Ambiguous Driving Scenarios Survey

Abstract

The Ambiguous Driving Scenarios Survey (ADSS), developed by Baby et al. (2024), is a specialized psychometric instrument designed to classify and evaluate ambiguous driving scenarios (ADS) within the context of Autonomous Vehicles (AVs) based on user perceptions. ADS are defined as complex situations where AVs struggle to accurately interpret the intentions or behaviors of other road users due to inherent uncertainty or perceptual confusion.

The development of the ADSS involved a rigorous, multi-stage refinement process. Initial research, encompassing extensive literature reviews and expert interviews, identified over 10,000 potential driving scenarios, which were subsequently filtered down to 6,590 through structured content analysis. Utilizing the user experience research technique known as card sorting, the researchers further reduced the item pool to 28 representative and distinct scenarios. The final survey instrument, validated using data collected from adult drivers in the Republic of Korea, confirmed a robust five-factor model through Confirmatory Factor Analysis (CFA). Comprehensive psychometric testing established high levels of reliability (internal consistency and split-half) and validity (convergent validity and discriminant validity).

Keywords

Automated Driving Systems, Autonomous Vehicle Drivers, Ethical Perception, Legal Perception, Moral Perception, Safety Perception, Utility Perception, Human Factors Engineering, Transportation Safety, Human-Technology Interaction.

Authors

Baby, Tiju, Ippoliti, Hatice Şahin, Wintersberger, Philipp, Zhang, Yiqi, Yoon, Sol Hee, Lee, Jieun, Lee, Seul Chan.

Purpose

The primary purpose of the ADSS is to systematically classify ambiguous driving scenarios based on how users perceive the importance of various factors when an AV responds to such a critical situation. By quantifying user expectations across moral, ethical, legal, utility, and safety dimensions, the survey provides a critical link between human judgment and automated decision-making.

This classification serves as a vital resource for multiple stakeholders. For policymakers, it informs the development of regulatory frameworks and safety guidelines. For researchers, it provides a standardized tool for studying human factors in automation. For AV manufacturers, it assists in the design and refinement of driving styles and operational algorithms that align with public perception and enhance trust in automated driving systems.

Construct

The scale measures the overarching construct of Ambiguous Driving Scenarios Perception, specifically focusing on how adult drivers evaluate the necessary priorities for Autonomous Vehicles when facing uncertain or confusing road events. The construct is operationalized through five empirically distinct perception factors:

• Moral Perception: Assessing the importance of right-versus-wrong outcomes in dilemma scenarios.
• Ethical Perception: Evaluating the necessity of adhering to established societal principles.
• Utility Perception: Measuring the importance of efficiency, convenience, and overall usefulness of the AV action.
• Legal Perception: Determining the perceived importance of complying with existing traffic laws and regulations.
• Safety Perception: Evaluating the perceived importance of minimizing physical harm or risk to road users.

Validity

The psychometric validation of the ADSS demonstrated robust evidence for both convergent and discriminant validity, supporting the integrity of the five-factor structure.

Convergent Validity: Convergent validity was established by confirming that items intended to measure the same construct were highly correlated. Factor loadings for all items exceeded 0.6. Furthermore, the Average Variance Extracted (AVE) for all five factors was found to be above the recommended threshold of 0.5 (Krishnan & Ramasamy, 2011), confirming adequate shared variance among indicators within each construct.

Discriminant Validity: Strong Discriminant Validity was confirmed by comparing the AVE of each construct against its correlations with other constructs. Specifically, the square root of the AVE for every factor exceeded the corresponding cross-correlations between that construct and all other factors in the model. This finding ensures that each perception factor (Moral, Ethical, Utility, Legal, Safety) measures a unique aspect of the overall construct.

Reliability

Reliability analyses confirmed the internal consistency and stability of the ADSS across its five core factors.

Internal Consistency: Internal consistency was assessed using Composite Reliability (CR) measures. The CR values ranged from 0.800 to 0.851 across all factors, indicating high internal consistency and suggesting that the multiple items used to measure each latent construct are highly reliable.

Split-Half Reliability: The stability of the instrument was further supported by the split-half reliability scores, which exceeded 0.812 for every factor. These high coefficients confirm the overall dependability of the survey for measuring perceptions related to Ambiguous Driving Scenarios.

Factor Analysis

The latent factor structure of the ADSS was derived using a rigorous sequence of psychometric modeling. Initially, Exploratory Factor Analysis (EFA) was utilized to test the preliminary model fit and identify the underlying dimensions of user perception, confirming the five-factor solution.

