study’s dataset. The gray area indicates the results from the CR-FDH method, which is more conservative and focuses on the upper boundary of the data points necessary for adoption. The yellow area represents the CE-FDH method, which is more inclusive and covers a broader range of data points that meet the necessary conditions. The chart reveals that for TECOM to be considered a necessary condition for the adoption of DDVs, the satisfaction level must exceed a certain threshold. Observations within the yellow area are above the ceiling line set by the CE-FDH method, indicating that a certain level of technological complexity must be manageable for consumers to consider adopting DDVs. Data points within the gray area also meet the necessary condition but are subject to a stricter criterion defined by the CR-FDH method. This visualization is crucial for understanding the maximum technological complexity consumers are willing to accept for driverless delivery vehicles. It provides insights for stakeholders to design DDVs that are perceived as less complex, which can ultimately promote their acceptance and integration into consumer lifestyles. By reducing the perceived complexity of DDVs, companies can foster a more favorable environment for the adoption of this technology.

4. Results

4.1. Dimension Model Appraisal

The final scale was evaluated for its reliability in the context of the study to calculate the reliability index, internal consistency reliability, and results of discriminant and convergent validity;

Table 4. Dimension model appraisal for adoption intention of driverless delivery vehicles.

Construct	Item	FL	CA	Rho_A	CR	AVE
Attitude	ATT1	0.885	0.921	0.922	0.941	0.761
	ATT2	0.872
	ATT3	0.863
	ATT4	0.867
	ATT5	0.874
Driverless delivery vehicle’s intention	DDVI1	0.882	0.930	0.930	0.947	0.782
	DDVI2	0.891
	DDVI3	0.881
	DDVI4	0.877
	DDVI5	0.890
Perceived ease of use	PEU1	0.884	0.926	0.926	0.944	0.771
	PEU2	0.864
	PEU3	0.885
	PEU4	0.878
	PEU5	0.877
Perceived usefulness	PU1	0.882	0.929	0.929	0.946	0.779
	PU2	0.880
	PU3	0.885
	PU4	0.883
	PU5	0.881
Technological complexity	TECOM1	0.876	0.933	0.933	0.949	0.788
	TECOM2	0.889
	TECOM3	0.889
	TECOM4	0.896
	TECOM5	0.887
Perceived trust	TST1	0.877	0.926	0.927	0.944	0.773
	TST2	0.866
	TST3	0.874
	TST4	0.885
	TST5	0.892

Note: FLs = factor loadings, CA = Cronbach’s alpha, CR = Composite reliability.

represents outer loadings for individual items with a value of composite reliability, Cronbach Alpha (CA), and exact reliability coefficient. As all outer loadings exceeded the threshold of 0.70, this confirms indicator reliability [91]. The internal consistency reliability was calculated through CA, composite reliability, and exact reliability coefficient, and all fall between 0.90 and 0.95 in the analysis. All AVEs were greater than 0.50, which signifies that convergent validity was achieved [17]. Following [87], the study compared the square root of AVE with the correlation between constructs to exploit discriminant validity.

The HTMT ratios of correlations and their 95% one-sided bootstrap confidence intervals were analyzed in order to assess discriminant validity (Ringle et al., 2023 [87]). The HTMT value was below the cutoff of 0.90. For conceptually similarly related constructs, the upper bound of the 95% confidence interval for the HTMT ratio marginally exceeded 0.90 in a few instances (see Table 3). Discriminant validity was still validated [87].

Table 5. HTMT criterion.

HTMT
	ATT	DDVI	PEU	PU	TECOM	TST
ATT
DDVI	0.478
PEU	0.263	0.383
PU	0.264	0.231	0.384
TECOM	0.65	0.183	0.327	0.372
TST	0.557	0.454	0.377	0.311	0.465

shows the heterotrait–monotrait criterion.

