The purpose of this research was to understand how an in-vehicle interaction model might impact driver distraction, considering how drivers input information into the vehicle and receive feedback from the vehicle. We conducted a task analysis on five vehicle models to evaluate the demands placed on the driver by the interaction model in that vehicle. Vehicles were selected by targeting different interaction models including designs that were mostly buttons, mostly touchscreen, and combinations thereof. For each vehicle, we collected information about the functions available within the vehicle, how drivers input information into the in-vehicle center console, the types of feedback received from the vehicle, and the steps required to complete each function. We conducted qualitative and quantitative analysis comparing the data from the five vehicles. From this dataset, we explore preliminary insights regarding how interaction models relate to driver demand, which may inform future investigations of in-vehicle information system design.

Human expert interviews are valuable but often limited by slow recruitment, scarce expertise, and network-based sampling, especially for abstract human-automation topics requiring multidisciplinary input. This paper presents a persona-based LLM expert-interview pipeline for structuring expert knowledge before human involvement. We constructed 28 discipline-specific personas and purposively selected seven differentiated “silicon experts” across human factors, UX design, algorithm engineering, accident investigation, regulation, accessibility, and product management. Using drivers’ minimum mental model (MMM) as a demonstration topic, each persona completed a structured interview protocol and returned JSON-formatted responses. The pipeline generated 59 candidate items, producing concrete, traceable, and auditable outputs. Results showed both convergence and persona-specific divergence: shared priorities included driver responsibility, operational design domain, and permitted non-driving-related tasks, while divergent contributions highlighted OTA updates, sensor limitations, and warning perceivability. The pipeline is positioned as a preliminary scoping tool for preparing subsequent human expert validation.

Stress-adaptive AI coaching may improve driving performance and experience by tailoring feedback to a driver’s internal state. This study investigated driver stress, driving performance, multimodal sensing, and temporal stress annotations under time pressure in a driving simulator. Participants completed repeated laps under three time-pressure conditions while physiological signals, vehicle telemetry, post-lap stress ratings, and retrospective temporal stress annotations were collected. Results showed that time pressure increased stress ratings, although substantial overlap remained across conditions. Subjective stress was more strongly associated with driving performance and vehicle-control behavior than task condition alone. Analyses stratified by subjective stress ratings suggested that pupil-related measures and driving variability features differentiate stress levels. Temporal annotations aligned with post-lap ratings, supporting their use as proxy labels for within-lap stress dynamics and future real-time stress modeling. These findings inform future stress-adaptive AI coaching systems and adaptive automotive interfaces.

Passengers are susceptible to motion sickness (MS) when engaging in non-driving-related tasks (NDRTs) in the cabin. Visual information about ego-vehicle motion can alleviate MS, yet it may also compete for cognitive resources needed for NDRTs, degrading NDRT performance. To reveal the potential negative effects of MS mitigation strategies, we designed two types of visual cues, i.e., implicit dynamic optic flow and explicit predictive symbols, and compared their MS-mitigation effects and impact on NDRT performance. An on-road driving experiment was conducted with 10 participants. Using a high-interaction NDRT, we revealed an NDRT performance degradation effect of a certain MS mitigation interface. Although all visual cues helped mitigate MS, explicit predictive cues significantly impaired task performance (p < 0.05), whereas implicit optic flow alone had minimal impact on NDRT performance. The results indicate that the attentional cost of MS mitigation cues should be carefully evaluated in the smart cabin, highlighting the need to balance sensory comfort with productivity in future intelligent cockpits.

As Advanced Driver Assistance Systems (ADAS) become increasingly common in production vehicles, surrounding drivers may increasingly encounter vehicles whose behavior is partly controlled by automation. In China, some ADAS-supported vehicles use an externally visible indicator, commonly referred to as the “small blue light,” to signal that ADAS is active. However, limited naturalistic evidence is available on whether such external ADAS status visibility influences surrounding drivers. This study investigated following drivers’ car-following behavior in naturalistic traffic when the lead vehicle operated with ADAS activated and the small blue light was on or off. Results showed speed-dependent changes under the blue-light-on condition. In particular, following drivers maintained larger time headway when the lead vehicle traveled at higher speeds with the light on. By providing naturalistic evidence from ADAS-supported production vehicles, this study contributes to understanding how external ADAS status visibility may shape following drivers’ longitudinal driving behaviors in real traffic.

We present WidgetGNN, a graph neural network–based approach for interactive-speed generation of adaptive human–machine interfaces (HMIs). Widgets are modeled as interacting nodes that iteratively optimize their spatial arrangement. The system combines continuous coordinate updates with a novel swapper network that enables discrete structural adjustments, improving cluster formation and layout quality. Unlike static layouts, WidgetGNN supports dynamic changes such as resizing, addition, and repositioning of widgets while preserving design rules. User experience (UX), brand, and structural rules can be flexibly defined and modified through system settings without requiring changes to the underlying model.

Future autonomous vehicle cabins may function as immersive spaces in which riders rest, work, or watch media while the vehicle manages navigation and control. Hazards create a communication challenge because riders may have little awareness about road conditions when the cabin needs to convey risk. We examine graceful degradation as a design lens for coordinating multimodal cabin responses across changing hazard states. We conducted a storyboard-based study with N = 12 participants who reviewed three scenarios with different time profiles: a gradual snowstorm, an intermediate pedestrian event, and a sudden incoming vehicle. Participants reflected on staged visual, auditory, and haptic cues. They associated graceful responses with cue strength that tracked hazard tempo, visible system intent, restrained haptics, and a prompt return to the baseline cabin state after danger passed. We position graceful degradation as a temporal interaction framework for rider-facing Level 5 autonomous vehicle cabins. We also identify design tensions and next-step studies around escalation thresholds, haptic intensity, sensory load, trust, and emergency cue offset.

Electric vehicles (EVs) reduce powertrain noise, creating safety challenges and opportunities for sound design. This paper examines whether generative audio supports personalised dynamic acoustics for future EVs. We report a research through design study in which AI-generated sound samples were selected, edited into seamless loops and embedded in a sonification (the process of translating non-auditory data into sound) prototype. The prototype connects a Processing vehicle interface to a Pure Data audio engine using Open Sound Control. Vehicle speed and throttle input modulate pitch, amplitude and load related parameters. A user study with 20 participants combined the Acceptance Scale with open questions. The results show moderately positive acceptance, with an average usefulness of 0.88 and a satisfaction of 0.62 on a scale from -2 to +2. Participants valued personalisation and responsiveness, but requested stronger recognisability and better throttle mapping. Generative AI is useful for early EV acoustic prototyping, while the final design requires expert refinement and safety evaluation.

