Advanced driver assistance systems (ADAS) with increasing automation maturity and availability in urban contexts are entering the market. Meanwhile, the situational context has been identified to play a crucial role in system comprehension and usage, yet its subcomponents and their relation to system comprehension remain an open research question. To gain insights in the role of the situation complexity regarding subjective system comprehension and different methodological aspects, this study applies a mixed quantitative and qualitative approach, focusing on signaled intersections as an exemplary scenario. An on-road study with forty-six participants was conducted, involving six traffic light scenarios (all experienced twice). Results indicate that while comprehension was generally high, the situational context, including environmental and traffic-related factors, affected subjective system understanding. The proposed approach sheds light on the role of mixed methods in ADAS research, which may provide insights for system developers and suggestions for user training content.
We demonstrate a web app user interface that simulates the benefits and trade-offs of Vehicle-to-Grid (V2G) given an individual’s electric vehicle driving habits. Users input their own driving and charging data (or explore example usage patterns), specify their battery type, and indicate their availability for V2G. The interface then generates a personalized V2G schedule and presents both the potential financial benefits (e.g., earnings from supplying energy to the grid) and impact on battery health. This interactive and personalized interface aims to mitigate known barriers to V2G participation and increase consumer interest in the technology.
Scenarios provide a fundamental link between driving simulators and real-world conditions, shaping the extent to which the findings of a user study can be applied to public roads. However, compared to other aspects of study design, scenario development in human-vehicle interaction research tends to receive less deliberate attention. To encourage more methodical scenario generation, this work introduces a mixed methods approach for extracting representative scenarios from an integration of three real-world data sources: aggregated crash statistics, interviews with experienced drivers, and naturalistic driving data. Through a case study on winter driving, we outline the derivation of a nighttime, two-lane road scenario from these data sources and conduct an initial driving simulator pilot study to assess its realism. We hope that this demonstration of scenario generation from quantitative and qualitative data inspires researchers to consider more rigorous methods for scenario design in future work.
Affording users the opportunity to customize automated vehicle (AV) behavior may meaningfully improve trust and acceptance. We examined this in a driving simulator study with 49 students and 18 community members (N = 67), randomly assigned to customization or non-customization conditions. Customization participants selected from a range of behavior options and experienced one of three driving styles. Results showed that the effect of customization on trust depended on the driving style. Specifically, those who customized and received the conservative style reported significantly higher trust than those who passively received it. Exploratory analyses using the Unified Theory of Acceptance and Use of Technology (UTAUT) revealed that AV interest explained 8% additional variance in behavioral intentions beyond UTAUT attitude variables. These findings support prior work emphasizing trust in AV adoption and extend it by positioning trust as a potential distal predictor of behavioral intentions, acting through performance and effort expectancy.
Modern vehicle infotainment systems are increasingly multimodal, enabling drivers to interact through voice, touch, and gesture. While these input methods enhance usability, most existing tools overlook the dynamic driving context-such as traffic density, road complexity, and cognitive load, leading to incomplete evaluations. This work-in-progress introduces the Context-Aware Usability Scale (CAUS), a psychometric instrument for evaluating multimodal in-vehicle interfaces with sensitivity to environmental factors. CAUS is developed through a three-phase process: (1) item generation from a systematic literature review using the PRISMA framework, (2) expert validation via Content Validity Index (CVI) methods, and (3) planned empirical testing in a high-fidelity driving simulator. The 20-item scale spans five core usability dimensions - efficiency, effectiveness, cognitive load, distraction, and satisfaction - and five context-sensitive items. Expert evaluations confirmed strong content validity (S-CVI > 0.90). The final phase will assess reliability and factor structure. CAUS offers an ecologically valid, scalable tool for evaluating automotive HCI, with applications in smart mobility and mobile interface design.