4.2. Common Method Bias (CMB)

Common method variance (CMV) was proactively addressed in this study to minimize its potential impact on the results, following the recommendations of Zheng et al. [92]. Several procedural measures were employed to mitigate CMV. First, the structure of the questionnaire was optimized before data collection, ensuring brevity and clarity. Respondents were assured of anonymity to reduce social desirability bias and evaluation apprehension [93]. The questionnaire emphasized the importance of providing honest responses to minimize response bias. The order of items was counterbalanced to avoid sequence-related bias, and the questionnaire was kept concise to maintain respondent engagement, thereby reducing the risk of fatigue or boredom [94]. Additionally, demographic questions were positioned at the end to prevent their influence on responses to the key constructs.

To further address CMV, analytical techniques were employed. Harman’s single-factor test was conducted on the entire dataset, revealing that the factor accounted for 22.7% of the variance, below the 50% threshold [91]. This result indicates that CMV was not a major issue. The collected data were also used to assess the degree of collinearity in the exploratory survey items. A variance inflation factor (VIF) exceeding 3.3 signifies high collinearity and represents a scenario potentially compromised by Common Method Bias (CMB) [95]. Therefore, if all VIFs derived from a thorough collinearity test within the internal model remain equal to or below 3.3, the model can be considered free of CMB [96]. As all VIFs associated with the latent constructs in our model are below 3.3, it is evident that our model is not affected by CMB. Table 2 displays the variance inflation factor.

4.3. Structural Model Assessment

An architectural method was used to verify the importance of the proposed paths and the model’s overall accuracy in the prediction. A structural model analysis was conducted to determine the degree to which the data fitted the structural model fit (SMF). Importance of variance explained (R₂), standardized root mean square residual (SRMR), and Normed Fit Index (NFI) The analyses of the conceptual framework and hypotheses were conducted for assessments of the components of the R², SRMR, and NFI of the conceptual framework and hypotheses. The path coefficients’ standard error and t-statistics were calculated with a bootstrapping approach (n = 579, sample =10,000). A significance level of 0.05 was used to determine critical variable values. The model fit measures reported in

Table 6. Structural model predictive and explanatory power.

Construct	R-Square	Q2 Predict	RMSE	MAE	SRMR	GFI	NFI
ATT	0.894	0.890	0.334	0.271	0.023	0.918	0.942
DDVI	0.907	0.887	0.337	0.269
PEU	0.894	0.894	0.327	0.256
PU	0.905	0.895	0.325	0.263

include NFI, SRMR, and GFI. Further support for the good fit of the measurement model came from the SRMR result of 0.023, which is below the threshold (<0.08, [82]). Additionally, NFI was 0.942, which showed a good fit since values above 0.80 are preferable [83,97]. The results demonstrate that the R² of the framework is 0.907, attitude is 0.894, PEU is 0.894, and PU is 0.905; values of the consumers’ adoption intention of the driverless delivery vehicles are important [98]. A rule of thumb also proposed by Hair et al. (2021) [99] and R² coefficient values of 0.19, 0.33, and 0.67 are indicative of weak, moderate, and robust explanations, respectively. Results from the present study show that all R² values for the variables could be described as having large (robust), moderate, and weak effects. The prediction capability of the model was then evaluated with the PLSpredict approach with 10 folds and 10 repetitions [100]. As shown in the assessment results in Table 6, all the different GFI are satisfied. Conducted up to October 2023, this result validation results in robust and reliable findings, ensuring that policymakers and stakeholders can base their conclusions on the insights from this study.

This research investigated structural models for collinearity. All VIF scores were below three, signifying that collinearity did not distort the estimates [87,101]. Then it analyzed the importance and relevance of the links within the structural model.

Table 7. Model estimates.

Direct Effects	Path Coefficients	p Values	Confidence Intervals		f-Square	Decision
			2.5%	97.5%
ATT → DDVI	0.117	0.004	0.038	0.196	0.013
PEU → ATT	0.475	0.000	0.399	0.552	0.303
PEU → DDVI	0.243	0.000	0.166	0.323	0.057
PEU → PU	0.305	0.000	0.228	0.382	0.104
PU → ATT	0.489	0.000	0.412	0.563	0.321
PU → DDVI	0.212	0.000	0.126	0.295	0.043
TECOM → DDVI	0.214	0.000	0.127	0.302	0.044
TECOM → PEU	0.532	0.000	0.463	0.599	0.378
TECOM → PU	0.266	0.000	0.188	0.347	0.077
TST → DDVI	0.196	0.000	0.114	0.279	0.038
TST → PEU	0.432	0.000	0.363	0.501	0.249
TST → PU	0.404	0.000	0.326	0.478	0.196

shows the PLS-SEM estimation results.