The success of sustainable micromobility depends on the rent–ride–return interaction between users and vehicles, yet the design features that underlie affective experience and rental intention remain underexplored. This study compared six morphological variables of shared electric bicycles (frame geometry, frame tube, handlebar curvature, storage placement, dock placement, and interface placement), with perceptual descriptors derived using large language models. To capture effects that emerge only in motion, 30 prototypes spanning these variables were rendered as 5-second animations and rated online (n = 98). Quantification Theory I revealed that dock placement exerted the strongest influence on rental intention, shaping perceptions of “effort-saving” and “lightweight”. A validation experiment partially supported these predictions and indicated that components with negligible perceptual effects can be delegated to generative AI for styling. These findings suggest the return maneuver, a point of infrastructure-vehicle integration governed by docking context, deserves closer design attention in shared micromobility adoption.

As advanced driver assistance systems (ADAS) become more complex in software-defined vehicles, drivers increasingly rely on visual alerts to understand system states, limitations, and interventions. However, an alert may show that something is wrong without explaining why it appeared or what the driver should do. This work analyzes vehicle-owner complaints from the NHTSA Vehicle Owner Complaint database to identify recurring interpretation breakdowns in visual ADAS alerts. Through a multi-stage filtering pipeline, a rule-based modality tagger, and hierarchical topic modeling, three breakdown mechanisms are identified: missing causal grounds and next steps, fragmented alerts within feature families, and limited transparency during abrupt interventions. These inform design implications for explainable visual ADAS alerts — What–Why–Now What explanations, progressive disclosure across cluster and center displays, and confidence/evidence cues for trust calibration — illustrated through scenarios that adapt explanation density to urgency and criticality.

The transition toward highly automated vehicles is primarily driven by safety and efficiency, often overlooking subjective aspects of driving such as enjoyment, emotional engagement, and the pleasure of driving itself. These factors remain central to the mobility experience of drivers. We evaluate a context-sensitive, smartphone-based conversational companion that proactively delivers location-based narrative content through a Wizard-of-Oz setup in a field study (N = 20). To balance exploration with driving safety, conversational interventions were triggered based on contextual factors such as location, the current driving situation and social context. Although the system ranked in the top ten percent for novelty and stimulation in the global metrics of the UEQ, correlation analysis revealed a significant negative correlation with car passion. This suggests that preserving the joy of driving requires personalized conversational agency, where systems must intelligently distinguish between exploratory and performance-oriented driving contexts.

As autonomous vehicles (AVs) diversify, cyclists may encounter AVs of different sizes. However, existing research on cyclist-AV has mainly focused on passenger cars and has still lacked cross-size exploration. This work conducted two studies to understand how cyclists currently interact with different-sized vehicles, as a starting point for informing how cross-size AVs should communicate with cyclists. Study 1 (survey N=92; focus groups N=8) derived a cyclist-centred four-category conceptual structure of vehicle size (Small, Medium, Large, Huge). Based on this, Study 2 (N=120) asked cyclists to self-report their perceptions and behavioural tendencies when interacting with different-sized vehicles. We found that larger vehicles were associated with lower perceived safety and stronger preferences for spatial distance. These results provide baseline evidence that vehicle size is a potential factor that influences cycling interactions. Therefore, future studies and designs should more explicitly account for vehicle size.

Driver distraction from in-vehicle information systems contributes significantly to road accidents. As physical controls (buttons and dials) are increasingly replaced by touchscreens and touchpads, understanding how different input modalities affect driver distraction and performance becomes critical. This study examines the effects of four input modalities (touchscreen, touchpad, button, and dial) on driver distraction. Thirty-seven participants performed secondary tasks using each input modality while driving in a simulator. Driver distraction was assessed using driving performance metrics, task performance, gaze behavior measures, and subjective ratings. Touchscreen and touchpad use resulted in significantly greater lane position variability than button input. All modalities significantly increased the human-machine interface (HMI) fixation and reduced road fixation relative to baseline driving. These findings indicate that the button condition was associated with favorable usability-related outcomes and may inform future investigations of function allocation in vehicle interface design.

Public Mobile Robots (PMRs) are increasingly operating on urban sidewalks, yet the physical design of these shared spaces has received little empirical attention as a determinant of pedestrian acceptance. This work-in-progress paper presents findings from a within-subjects experiment (N = 500) examining how three sidewalk separation types, Shared Space, Semi-separated Space, and Fully-separated Space, affect pedestrian subjective perceptions and coexistence acceptance. The three separation types represent conceptual conditions spanning a spectrum currently under policy and infrastructure discussion across multiple countries, rather than physical designs we implemented or evaluated in situ. Repeated-measures ANOVA revealed that Shared Space consistently produced the least favorable evaluations across all seven measured constructs. Semi-separated and Fully-separated Space conditions yielded comparable outcomes for Perceived Usefulness, Sidewalk Design Acceptance, and Coexistence Intention, suggesting that the presence of any spatial demarcation, rather than its degree, may represent the more consequential factor in early acceptance formation. We discuss what these findings suggest for the design of PMR coexistence environments and identify open questions for future research, including how different forms of spatial demarcation may be experienced across diverse pedestrian populations and real-world deployment contexts.

While the concept of the Human Digital Twin has been proposed within the automotive domain, a structured framework to define and operationalize it is currently lacking. This paper interrogates the concept of Human Driver Digital Twins (HDDTs), systematically distinguishing them from Human Driver Digital Models (HDDMs) and Human Driver Digital Shadows (HDDSs) based on data integration levels and bidirectional feedback loops. We examine a broad spectrum of automotive use cases to assess how the HDDT paradigm can enhance existing applications and where conventional driver modeling methodologies encounter functional limitations. Furthermore, we acknowledge engineering challenges, such as data sovereignty, standardized interaction protocols, and modular internal architectures. Ultimately, this work highlights how an HDDT framework allows systems to transcend the boundaries of static driver modeling, enabling persistent, adaptive, and personalized representations of the human driver.

In-vehicle interaction is often envisioned as requiring as little explicit user input as possible. In multi-user scenarios, however, varying user states and expectations can lead to conflicting automation scenarios. Including more than one passenger introduces complexity to the interaction space that has been neglected by existing literature. We tackle this with a systematic Simulation-Augmented Decision Space Exploration (Sim-DSE) approach: Relevant in-vehicle interactions are identified in an expert workshop (N = 5) and utilized to model a suitable decision space for in-car automations. We then simulate 1000 permutations in this decision space using an agentic AI framework to derive the reasoning behind these automations. Next, we validate the simulated outcomes and their rationales through human feedback in an online survey (N = 100) and distill the results into design guidelines for multi-user automation. The method and guidelines help future researchers and developers to mitigate automation conflicts in multi-user scenarios.