Anomaly detection can enhance autonomous driving safety through identifying rare events and potentially dangerous deviations from expected gaze distributions. We propose a statistical anomaly detection model that processes temporal gaze dynamics and creates linear inference to contextual features. The model is trained using a reconstruction technique to familiarize it with normal data, enabling it to detect anomalous gaze patterns. The reconstruction-based approach allows the model to learn what constitutes typical gaze behaviour in a given context. Our research on both Level 3 (L3) modes demonstrates that drivers usually change their gaze direction roughly after a couple of seconds, presumably as part of their regular safety checks of the surrounding environment, or because their attention was drawn by noticeable objects. The findings support the proposition that the driver’s gaze shifts toward prioritizing central-road information over peripheral cues and tends to suppress physical distractions by maintaining a more fixed gaze and focusing straight ahead.
This study evaluates the effectiveness of large language model-based personas for assessing external Human-Machine Interfaces (eHMIs) in automated vehicles. 13 different models namely BakLLaVA, ChatGPT-4o, DeepSeek-VL2-Tiny, Gemma3:12B, Gemma3:27B, Granite Vision 3.2, LLaMA 3.2 Vision, LLaVA-13B, LLaVA-34B, LLaVA-LLaMA-3, LLaVA-Phi3, MiniCPM-V and Moondream were tasked with simulating pedestrian decision making for 227 vehicle images equipped with eHMI. Confidence scores (0-100) were collected under two conditions: no memory (images independently assessed) and memory-enabled (conversation history preserved), each in 15 independent trials. The model outputs were compared with the ratings of 1,438 human participants. Gemma3:27B achieved the highest correlation with humans without memory (r = 0.85), while ChatGPT-4o performed best with memory (r = 0.81). DeepSeek-VL2-Tiny and BakLLaVA showed little sensitivity to context, and LLaVA-LLaMA-3, LLaVA-Phi3, LLaVA-13B and Moondream consistently produced limited-range output.
As autonomous vehicles (AVs) become increasingly integral into traffic environments, external human-machine interfaces (eHMIs) have emerged as a critical topic in facilitating safe interactions with vulnerable road users (VRUs), specifically children, elderly, and persons with disabilities (PwDs). While recent studies have explored eHMI concepts designed to address the needs of these demographics, much of the existing research centered on the interface solutions with less emphasis placed on the underlying methodologies used to evaluate them. As a result, the methodological approaches used to evaluate such interfaces remaining varied and underexamined. Rather than focusing on interface design outcomes, this paper presents a systematic review of 19 studies with the attention shifted to the methodologies used, focusing on study methodologies, evaluation tools, and participant engagement practices. The findings revealed notable inconsistencies in evaluation methods, limited adoption of inclusive design practices, and a lack of standardized protocols. By highlighting these gaps, this body of work provides actionable suggestions to guide future research and promote more standardized practices in the development of eHMIs for inclusive evaluation approaches for future eHMI research.
The emergence of large language models has introduced new opportunities in software development, particularly through a revolutionary paradigm known as vibe coding or “coding by vibes”, in which developers express their software ideas in natural language and where the LLM generates the code. This paper investigates the potential of vibe coding to support novice programmers. The first author, without coding experience, attempted to create a 3D driving simulator using the Cursor platform and Three.js. The iterative prompting process improved the simulation’s functionality and visual quality. The results indicated that LLM can reduce barriers to creative development and expand access to computational tools. However, challenges remain: prompts often required refinements, output code can be logically flawed, and debugging demanded a foundational understanding of programming concepts. These findings highlight that while vibe coding increases accessibility, it does not completely eliminate the need for technical reasoning and understanding prompt engineering.
Artificial compensation of visual feedback can improve a driver’s steering maneuvers in degraded visual conditions. This paper examines whether this benefit takes effect in distracted or drowsy driving. We also investigate whether the continuous presentation of artificial feedback over a long period causes a “learning effect,” in which the effect of such artificial feedback increases or decreases, or a “dependency,” in which driving behavior deteriorates after the artificial feedback is removed. In addition, we examine the effect of the strength of artificial feedback using two different feedback luminance levels. Simulated driving experiments, consisting of over one hour of total driving, show the persistence of the artificial feedback effect but do not support the emergence of any “learning effect” and “dependency.” We also confirm that this effectiveness differs depending on feedback strength, suggesting the importance of proper control of feedback luminance.