The hypothesized relationships were validated as significant in the conceptual framework findings, as all the TAM constructs significantly influenced consumers’ adoption intention of the driverless delivery vehicles, thereby supporting Hypotheses H1, H2, and H3. PEU and PU had a significant effect on consumers’ attitudes toward the use of driverless delivery vehicles. These results corroborate Hypotheses H1a and H2b. The results show that TST and TECOM had a significant impact on consumers’ intention to adopt the driverless delivery vehicle, supporting Hypotheses H5 and H6. The findings also indicated that perceived trust and technological complexity were more important than PU and PEU. These outcomes provide additional support for Hypotheses H5a, H5b, H6a, and H6b.

A mediation study was executed to estimate the link between TECOM, TST, PEU, PU, and consumers’ adoption intention of driverless delivery vehicles via ATT, PEU, and PU (see

Table 8. Mediation analysis.

	Paths	p Values	CI		Decision
			2.5%	97.5%
TECOM → PU → DDVI	0.056	0.000	0.031	0.086	Partial mediation
TST → PU → DDVI	0.085	0.000	0.048	0.125	Partial mediation
PEU → ATT → DDVI	0.055	0.005	0.018	0.095	Partial mediation
TECOM → PEU → DDVI	0.129	0.000	0.085	0.177	Partial mediation
PU → ATT → DDVI	0.057	0.006	0.019	0.100	Partial mediation
TST → PEU → DDVI	0.105	0.000	0.070	0.143	Partial mediation
PEU → PU → DDVI	0.064	0.000	0.035	0.097	Partial mediation

). As expected, technological complexity had a significant indirect influence on consumers’ adoption intention of driverless delivery vehicles via PU and PEU. Perceived trust had a significant indirect influence on consumers’ DDVI via PU and PEU. PEU had a significant indirect influence on consumers’ DDVI via PU and ATT. Finally, PU had a significant indirect influence on consumers’ DDVI via attitude.

The PLS-SEM latent variable analysis results were employed to conduct the NCA. Descriptive indicators for this analysis are stipulated in

Table 9. NCA effect sizes.

	CE-FDH			CR-FDH
Latent Variable	Original Effect Size	95.0%	p Value	Original Effect Size	95.0%	p Value
ATT	0.412	0.042	0.000	0.284	0.031	0.000
PEU	0.394	0.044	0.000	0.305	0.035	0.000
PU	0.391	0.045	0.000	0.286	0.036	0.000
TECOM	0.354	0.031	0.000	0.257	0.028	0.000
TST	0.394	0.037	0.000	0.266	0.027	0.000

. Since CE-FDH was utilized in the analysis, its accuracy was automatically set at 100%. The effect sizes d, and their significance are detailed in Table 7. According to practical standards, an effect size d greater than 0.1 is considered the threshold for identifying necessary conditions [19,22]. Ref. [22] proposed that effect sizes falling between 0.1 and 0.3 may be considered medium-sized effects, those between 0.3 and 0.5 may be considered large effects, and those greater than 0.5 may be considered very large effects. The findings indicated that ATT, PEU, PU, TECOM, and TST were significant NCA (p < 0.05) for consumers’ adoption intention of driverless delivery vehicles. Therefore, all the hypotheses were fully supported.

The bottleneck table offers a comprehensive overview of the ceiling lines, highlighting the specific values of conditions crucial for attaining the desired outcomes (refer to

Table 10. Bottleneck tables for adoption intention of driverless delivery vehicles.