Modern automotive software architectures have evolved into highly complex networks of software components, electronic control units, signals, and thousands of properties. Understanding these multi-format, large-scale, highly-nested software structures is a significant cognitive challenge for domain experts. To address this, this Work-in-Progress paper presents an interactive software visualization artifact developed following the Design Science Research Methodology. The designed artifact ingests heterogeneous automotive vehicle data (ARXML, YAML, and binary files) and establishes a unified pipeline to transform these raw, disconnected metadata into comprehensive network models. By applying knowledge graph principles, the system explores graph visualization libraries to build stable mental models of complex software structures. Finally, this study proposes to evaluate the artifact through a comprehensive usability study with domain experts, while considering its support for users’ cognitive mapping.

Driver emotions such as stress, frustration, and anxiety are associated with riskier behavior and reduced situational awareness, motivating in-vehicle systems that regulate affect in real time. However, emotion-only approaches capture how a driver feels, but not why, limiting appropriate support delivery. We present Emotune, a work-in-progress framework integrating emotional state assessment, driving-context awareness, and Large Language Model (LLM)-assisted intervention selection within the CARLA simulator. The framework uses a human-guided workflow, in which predefined interventions are selected by locally deployed LLM agents. As a methodological contribution, it enables studying how contextual information can inform emotion regulation in driving. We report a pilot study comparing a baseline emotion-aware agent with a context-aware agent that grounds interventions in driving situations. No statistically significant between-condition differences were found in the pilot study. Qualitative feedback instead highlighted contextual relevance, intervention timing, and perceived empathy as priorities for future investigation.

As vehicles progress toward higher levels of autonomy, human–vehicle interaction will increasingly shift from driving-related tasks toward conversational engagement, highlighting the importance of conversational agent design. This paper presents a multi-method investigation of user-centered and brand-consistent visualization strategies for in-car conversational agents (ICAs), combining a quantitative survey based on personality and brand frameworks, an exploratory benchmarking analysis of existing automotive conversational agents, and an experimental study in a simulated driving environment examining responses to varying levels of visual abstraction and anthropomorphism. Preliminary results suggest that functional reliability and voice interaction quality form the primary acceptance baseline, while visual representation serves as a secondary and strategically differentiating layer. Rather than highly anthropomorphic or complex visualizations, adaptive concepts offering adjustable expressiveness, personalization, and modular abstraction emerge as a robust design direction for integrating ICAs into vehicle identity, user experience, and long-term human–machine relationships.

Despite a decade of trust research at AutomotiveUI spanning the full SAE automation spectrum, a cross-level synthesis of the field’s own findings has not been explored. This PRISMA-guided scoping review examined 30 full papers screened from 338 candidates across ACM AutomotiveUI proceedings (2016–2025), coded across SAE level groups and two temporal eras. The analysis suggests three distinct psychological frames operating under the single label of trust, shifting from supervisory reliability at L1/2 through calibration at L3 to social delegation at L4/5. Across the reviewed studies, trust-calibrating HMI information followed a similar progression from system-status information toward system-reasoning information, a pattern this review characterized as the HOW–WHY information gradient. Seven of the nine L4/5 studies nonetheless placed participants in driver-seat configurations despite the passenger role of high automation, a recurring methodological pattern characterized here as a potential driver-seat paradox. Across multiple studies, subjective trust ratings and objective behavioral outcomes did not consistently align, suggesting an unresolved theoretical gap in understanding how trust is reflected in observable behaviour. Resolving this disconnect, establishing passenger-role designs at L4/5, and formalizing design knowledge at L1/2 are key directions for future trust research.

Rear-end collisions are prevalent and are closely related to car-following (CF) behavior. In this study, we proposed three in-vehicle human-machine interfaces (HMIs) to provide progressively richer information about the lead vehicle (LV), including speed information (HMI-vel), time headway (THW) information on top of HMI-vel (HMI-THW), and coloured risk warnings based on THW thresholds on top of HMI- THW (HMI-alert). Cluster analysis was conducted to identify heterogeneous driver groups in a driving simulator study with 24 participants. Results showed that, in general, enhancing drivers’ perception of lead-vehicle states improved the safety margin during car-following segments without intensive braking, but certain HMIs impaired speed stability, with the effects varying across driver groups. Furthermore, the safety benefits of HMIs, reflected by reduced collision risk, differed across driver groups. These findings highlight the importance of considering driver heterogeneity and adaptive HMI designs to support safe car-following behaviors.

Exploring historical design trends helps reveal what shapes design evolution and may provide insight into future direction. This work-in-progress presents initial findings from a larger study spanning a 60-year period and explores patterns regarding the material use and type of interaction points within the car interior. Interior images from 57 Wheels “Car of the Year” winners, were collected, objectively coded, then analysed by their visual patterns. Preliminary findings and observations indicate a cyclic use of metal on some panels, the evolution and reduction of novel interaction points across the dashboard and center console, and the objective matrix provides opportunity to visualise further patterns. These preliminary findings contribute to the community by providing an understanding of past trends by tracking their propagation, teasing out meaning and decisions across different automotive market sectors, and to hint at interaction features that have persisted, offering relevance for future vehicle interiors.

Safe mobility is important for older adults, yet age-related declines may lead many older adults across cultural contexts to reduce or stop driving. Autonomous vehicles (AVs) could extend older adults’ mobility; however, this potential can only be realized if this technology is designed in ways that older adults prefer, accept, and find usable. This paper explores how interface modalities influence older adults’ perceptions of AVs in two cultural contexts: Sweden and Japan. We conducted a video-based online study with 44 older adults with driving licenses (22 from Sweden, 22 from Japan) comparing four passenger interface conditions across two scenarios with different visibility conditions: no interface, text-based, voice-based, and an embodied robot interface that communicates information about an AV’s actions. Our results indicate that providing an interface significantly improves older adults’ acceptance and trust relative to no-interface, with voice-based explanations yielding the highest mean acceptance in both countries. However, differences emerged: Japanese participants reported higher trust in the robot interface and more often selected a child-like voice, whereas Swedish participants more often selected adult voices. These findings suggest that tailoring AV human–machine interfaces to older users’ cultural expectations and communication preferences can enhance their comfort and adoption.