Urban Air Mobility (UAM) is expected to operate without onboard pilots, raising concerns about passenger trust in automation. Building on findings from autonomous vehicle (AV) research, this study investigates whether explanation-based human-machine interfaces (HMIs) can enhance trust in UAM. We developed a virtual simulation with five conditions, varying the type (how vs. how+why) and timing (pre-event vs. post-event) of system-generated explanations. A baseline condition with no explanation was also included. Participants will experience multiple UAM issue events (e.g., fog, gusts), during which trust, transparency, and competence are measured. We anticipate that both how and why explanations will increase trust, with why and pre-event explanations expected to have stronger effects. Our findings will inform the design of explainable UAM systems that foster psychological comfort and acceptance.
External human-machine interfaces (eHMIs) are designed to explicitly communicate autonomous vehicles’ (AVs) intentions, thereby enhancing safety in complex traffic interactions. This study evaluated the effectiveness of a dynamic text-based eHMI on an autonomous shuttle operating in a naturalistic setting at unsignalized crosswalks in South Korea. Through field observations, we identified scenarios in which traditional yielding or stopping messages were insufficient, especially under conditions of continuous pedestrian flow causing vehicle delays. Using this scenario, explicitly communicating the vehicle's imminent departure, was tested against a static control condition. Post-interaction surveys of 60 pedestrians revealed that the dynamic eHMI significantly improved message visibility, comprehension, and perceived system support. Additionally, pedestrians exposed to the dynamic eHMI prioritized explicit textual cues over implicit vehicle cues when deciding to cross, leading to increased trust in AV technology. These results highlight the practical value of context-sensitive, explicit eHMIs for enhancing real-world AV-pedestrian interactions.
Deciding whether to allow an automated vehicle (AV) to merge in front can present a complex negotiation for human drivers. To address this, we explored the human-machine interface (HMI) design for merge negotiation between a manually driven vehicle and an AV from the human driver’s perspective. We developed five HMI designs, each integrating different combinations of visual cues, haptic alerts, and explicit approve and reject controls. They were evaluated together with two baseline conditions in a video-based driving simulator. The results of Likert-scale ratings indicated that HMI designs with explicit accept and reject controls received higher mean ratings in communication clarity, perceived adequacy, safety perception, trust in AV behaviour, and decision-making efficiency than those without such controls. Open-ended feedback further suggested that these HMIs may foster a stronger sense of control and reduce perceptions of aggressiveness. Based on the findings, we outlined three preliminary design considerations for HMIs that support merge negotiation between AVs and human drivers. This work offers early guidance and sets the stage for future research on integrating diverse modalities and driver inputs (e.g., explicit approve and reject controls) into HMI design for merge negotiation.
As high-level, automated vehicles (AVs) become more present on our roads, resolving ethical, legal, and social implication (ELSI) conflicts in decision-making is a complex challenge. To possibly find solutions to such challenges, this workshop explores how Explainable AI(XAI), and Human-Machine Interfaces (HMIs) can improve transparency and increase trust, particularly in ambiguous situations. We propose a scenario-based workshop that invites participants to reflect on decision-making, expectations for explanations, and possible communication through HMI. Outcomes of this workshop will be the first step to meaningfully add XAI to ensure human-centered decisions of AVs.
This paper investigates how in-car conversational agents can support parents driving with young children—a group prone to distraction and cognitive overload. In a high-stress driving simulator study, 22 participants experienced a navigation assistant using four modalities: visual, verbal, non-verbal auditory, and haptic. Feedback was collected using a modified NASA TLX (DALI) and interviews. Verbal navigation instructions were rated most effective, reducing stress and distraction, while haptic feedback was seen as a useful secondary channel for urgent cues. Non-verbal auditory signals were largely ineffective in noisy environments. Participants favored multimodal combinations that balanced clarity with low cognitive demand. Based on these findings, we offer design recommendations for adaptive, family-aware in-car systems, emphasizing verbal communication, context sensitivity, and user customization.