Bottleneck Tables	ATT	PEU	PU	TECOM	TST
0.000%	NN	NN	NN	NN	NN
10.000%	NN	NN	NN	NN	NN
20.000%	−2.432	−2.287	NN	NN	NN
30.000%	−2.120	−1.986	−2.258	−1.997	−2.192
40.000%	−1.809	−1.686	−1.898	−1.681	−1.852
50.000%	−1.498	−1.386	−1.537	−1.365	−1.511
60.000%	−1.186	−1.085	−1.177	−1.048	−1.171
70.000%	−0.875	−0.785	−0.817	−0.732	−0.830
80.000%	−0.564	−0.484	−0.456	−0.415	−0.489
90.000%	−0.252	−0.184	−0.096	−0.099	−0.149
100.000%	0.059	0.116	0.265	0.217	0.192

). To assess these outcomes, we followed the methodology suggested by ref. [102], utilizing the 30th percentile as the cut-off point between low and high outcome levels. Consequently, high levels of consumers’ adoption intention of driverless delivery vehicles can be achieved with values of attitude of at least 2.120. Perceived ease of use needs to have scores of at least 1.986, perceived usefulness needs to have scores of at least 2.258, perceived trust needs to have scores of at least 2.192, and technological complexity needs to have scores of at least 1.997, respectively.

The results from both Necessary Condition Analysis (NCA) and Partial Least Squares Structural Equation Modeling (PLS-SEM) provide valuable insights into the factors influencing consumers’ adoption intention of driverless delivery vehicles (See

Table 11. Summary of findings for adoption intention of driverless delivery vehicles.

Construct	NCA Results	PLS-SEM Results
ATT	Necessary condition that is significant and relevant	Significant determinant
PEU	Necessary condition that is significant and relevant	Significant determinant
PU	Necessary condition that is significant and relevant	Significant determinant
TECOM	Necessary condition that is significant and relevant	Significant determinant
TST	Necessary condition that is significant and relevant	Significant determinant

). According to the NCA results, constructs such as attitude (ATT), perceived ease of use (PEU), perceived usefulness (PU), technological complexity (TECOM), and perceived trust (TST) are identified as necessary conditions for adoption. These conditions are significant and relevant, indicating that they must be met for consumers to consider adopting driverless delivery vehicles. For instance, a positive attitude towards the technology is crucial, as it sets a foundational requirement for adoption. Similarly, consumers must perceive the vehicles as easy to use, useful, and trustworthy while also finding the technology manageable in complexity. These conditions highlight the fundamental aspects that must be addressed to facilitate consumer acceptance.

The PLS-SEM results further reinforce the importance of these constructs by identifying them as significant determinants of adoption intention. A positive attitude serves as a necessary condition and strongly influences consumers’ willingness to adopt driverless delivery vehicles. The ease of use and perceived usefulness of the technology are critical factors that drive adoption, as consumers are more likely to embrace a service, they find beneficial and easy to interact with. Technological complexity and perceived trust also play vital roles in shaping adoption intention. Reducing complexity and building trust through transparent communication and reliable service can significantly enhance consumer confidence in driverless delivery vehicles. These findings underscore the multifaceted nature of consumer adoption behavior and highlight the need for a comprehensive approach to address both necessary conditions and significant determinants.

5. Implications and Conclusions

5.1. Discussion

To better understand how accurately the proposed model predicts consumers’ intentions to adopt autonomous delivery vehicles, a mix-method of the PLS-SEM and the NCA was performed. Ref. [19] propose that improving the quality of PLS-SEM and NCA would facilitate theorizing and theory testing with these methods. Requisite conditions have been presented in theoretical reasoning not only in operations management [103], but also in a range of areas from marketing [104] to information systems research [105]. This study further broadens TAM-based frameworks by considering perceived trust and the complexity of technology. Utilizing PLS-SEM along with NCA provided a means to delve into five sets of exogenous constructs, known as attitude, technological complexity, PEU, perceived trust, and PU, marking them as pivotal and required agents. The PLS-SEM outcomes affirm that the constructs significantly affect DDVI with PU, attitude, and PEU as mediators. As a result, improvement in every one of these constructs contributes to an increased target of adoption intention. The NCA results, on the other hand, indicate that these three constructs serve as fundamental prerequisites for adoption intention. Bottleneck analysis showed that high adoption intention can be obtained with relatively low attitude, perceived usefulness, and technological complexity. The findings suggest that consumers are more likely to adopt autonomous delivery vehicles when they perceive them as environmentally friendly, highlighting the importance of sustainability in driving consumer acceptance.