Subsequently, Confirmatory Factor Analysis (CFA) was employed to validate the proposed model structure. The CFA supported the refined five-factor model with substantially improved fit indices. A final, simplified model was achieved by addressing high modification indices and removing five items that demonstrated psychometric issues (factor loadings below 0.30). The resulting 28 scenarios are categorized as follows based on their primary factor loading:

• Moral Perception: Scenarios 5, 11, 20, 22.
• Ethical Perception: Scenarios 1, 4, 17, 23.
• Legal Perception: Scenarios 5, 11, 12, 26.
• Utility Perception: Scenarios 4, 7, 13, 17, 26.
• Safety Perception: Includes all scenarios except 1, 4, 5, 7, 13, 20, 22, and 23.

Instrument

Test Type: Original Inventory/Questionnaire

Format: The survey is administered electronically and comprises three distinct sections. The first section gathers information on user perceptions related to AVs (moral, ethical, legal, utility, and safety). The second section collects general demographic information. The third and main section presents the 28 core Ambiguous Driving Scenarios. Each scenario includes a pictorial representation, a descriptive text, and five corresponding questions designed to assess the user’s perception of importance regarding the AV’s response, rated on a five-point Likert scale (1 = “Very important” to 5 = “Not at all important”). The total number of items assessed is 28 scenarios, each generating 5 data points (140 total responses).

Language Available: English (Primary publication language).

Population Group: Human; Male; Female

Age Group: Adulthood (18+ years), with specific data collected across Young Adulthood (18-29 years), Thirties (30-39 years), and Middle Age (40-64 years).

Population Details: Respondents were Autonomous Vehicle Users residing in the Republic of Korea, with an age range spanning from 20 to 49 years.

Test Methodology: The development and validation utilized comprehensive psychometric methods, including Test Validity, Convergent Validity, Discriminant Validity, Test Reliability, Internal Consistency, Split-Half Reliability, Factor Analysis, Confirmatory Factor Analysis, and Exploratory Factor Analysis.

Keywords

Drivers, Highway Safety, Human Factors Measures, Human-Computer Interaction Measures, Human-Technology Interaction, Autonomous Vehicles, Ethical Decision Making, Card Sorting, Psychometrics.

Authors

Author ORCID Identifier:

• Baby, Tiju: ORCID
• Wintersberger, Philipp: ORCID

Affiliation Email addresses:

• Baby, Tiju – Division of Media, Culture, and Design Technology, Hanyang University ERICA
• Ippoliti, Hatice Şahin – Department of Computing Science, University of Oldenburg
• Wintersberger, Philipp – Digital Media Department, University of Applied Sciences Upper Austria
• Zhang, Yiqi – Department of Industrial and Manufacturing Engineering, Pennsylvania State University
• Yoon, Sol Hee – Department of Safety Engineering, Seoul National University of Science and Technology
• Lee, Jieun – Department of Safety Engineering, Pukyong National University
• Lee, Seul Chan – Division of Media, Culture, and Design Technology, Hanyang University ERICA

Correspondence Address: Lee, Seul Chan: Hanyang University ERICA, Lion’s Hall 202, 55, Hanyangdaehak-ro, Sangnok-gu, Gyeonggi-do, Ansan-si, Korea, Republic of. Email: [email protected]

Permissions & Fee and Test Year

Test Year: 2024

Test Items Availability: The complete test items, including pictorial representations and scenario descriptions for the 28 Ambiguous Driving Scenarios, are proprietary and not publicly available. Interested parties must contact the corresponding author, Lee, Seul Chan, or the publisher to obtain permission and access to the instrument.

Reference’s

Baby, T., Ippoliti, H. Ş., Wintersberger, P., Zhang, Y., Yoon, S. H., Lee, J., & Lee, S. C. (2024). Development and classification of autonomous vehicle’s ambiguous driving scenario. Accident Analysis and Prevention, 200, 1–20. DOI: 10.1016/j.aap.2024.107501

Krishnan, A., & Ramasamy, A. (2011). Reference cited in the original article for defining acceptable thresholds for AVE in convergent validity assessment.

Items of the Ambiguous Driving Scenarios Survey

IMPORTANT: The following scale items must be preserved in their original language and must not be changed in any way.

The specific scenario descriptions and the five related survey questions for each of the 28 ambiguous driving scenarios were not included in the provided source table. Access to the full instrument requires contacting the corresponding author or publisher.