Automated mobility solutions are rapidly advancing globally, offering potential benefits in safety, sustainability, efficiency, and inclusivity. However, these benefits are not guaranteed and could decrease if improperly managed. This paper introduces the Fast Automated and Smart mobility impact Tool (FAST), a method to assess and visualize automated mobility’s impact on travelers and society. Developed through an iterative design process involving literature review, tool design, and expert validation, FAST enables rapid initial assessments to stimulate discussion and inform decision-making. Preliminary validation shows potential but highlights the need for objective clarification, impact indicator refinement, and improved user group distinctions. Future work includes additional validation, framework elaboration, and interactive tool development to better support policy makers, researchers, and designers in evaluating automated mobility options. FAST can empower UI researchers to initiate dialogue with policy makers and industry, steering toward meaningful human-machine interactions, improving travel behavior, and facilitating acceptance of desired automated mobility futures.

With the development of automated vehicle technologies, automated vehicles (AVs), human-driven vehicles (HDVs), and vulnerable road users (VRUs) are expected to coexist in mixed traffic for the foreseeable future. Amongst VRUs, cyclists face particularly high safety risks, making it important to understand how AVs should interact with them. HDV behaviour can serve as a useful benchmark, providing a reference point against which AV interactions can be evaluated and refined. However, it is currently not yet well understood whether AVs and HDVs exhibit different kinematic behaviours and interaction outcomes when interacting with cyclists. This study uses naturalistic trajectory data from the Waymo Motion Dataset to compare cyclist-AV and cyclist-HDV crossing-path interactions. The preliminary results indicate that vehicle kinematics may serve as important implicit cues. This work offers potential guidance for future research on understanding the role of vehicle kinematics in shaping safe and interpretable cyclist-AV interactions.

Navigating unregulated intersections with a heavy vehicle is a challenging task harboring significant risks for the vehicle itself and other traffic participants. In this study we introduce a dynamic, evidence-based risk assessment algorithm predicting increased general risk at upcoming intersections based on environmental cues. We qualitatively evaluated two HMI configurations—a standard dashboard versus an adaptive Head-Up Display (HUD)—with professional drivers in a driving simulator. Drivers perceived the predictive risk indicator as highly useful, preferring the HUD layout. Despite physical simulator constraints, the results strongly support the utility of predictive general risk assessment. Future work will integrate real-time driver monitoring of eyes-off-road (EOR) metrics to dynamically adjust risk indicators based on driver attentiveness.

Parkinson’s disease (PD) can impair abilities essential for safe and independent driving. However, little is known about how individuals with PD experience driving and perceive current driver assistance systems. At the same time, advances in driver assistance technologies may offer new opportunities to support mobility, independence, and continued participation in road traffic. To investigate these experiences and perceptions, we conducted an online survey with N = 52 individuals diagnosed with PD, followed by N = 9 semi-structured interviews exploring driving experiences, changes in driving ability, and attitudes toward assistance technologies. Our findings show that driving remains important for independence and quality of life, with many participants adapting their driving behavior rather than ceasing driving altogether. While driver assistance systems were generally perceived positively, participants emphasized the need for solutions that accommodate the diverse and evolving symptoms of PD. Based on these findings, we discuss opportunities and identify research directions for the design of PD-aware driver assistance technologies.

Digital Twins in the automotive domain have been developed to reproduce vehicle behaviour and optimise performance, yet the subjective driving experience remains largely unaddressed. Existing Driver Digital Twins reproduce observable control actions but do not represent the internal processes through which subjective experience arises, limiting their ability to account for why the same physical situation can be experienced differently by different drivers. This paper argues that subjectivity is a cognitive phenomenon and that a model of human cognition, specifically cognitive architectures, can provide a relevant modelling layer for a Human Driver Digital Twin. Four dimensions of driver subjectivity are discussed and mapped to cognitive architecture mechanisms. This represents a step toward Driver Digital Twins with greater human fidelity, though much remains to be specified before such systems can be realised. Grounding driver models in cognition opens directions for designing systems that are responsive to the human dimension of driving.

Naturalistic driving video captures how drivers supervise under real road conditions, but converting long, multi-camera in-cabin footage into behavioral features still relies on manual coding, which is costly and hard to scale. We present a human-in-the-loop workflow that converts raw in-cabin video into two reviewable supervisory features: area-of-interest (AOI) gaze labels and hand-on-wheel state. Instead of a single end-to-end model, the workflow keeps region selection, gaze labeling, participant-specific calibration, temporal stabilization, and window-level metric extraction as distinct, observable stages, so an error can be localized. Specifically, the gaze module pairs Sample and Computation Redistribution for Efficient Face Detection (SCRFD) face detection with a YOLOv8-cls classifier, while the hand module uses GroundingDINO with a temporal state stabilization strategy. The workflow was validated in a naturalistic driving study with millions of video frames from 15 drivers. The source code of the workflow has been released at https://github.com/zdu881/autodri.

This Work-in-Progress examines whether personality-informed prompting changes perceived LLM emotional support in driving scenarios designed to elicit stress. A condition-order-balanced, within-subject CARLA simulator study (n = 14) compared a baseline with a Driver Personality Profile (DPP) condition; a supplementary online video pilot (n = 12) examined response perception without driving control or live-system latency. No statistically detectable condition differences emerged for usefulness, ease of use, privacy concerns, or social/emotional presence, and only 7 of 14 simulator participants identified the DPP condition correctly. A post-hoc lexical audit showed that both conditions frequently reused generic supportive scaffolding. We discuss output anchoring as one tentative interpretation, not an established phenomenon: the evidence cannot distinguish constraint-dominated generation from a weak personalization manipulation or limitations of the 7B model. The findings motivate stronger, independently validated personalization manipulations and privacy-aware in-vehicle support.

This study explores the potential of Google Veo 3, a generative video model, to synthesise 8-second dashcam-style urban traffic scenes based solely on text prompts in 76 cities across six continents. YOLOv11x was used to count road users, traffic lights, and stop signs, revealing variations across cities: Karachi had the most objects detected (79), while Muscat had only four cars. Audio analysis using dBFS showed that Montevideo was the loudest, while Copenhagen was the quietest. Through a qualitative visual analysis, the authors assessed and confirmed the perceived authenticity of most traffic scenes and highlighted AI errors, including the inability to handle non-English languages in these videos. Moreover, we compared 10 synthetic videos of New York City and Kampala, each, and verified that Veo 3 is consistent. To summarise, Veo 3 is capable of synthesising authentic, logical traffic scenes worldwide; nevertheless, it still poses non-negligible errors.