5.2. Theoretical Implications

This research contributes significantly to the body of literature. This study fills two important research gaps by examining consumer adoption intention of driverless delivery vehicles. The prior studies mainly conducted explanatory analyses rather than aiming to predict intentions and behaviors [106,107]. Using the PLS-SEM toolbox, we performed a causal predictive analysis of the model [17,87] that allowed us to make solid conclusions regarding the ability of the model to predict DDVI. This research specifically determined the consumer intention to adopt driverless delivery vehicles, as they are a key driver for increased efficiency, reduced costs, and less traffic congestion in the future [6]. Consequently, our study presents an improvement to existing knowledge, which has mainly focused on the adoption period and cross-sectional representations of such actions [108]. Existing research has not yielded actionable insights for predicting the adoption intention of driverless delivery vehicles.

Theoretical Contributions to Literature Finally, based on the available literature on the antecedents of DDVI, this study makes further contributions by bringing together and expanding the TAM with technological complexity and perceived trust. The Technology Acceptance Model is a common model of technology use based only on the outer factors (PEU, attitude, and PU). This neglects the technological complexity and perceived trust that shape concerns around the reliability, safety, and usability of, as well as their desire to feel safe and secure, ever increasingly a factor in the introduction of driverless delivery vehicles [69] (more on the factors of adoption here and here). According to PLS-SEM and NCA, our findings suggest that perceived trust and technological complexity are significant predictors and may be regarded as necessary for consumers’ intention to adopt driverless delivery vehicles. Furthermore, integrating technological complexity and perceived trust as an additional layer to the TAM proposed in our study would be relevant for predicting other responsible behaviors and operational capabilities, such as the technology’s safety, reliability, and performance [55,57].

5.3. Methodological Implications

For predictive model assessment, the PLS-SEM toolbox was applied in this study [17,109]. An exploration of the antecedents proposed with Necessary Condition Analysis (NCA) combined with PLS-Settings allowed determining both the necessity and sufficiency of the identified drivers [19,107,110]. NCA should be used particularly in cases where the intention is to extend an existing model, such as the TAM, with other antecedents. The exploration of these new antecedents, as both critical and superlative factors, becomes apparent from the joint analysis of NCA and PLS-SEM outputs [107]. However, in line with recent literature [22,89,103], we advise cautious interpretation of NCA results. More specifically, bottleneck tables may show that, for some required antecedents, even when their impacts are statistically significant, only small amounts are needed to produce the desired effects (e.g., a study on technology acceptance). Therefore, it is essential that the explanation of these outcomes reflects the actual clinical relevance of the conditions examined.

5.4. Managerial Implications

To foster consumer adoption of driverless delivery vehicles, companies must prioritize the development of user-friendly interfaces and processes. Intuitive designs, clear instructions, and streamlined delivery experiences can significantly boost perceived ease of use (PEU). Providing comprehensive training and support, such as tutorials and customer service, can further alleviate initial anxieties and enhance user confidence. By simplifying interactions and reducing cognitive load, businesses can create a more seamless and accessible experience for consumers.

A positive attitude towards driverless delivery vehicles is crucial for their acceptance. Companies should leverage marketing campaigns that highlight the benefits of convenience, speed, and environmental sustainability. Engaging with the community through pilot programs and events can also help consumers experience the technology firsthand, thereby changing negative perceptions. Additionally, emphasizing safety features and obtaining third-party endorsements can build consumer confidence and foster a positive attitude.

Trust is a fundamental factor influencing consumer adoption. Companies must maintain transparency about the technology, safety measures, and data privacy policies. Regular updates and clear communication can help build consumer confidence. Obtaining certifications and endorsements from reputable organizations, as well as addressing customer feedback promptly, can further enhance perceived trust and credibility.

Technological complexity can be a significant barrier to adoption. Companies should focus on simplifying the underlying technology to make it more accessible and less intimidating. Providing educational materials and workshops to help consumers understand the technology can also reduce perceived complexity. Gradual introduction of features, starting with simpler functionalities, can help consumers become more comfortable with the technology over time.

Highlighting the practical benefits of driverless delivery vehicles, such as faster delivery times and cost savings, can enhance perceived usefulness. Offering customization options and seamless integration with existing e-commerce platforms can further improve the value proposition. By demonstrating the tangible benefits and convenience of the technology, companies can encourage greater consumer adoption.