Distracted driving is a leading cause of road crashes, yet traditional analysis often relies on labour-intensive manual video annotation. This study investigates the use of the Vision-Language Model (VLM) PaliGemma 2 to automatically detect driver distraction. Using video data from the Australian Naturalistic Driving Study (ANDS), which captures the driver's steering wheel, we sampled frames and processed them through PaliGemma 2 to generate textual descriptions of behaviour. The model was fine-tuned by updating only its attention layers, improving recognition of distractions like phone use and adjusting controls while retaining general knowledge. Outputs were standardised to enable systematic analysis. Preliminary results show that VLM-based detection accurately identifies key distraction behaviours, greatly reducing the need for manual labelling. These findings support the use of VLMs in road safety research, driver monitoring, and AI-driven interventions, and mark one of the first applications of large-scale multimodal models to naturalistic driving data.
Modern automotive infotainment systems offer a complex and wide array of controls and features through various interaction methods. However, such complexity can distract the driver from the primary task of driving, increasing response time and posing safety risks to both car occupants and other road users. Additionally, an overwhelming user interface (UI) can significantly diminish usability and the overall user experience. A simplified UI enhances user experience, reduces driver distraction, and improves road safety. Adaptive UIs that recommend preferred infotainment items to the user represent an intelligent UI, potentially enhancing both user experience and traffic safety. Hence, this paper presents a deep learning foundation model to develop a context-aware recommender system for infotainment systems (CARSI). It can be adopted universally across different user interfaces and car brands, providing a versatile solution for modern infotainment systems.
Heavy-duty vehicle operators often suffer from elevated muscle strain when making low-speed or tight turns due to high steering torque demands. This study presents a pilot evaluation of a novel ball-shaped steering mechanism designed to improve usability and reduce muscle stress compared to a conventional steering wheel (CSW). Our prototype maps ball rotations (pitch, roll, yaw) into throttle, braking, and steering commands, simulating a potential drive-by-wire setup. We ran a small pilot comparing two different mappings (roll-based vs. yaw-based) against a conventional steering wheel. Early observations suggest that both ball-based mappings reduce arm-acceleration (as a rough proxy for muscle stress) by approximately 60-70%. However, survey data indicates that participants still preferred the conventional wheel for precise control. We discuss the mechanical design, preliminary pilot data, and planned improvements, such as haptic return springs and more robust EMG measurements, to guide future heavy-vehicle applications.
The rapid advancement of autonomous vehicle (AV) technologies is fundamentally reshaping paradigms of human-vehicle collaboration, raising not only an urgent need for innovative design solutions but also for policies that address corresponding broader tensions in society. To bridge the gap between HCI research and policy making, this workshop will bring together researchers and practitioners in the automotive community to explore AV policy directions through collaborative speculation on the future of AVs. We designed The UnScripted Trip, a card game rooted in fictional narratives of autonomous mobility, to surface tensions around human-vehicle collaboration in future AV scenarios and to provoke critical reflections on design solutions and policy directions. Our goal is to provide an engaging, participatory space and method for automotive researchers, designers, and industry practitioners to collectively explore and shape the future of human-vehicle collaboration and its policy implications.
As fully driverless robotaxi services emerge, understanding user needs under real-world conditions is critical. This study employed the think-aloud method to capture real-time cognitive and emotional responses during users' first ride in a fully driverless robotaxi. Analysis of 30 participants' verbal reports revealed three key user needs: perceived safety, efficiency, and comfort. We found that users' trust can be enhanced by conservative driving behaviors and transparent human-machine interface (HMI) design. Conversely, inconsistencies between user expectations and driving behaviors, potentially stemming from technical limitations, and individual differences, can undermine trust. Further, while conservative driving enhanced perceived safety, it can also reduce efficiency, especially in time-sensitive scenarios. Finally, comfort can be shaped by both driving behaviors and HMI interactivity. These findings highlight the importance of user-adaptive interfaces and context-aware driving strategies to balance perceived safety, efficiency, and comfort, thereby supporting the acceptance and deployment of driverless mobility services.