In-vehicle pedestrian detection displays show what a vehicle senses, but a sensor detects more than a display can usefully present. We argue that what a display surfaces should be prioritized by human monitoring: confirm the pedestrians an occupant reliably detects, and flag those they miss. In an SAE Level 4 driving simulator, participants monitored an automated drive and reported pedestrians under intermittent visual-search load. Across 488 encounters from 27 participants, moving pedestrians were detected 90–98% and a stationary, occluded pedestrian 44%, despite passing within 7 m. Detection tracked visual salience, not proximity. Projecting position into the gaze record shows the miss is split: in about half, gaze reached the pedestrian and went unreported; in the rest, gaze never arrived. Adding dashboard detail raised trust without raising workload but did not recover the missed pedestrian, implying the safety cue belongs in the environment, not on a dashboard panel.

As autonomous-driving systems take on increasing control authority, drivers are repositioned as human supervisors who must monitor decisions they did not make. Yet, today’s in-vehicle interfaces expose only the system’s final actions, leaving its underlying perception and reasoning invisible — a structural information asymmetry that manifests as a chronic fear of the unknown and undermines calibrated trust. Drawing on a formative study with 18 experienced ADAS users, we show that drivers’ trust is shaped more strongly by why and what-next explanations than by mere object recognition, and that the largest gap between current and desired information lies in intent, reasoning, confidence, and limits. Building on these findings, we derive three design principles that align the autonomous-driving pipeline with the driver’s reasoning order, and instantiate them in the Transparent ADAS Interface — a two-area design that communicates the vehicle’s behavior and intent alongside what the system perceives and how confidently it does so.

Although extensive research has been conducted on driver fatigue detection, few studies have explored how fatigue affects drivers’ car-following behaviors and, in turn, how to detect fatigue based on car-following features. In an on-road experiment, we collected multimodal physiological data and eye-tracking data from 5 long-haul truck drivers, along with vehicle motion and surrounding traffic data. Drivers’ fatigue states were labeled based on the percentage of eyelid closure over the pupil (PERCLOS). Comparative analyses of car-following behavioral parameters and physiological indicators under different fatigue states were performed. Furthermore, we explored the feasibility of estimating drivers’ fatigue levels using car-following behavioral parameters and compared its performance with models based on physiological features. The results show that the models based on car-following features showed better performance than models based on physiological indicators in this preliminary endeavor. These findings provide theoretical evidence for the feasibility of detecting surrounding human-driven vehicles’ fatigue states based on their driving characteristics using lightweight, interpretable, and practically deployable models.

While conversational AI agents (CAs) have the potential to support drivers through hands-free and eyes-free interaction, their role in long-haul trucking remains unexplored. This paper investigates how long-haul truck drivers perceive CAs and identifies design considerations for their successful deployment. We conducted ride-along observations and semi-structured interviews with five long-haul truck drivers during regular driving operations. The results show that drivers today regulate verbal communication according to driving demands and expect communication partners to adapt to changes in workload and attention. Participants were most receptive to CAs supporting driving-related tasks, including traffic information, rerouting and route planning. Trust, perceived complexity, and discomfort with conversational interaction emerged as key barriers to adoption. We conclude that CAs may be best suited to supporting strategic and selected tactical tasks, while being less appropriate for operational tasks. Effective CA design must adapt to context and workload, support interruptions, and provide transparent and trustworthy support.

Visual interface design strongly affects operator performance in teleoperated remote inspection, yet evaluations have typically relied on user-reported ratings alone. We report a within-subjects laboratory study (n = 24) comparing five visual interface configurations for a teleoperated mobile platform, including a static depth-cue overlay tested in both single-view (first-person view; FPV) and dual-view (FPV plus third-person view; TPV) configurations. Helpfulness, on-FPV gaze dwell, and task performance all ranked the TPV-only condition below each FPV-based condition, providing convergent behavioural evidence for the user-reported condition ordering. Within FPV-based conditions, the overlay systematically redirected gaze into its target region, holding across the large majority of participants in both viewing configurations, yet user-reported helpfulness showed no corresponding augmentation effect. This dissociation between behavioural and self-report measures was echoed in modest but consistent trial-level correlations between helpfulness and the behavioural measures. These findings support combining gaze allocation and task performance with user-reported ratings in teleoperation interface evaluations.

Accurately measuring comfort and perceived safety in automated vehicles (AVs) is critical for understanding user experience, accommodating various needs, and eventually contributing to the public acceptance of AVs. However, methodological practices vary largely across the field. To address this gap, we organised a workshop at AutomotiveUI 2025, where three expert talks covered subjective measurements, objective and physiological measures, and AI-driven models, followed by two structured group discussion sessions on current practices and future methodologies. This work-in-progress synthesises key insights from both the expert talks and community discussions. Critical themes include, for example, the need for context-specific, deliberate selection of subjective measures; the promise and practical limitations of physiological signals; the ground truth challenge that complicates subjective-objective combination; and the emerging role of artificial intelligence (AI) for real-time inference and multimodal combination. We offer these insights to provide the community with a common methodological landscape and to promote more robust practices.

Autonomous vehicles (AVs) must communicate intent to pedestrians in mixed traffic to ensure safe interaction, and vehicle deceleration motion is a promising implicit communication channel. Designing acceptable AV deceleration remains difficult because it is continuous with interdependent kinematics, making fixed‑design experiments inefficient. We therefore parameterised deceleration using a minimum‑jerk model (MJM), which generates smooth speed profiles from tuneable parameters. We will conduct an adaptive experiment in an immersive pedestrian simulator, where Bayesian optimisation sequentially proposes AV deceleration profiles expected to receive higher pedestrian ratings for safety and efficiency, based on previous ratings. Preliminary results from fitting MJM to 62 naturalistic human profiles indicate that the MJM reproduces the overall human profile shape (median RMSE = 0.88 m/s) with only 2 parameters (i.e., stopping distance, duration), supporting its use as a profile generator for the experiment. This work can contribute a human-in-the-loop framework for designing AV motion as an implicit interface.

This paper presents a method to integrate virtual car interiors with interactive User Interfaces (UIs) for holistic simulation in a mixed reality (MR) environment, coupled with an adjustable seating buck. For user-centric development of automated vehicles, early-stage feedback on interiors, including Human-Machine Interfaces (HMIs), is highly essential. Methods to simulate automotive interfaces integrating ergonomics and mixed reality techniques that include users in early phase testing are still not explored enough. The proposed method is implemented in Unreal Engine (UE). It showcases the workflow for integrating User Interface (UI) prototypes developed on external platforms like ProtoPie and Figma into the UE platform, thus not only representing the UI visually but also integrating the UI prototype’s interaction logic with the simulation environment bi-directionally. This UI prototype simulation is coupled with the adjustable seating buck to evaluate ergonomic aspects, and integrated with a mixed reality workflow to represent both the user’s body and the desired 3D model variant of the interior. A Pilot study was performed to validate the proof of concept (POC).