5.5. Policy Implications

Governments should establish clear safety standards and regulations to ensure the reliability and security of driverless delivery vehicles. Enforcing strict data privacy laws and requiring rigorous testing and certification processes can build consumer trust. By setting high standards for safety, data protection, and transparency, policymakers can create a conducive environment for the adoption of this technology.

To accelerate the adoption of driverless delivery vehicles, policymakers should provide tax incentives, subsidies, and grants for research and development. Investing in infrastructure, such as dedicated lanes and charging stations, can also facilitate the widespread use of the technology. By offering financial support and improving the operational environment, governments can encourage both businesses and consumers to embrace driverless delivery vehicles.

Public campaigns and educational initiatives are essential for promoting the benefits and safety of driverless delivery vehicles. Incorporating information about autonomous vehicles into school curricula and community programs can help familiarize the public with the technology. By working closely with industry stakeholders to develop accurate and relevant educational materials, policymakers can reduce misconceptions and build a more informed consumer base.

Policymakers should develop guidelines to simplify the user interface and experience of driverless delivery vehicles. Regulating technology providers to promote ease of use and transparency can help reduce complexity. Supporting innovation in user interface design and user experience can further enhance the accessibility of the technology for a broader range of consumers.

Integrating driverless delivery vehicles with public services and e-commerce platforms can highlight their practical benefits. Policymakers should provide regulatory support to facilitate seamless integration and encourage the development of feedback mechanisms to monitor effectiveness. By emphasizing the tangible benefits and ensuring compatibility with existing systems, policymakers can enhance the perceived usefulness of driverless delivery vehicles.

5.6. Limitations and Future Examination

Although the shortcomings of this research are duly noted, the limitations presented in this section may inspire future research initiatives. Additional studies are recommended to broaden our study to other seats and to improve our understanding of the larger lessons of our results [111]. Moreover, the model presented in the current study could be utilized to evaluate consumers’ willingness to adopt driverless delivery cars and to promote green behavior. This study generally highlighted the advantages of the current PLS-SEM methodological improvements on predicting outcomes while neglecting the heterogeneity of data structure. This study also has limitations that future studies may overcome through moderator or multigroup analyses. Additionally, it is important to clarify that our analysis spans a narrow time range, from the end of November 2023 until mid-January 2024. Accordingly, we recommend using a longitudinal study to explore the antecedents of DDVI across a longer period of time. Moreover, future research can further refine our work by extending different PLS-SEM methods to PLS-SEM and machine learning, including but not limited to cIPMA [103] and IPMA [20]. We noted that while NCA effectively identifies necessary conditions, it does not address sufficiency or non-binary conditions. For PLS-SEM, we highlighted its assumptions of linearity and potential susceptibility to common method bias. Future research could explore these limitations using additional methods or data sources to validate our findings. The TAM framework does not account for the changing perceptions of risk associated with new technologies, such as potential safety concerns or data privacy issues. This oversight could limit the study’s ability to fully capture the complexities of consumer adoption intentions for DDVs. Future research should consider incorporating additional constructs that address dynamic risk factors to provide a more comprehensive understanding of technology acceptance in emerging contexts. This study uses a cross-sectional design, limiting our ability to establish causality or track changes over time. The findings reflect a specific period and may not capture long-term trends. Future research should employ longitudinal designs to better understand how user attitudes and behaviors evolve over time, providing deeper insights into the dynamics of autonomous vehicle adoption. The study focuses on a specific region and demographic, which may limit the generalizability of our findings. Future research could address this limitation by conducting cross-cultural studies to examine how cultural differences influence the adoption of autonomous delivery vehicles. Additionally, the study captures a snapshot of consumer perceptions at a particular point in time. Longitudinal research would be valuable in tracking changes in consumer attitudes and behaviors as these technologies evolve and become more prevalent. The current study focuses on quantitative analysis. Future research should consider employing a mixed-methods approach, combining quantitative data with qualitative insights from interviews or focus groups. This triangulation would provide a more comprehensive understanding of the factors influencing the acceptance and use of driverless delivery vehicles.