In this paper, we present a novel approach of assessment of user interfaces, designed to positively affect passenger comfort, using electrogastrography (EGG). EGG is a method that records electrical activity of the stomach, and has been shown to reliably assess motion sickness, which is one of the main contributors to physical discomfort in AVs. A user study with 26 (22 complete data sets for analysis) participants was conducted in a driving simulator, assessing three types of auditory-tactile user interfaces (UI) (baseline, abstract, representational) for highly automated driving using EGG features such as dominant fre-quency, tachygastria, and crest factor of spectral density. The results indicate that the percentage of power spectrum density in the tachygastric range increased when using an abstract UI during events requiring lateral AV maneuvers. These findings suggest that EGG can serve as an objective method for evaluating passenger-centric interfaces in future vehicles.

AI-based assistance is expanding in-vehicle functions and increasing how often drivers use human-machine interfaces (HMIs) in everyday driving. This growing role makes research on safe and effective interface design increasingly important to automotive development. Academic research has produced a body of work relevant to this need. However, automotive-specific research has documented continuing challenges in communicating and incorporating this knowledge into vehicle-development practice. To examine this research-practice gap from practitioners’ perspectives, this exploratory study conducted semi-structured interviews with 12 experts from original equipment manufacturers (OEMs) and related industry organizations. The thematic analysis identified five themes concerning HMI value, decision-ready evidence, collaboration structure, internal governance, and layered evidence use. These themes led to three practical recommendations: make findings actionable, create outputs beyond academic papers, and engage OEM partners early. Overall, the findings aim to strengthen collaboration between academic researchers and industry practitioners and support future automotive HMI development.

Takeover support requires identifying the situation awareness (SA) deficit affecting a driver’s response. A single SA measure cannot identify the deficit or support need. This study therefore defines support-relevant SA across two layers. The traffic-event layer identifies support timing and location through temporal (materialized and unmaterialized hazards) and spatial (forward path and adjacent lane) dimensions. The information-processing layer uses Quantitative Analysis of Situation Awareness (QASA) to separate actual SA (ASA; situation knowledge), perceived SA (PSA; confidence), and response bias (response criterion). Their intersection yields 12 support-relevant targets. Using physiological signals from 66 participants, the study applied the Weighted Importance Score and Frequency Count (WISFC) framework across six model families to compare retrospective window-level feature rankings. ASA and response bias rankings showed preliminary variation, whereas PSA shared pupil and gaze features across dimensions. These patterns suggest which targets may be physiologically distinguishable or require contextual evidence for takeover support.

In-vehicle assistants (IVAs) are moving from fixed command grammars toward large language model (LLM) driven conversational agents, where the quality of the system’s response depends heavily on how the driver phrases the request. Yet we know little about how drivers actually formulate spoken commands, and almost nothing about how this varies across cultures and languages. We present the design and motivation of PromptDrive, a web-based, remote, unattended driving-simulator platform that elicits spoken command utterances under three priming styles (Direct, Goal-based, and Contextual) delivered through a simulated head-up display. We ran a cross-cultural, multilingual pilot study with university students in Germany and Mexico, assembling a trilingual (German, Spanish, English) corpus of command utterances for two vehicle functions, navigation and climate control. In this work we frame the cross-cultural IVAs prompting problem, situate it in prior work on cross-cultural HCI, and prompting, and motivate why command-phrasing patterns may differ systematically between a German and a Mexican driver population. We report preliminary observations from 30 participants and 383 utterances: priming style shifts how drivers phrase requests, with Direct cues yielding the shortest commands and Goal-based cues the most specific, while retrospective workload did not differ across styles and drivers preferred the Direct framing. Because the achieved sample fell short of the planned size and the cells are unbalanced, these results are exploratory; the paper’s primary contributions are the conceptual framing, the elicitation protocol, and evidence that the platform supports unattended multilingual collection at a distance.

This paper investigates how different Explainable Artificial Intelligence (XAI) modality combinations affect users in fully automated vehicles across different types of ambiguous driving scenarios (ADS). In an online study (N = 123), three explanation modality combinations were compared: visual-textual (VT), visual-auditory (VA), and visual-textual-auditory (VTA). VT resulted in significantly lower situational trust and significantly higher cognitive load than VA and VTA. Explanation satisfaction increased significantly from VT to VA and was highest for VTA. While baseline levels of trust, satisfaction, and cognitive load differed between ADS categories, no significant modality–scenario interaction was found, indicating that modality effects generalize across forms of ambiguity. Based on these results, five design guidelines for XAI in fully automated vehicles were derived. Overall, multimodal explanations appear beneficial, particularly during initial use, but should adapt to user needs and context.

As vehicles become increasingly automated, the driver’s role is shifting from active controller to passive supervisor. Automation reduces manual effort, but it also introduces the risk of disengagement and decreased situational awareness. This paper presents EchoDrive, a gamified in-cabin system designed to support situational awareness (SA) in vehicle occupants during automated driving through spatialized audio cues and gesture-based responses. The system allows drivers to initiate sonar-like scans and point toward the audio cues responses of the detected traffic participants, fostering periodic perceptual engagement. A within-subjects simulation compared EchoDrive to a mobile game baseline. Results showed that, compared to a visually demanding smartphone game baseline, EchoDrive increased subjective driver SA, reduced reported cognitive workload, and improved takeover readiness, while being positively received by participants. This work highlights the potential of lightweight multi-sensory interaction to support driver presence in automated vehicle contexts, though the asymmetric comparison limits attribution to specific design elements.

Robotaxis are beginning to move beyond restricted pilot programmes and enter everyday traffic in selected cities. Yet much of what we know about interacting with autonomous vehicles (AVs) has been established in simulators and labs, leaving the lived experience of riding a deployed robotaxi underexplored. In this work, we conducted an autoethnographic study of five robotaxi rides across two platforms in Shenzhen. Drawing on our first-hand experience as passengers as well as our expertise in AV research, we surface three tensions: a habitual urge to monitor the road that persists beyond prevailing trust-vigilance accounts; an uncanny unease as the vehicle negotiates traffic too convincingly like a human driver; and the friction of pick-up and drop-off, where the passenger becomes a pedestrian the system no longer attends to. In response to each tension, we offer design considerations that extend current interaction design paradigms to capture the nuances of robotaxis as they operate in real-world settings.

In future driving, manual drivers will interact with Autonomous Vehicles (AVs) through Vehicle to Vehicle (V2V) communication. This study evaluates the human-machine interaction trade-offs of interface modality stacking for center-stack touchscreens during a cooperative overtaking task. Using a Virtual Reality driving simulator with 26 participants performing a visual secondary distraction task, we compared three configurations: Visual (V), Visual+Haptic (VH), and Visual+Auditory+Haptic (VAH) across normal and time-sensitive requests. Results show that the VAH setup improved overall driver response times, and enhanced request-type classification accuracy for time-sensitive requests. However, an interaction effect revealed that modality stacking sequentially escalated perceived pressure and subjective workload (temporal demand and frustration) specifically during time-sensitive requests. A marginal trend indicated increased system intrusiveness as sensory layers accumulated. These suggest a design trade-off for cooperative driving Human-Machine Interfaces (HMIs): while multi-sensory density dramatically recovers tactical accuracy, it carries an immediate psychological cost that designers must try to mitigate.

Motorcyclists are highly vulnerable in complex traffic, where hazards are often detected too late for a timely response. We evaluated MOTEX, a helmet-integrated HMI that provides proactive directional cues through peripheral visual, auditory, and multimodal feedback. In a within-subject simulator study, 14 participants completed 16 scenarios based on common motorcycle crash typologies. The feedback mode did not significantly affect the reaction time or the minimum time-to-collision, although the visual and multimodal conditions showed numerically faster responses and greater safety margins than the audio and control conditions. Participants reported low cognitive workload, moderate trust, and generally positive evaluations of cue clarity. Perceived benefits of situation awareness were mixed, with most of participants reporting a neutral effect. These findings suggest that MOTEX may function less as a direct determinant of behavioural response and more as a supportive “second pair of eyes’’ for spatial confirmation.

External Human-Machine Interfaces (eHMIs) for automated vehicles must remain quickly interpretable if they are to scale to realistic traffic. This study examines a deliberately simple directional light-based windshield eHMI as a first step toward that goal. The eHMI uses a minimal light strip whose highlighted segment shifts based on the pedestrian’s relative position, aiming to convey yielding intent without text or complex visual elements. In a virtual-reality study with 30 participants, pedestrians encountered an automated vehicle with and without the eHMI; half received a brief onboarding video, while the others learnt through repeated exposure. Results suggest that the eHMI supported earlier crossing initiation and increased perceived safety and trust. Onboarding produced immediate benefits, whereas participants without onboarding showed gradual improvements over repeated encounters. These findings indicate that the signal is learnable in a controlled dyadic setting and motivate future VR studies of multi-pedestrian interaction, social influence, and traffic complexity.

A city’s traffic runs on planning decisions almost nobody sees, and autonomous vehicles widen that gap: they read the road network and route through it, while the people sharing the street understand less and less of how it works. Path Weavers is a concept for closing that gap through play. It turns a CAVE into a living city where two or more players take opposing roles. One side builds the road network, the other disrupts it with urban events, and the vehicles navigate whatever infrastructure exists. The room itself does the explaining: traffic state shows through color, sound and floor feedback rather than menus or screens. Beyond a short briefing, players get no instructions, so the central problem is legibility, which we address with a layered Explainable UI. This paper presents the concept, the decisions that shaped it, and the open questions we bring to the AutoUI community.

As urban air mobility (UAM) advances toward operational deployment, electric vertical takeoff and landing (eVTOL) pilots may face heightened cognitive demands from dense low-altitude traffic, strict energy constraints, and safety-critical flight phases with complex control-mode transitions. Real-time pilot state monitoring is therefore essential for safe operations. Prior electroencephalography (EEG)-based workload-estimation models for pilots predominantly focus on temporal dynamics, while insufficiently modeling inter-channel dependencies, which are crucial for effective spatial aggregation across EEG electrodes. In this study, we propose the Cortical Topology-Aware Spatiotemporal Graph Transformer (CTA-STGT), which explicitly encodes structured inter-channel relationships via a topology-aware graph convolutional network constrained by a cortical-topology-derived adjacency matrix. Temporal patch embeddings and spatial graph embeddings are projected into a unified latent space and then jointly fused by a Transformer encoder to capture spatiotemporal dependencies. Evaluated on EEG data from a 6-degree-of-freedom eVTOL simulator experiment with three workload levels based on NASA Task Load Index (NASA-TLX) scores, CTA-STGT achieved an accuracy of 0.89 and an F1-score of 0.89, outperforming the baseline models. These results provide a technical basis for adaptive workload-mitigation support in high-workload eVTOL scenarios.

To effectively correct older drivers’ self-estimation, we proposed a review method that considers the attitude change associated with the correction of their overestimation. First, reanalyzing previous data, we hypothesized that the attitude change in optimistic comparison with same-age drivers is associated with the degree of overestimation correction. To verify this hypothesis, we conducted an attitude survey before and after a self-estimation correction education. The results supported this hypothesis. Based on these findings, to reduce their optimistic attitude toward their own driving compared to same-age drivers, we proposed a review method that presents and compares the driving evaluations of other same-age drivers alongside the participant’s own driving evaluation in a driving simulator. To evaluate the proposed method’s feasibility, we conducted a preliminary experiment with three older drivers, which indicated its potential to effectively reduce optimistic comparative attitude compared to same-age drivers and correct overestimation.

Conditionally automated driving offers engagement in non-driving-related activities such as relaxation. The present driving simulator study (N = 12) compares three relaxation methods, Closing the Eyes, Sounds of Nature, and Diaphragmatic Breathing, considering their impact on driving performance, physiology, and subjective experience. The best results were achieved with Sounds of Nature in the majority of the different evaluation categories.

Explanations in automated vehicles have been proposed to support users’ understanding, trust, and acceptance of automated driving behavior. However, many explanation interfaces remain static interventions, even though users’ information needs may vary depending on driving context, familiarity, perceived criticality, and preference. This work presents a simulator platform for prototyping adaptive in-vehicle explanations in automated driving. The platform combines a Unity-based driving environment, a motion-based, VR-enabled setup, scripted traffic incidents, an in-vehicle explanation interface, and a feedback loop to adapt explanations. The system uses a large language model-based explanation pipeline that receives scenario information and user feedback to determine whether an explanation should be presented, how detailed it should be, and how it should be delivered. We describe the platform, outline the design space it enables, and reflect on how such systems can support AutomotiveUI research on adaptive explanations, explanation suppression, user control, and human-AI interaction in automated vehicles.

This paper explores how accessibility challenges for older adults in automated buses can be understood through different forms of user representation. Two explorative studies were conducted in an actual stationary automated bus: a user enactment study with three older adults with mobility, vision, and hearing impairments, and an empathy simulation study with six non-target participants using a kit to simulate age-related capability loss. Findings from both studies highlight the importance of multimodal communication, interior design support, and human assistance for older passengers. Beyond these design insights, the paper contributes a methodological reflection on how user enactment and empathy simulation produce different forms of accessibility-related user knowledge. Empathy simulation enables safer exploration of critical interaction scenarios and challenges existing design assumptions, while user enactment with target users grounded such insights in lived experience, situated strategies, and actual capability variation.

Research on sexual behavior and transport has either focused on fear of sexual harassment and perceived safety inside public transport or on consensual sexual activity inside private transport framed as distracted driving. Meanwhile, evidence from adult content platforms suggests that consensual sexual activity inside public transport is trending. This is particularly concerning in relation to public and shared automated vehicles (AVs), where enclosure, shared use, and the absence of an authority figure can warp perceptions of accountability and intervention potential, with important consequences for passenger comfort and perceived safety and for public acceptance of the technology. Our workshop aims to address this concern by approaching consensual sexual activity as a design stress test for how AVs are perceived and used as collective spaces. We aim to synthesize design principles, identify open research questions, and establish a research agenda for designing public and shared AVs as comfortable and respectful collective spaces.

As automated vehicles (AVs) are increasingly deployed in real-world environments, users often do not understand what decisions their vehicles are making, or why. Explainable AI (XAI) offers a promising path toward transparency, yet most existing approaches remain algorithm-centric and inaccessible to end users. This workshop explores how Human-Centered XAI (HCXAI) and the Question-Driven Design framework can be used to systematically discover user explainability needs for AVs. Through scenario-based, participatory activities, participants will adopt different stakeholder perspectives and generate the questions they would want an AV to answer in ambiguous driving situations. Building on the prior workshop at AutomotiveUI 2025, this workshop aims to produce an initial taxonomy of AV-specific explanation needs grounded in real user questions, creating a shared resource that can guide future research, UI design, and evaluation of explainable automated driving systems.

Public Mobile Robots (PMRs), including last-mile delivery, street-cleaning, and guide robots, are increasingly being deployed in shared public spaces such as sidewalks, corridors, and building entrances, where they must communicate and coordinate with nearby pedestrians. Unlike automated vehicles, PMRs operate in close proximity to people, introducing new challenges for intent communication, trust, and socially acceptable interaction. While the AutoUI community has established extensive knowledge on external Human–Machine Interfaces (eHMIs) for automated vehicles, it remains unclear how these principles transfer to PMRs. This workshop aims to examine the applicability of vehicle-based eHMI concepts to PMRs by identifying transferable design principles, challenges requiring adaptation, and new research directions. Through invited talks, comparative discussions, and scenario-based group activities, participants will develop a shared research agenda, propose preliminary design considerations for PMR eHMIs, and strengthen collaboration between the AutoUI and Human–Robot Interaction communities.

Driver monitoring systems are typically designed to detect driver states and as appropriate, issue corrective warning and/or take action (alert, takeover, etc.). Current warning strategies are designed by a rule-based deterministic model, given ease in certification and predictability, with little or no personalization and adaptation to a driver’s individualistic profile and preferences. However, long-term effectiveness of these systems, including wider adoption, depends not only on detection accuracy, but on whether drivers perceive the system to be useful, trustworthy, and supportive. In this workshop, we discuss the benefits of a value-in-use approach that provides context-sensitive interventions to account for driver, situational, and environmental variations. The term “value-in-use” refers to the concept that a system’s worth is determined by what it can deliver to the driver in a real-world setting. Drivers are more likely to accept and benefit from a driver monitoring system that is viewed as a partner in safe driving rather than an intrusive or punitive supervisor. This workshop will present a conceptual roadmap that addresses the questions, “Does the system create value for the driver for real world use” and “What relationships (or partnerships) should exist between humans and systems to provide value?” The workshop advocates moving beyond compliance-oriented warning logic towards a user-centered paradigm in which driver monitoring systems create value through adaptive, relational, and meaningful support, in use. Participants will be asked to provide use case scenarios based on a shared understanding of the possible partnerships that can evolve during a trip.

Automated mobility is reshaping society, yet its development has been primarily OEM-driven, often prioritizing business over societal and individual gains. The goal of this workshop is to facilitate debate around automated mobility impact with the use of the Fast Automated and Smart mobility impact Tool (FAST), and to discuss the role of design-research in working towards smart mobility futures. FAST enables stakeholders to evaluate systems against individual and societal values. Participants will use FAST to brainstorm, assess, and refine automated mobility propositions (from demand-responsive transport to automated mobility scooters, or cars), balancing benefits against risks. The workshop familiarizes participants with the potential impact of automated mobility and provides them with a hands-on tool for discussions around the topic. Outcomes include evaluated use cases, actionable improvements, and recommendations for industry, policy, and the Human Factors and Design research community.

Automated buses are increasingly being deployed in public transport systems. In Linköping, Sweden, they have been in daily operation since 2020. However, their deployment has also raised critical accessibility and inclusive design challenges. There is growing concern regarding how automated buses can accommodate users with diverse mobility, sensory, and cognitive capabilities, especially older adults and people with impairments, throughout the entire user journey. In this workshop, we apply a speculative design approach to engage participants in investigating a fictive accident involving older adult passengers. We translate these speculative scenarios into design research questions addressing safety, accessibility, and inclusivity. Using the automated bus deployment in Linköping as the setting, the workshop aims to generate insights that bridge the gap between research and real-world deployment for a more accessible and inclusive autonomous public transport system.

Semi-automated vehicles utilizing interactive interfaces represent a critical, emergent research field for enhancing road safety and user experience. This half-day workshop aims to advance the design and evaluation of anthropomorphic, intelligent in-car assistants to foster trust and mitigate risks like driver distraction or over-trust. Through paper presentations and interactive discussions, participants will explore agent morphology, spanning voice-only systems, screen-based interfaces, and physical robotic agents, and investigate how these agents can effectively communicate vehicle intentions and driving context. Additionally, the workshop will address the impact of these in-car assistants on driver workload and control takeovers, while establishing robust evaluation metrics, ethical guidelines, and global standardization frameworks. By bridging robotics, vehicular communication, human-machine interaction and automatic control, this workshop seeks to synthesize practical design guidance and outline a future research agenda for safer and more efficient road transportation.

Designing and evaluating in-vehicle Human-Machine Interfaces (HMI) requires researchers and UX designers to simultaneously navigate driving simulation software, interaction design, and user study methodology, a steep learning curve for those without a deep technical background. This workshop introduces SCARline, an open platform built atop a driving simulator that abstracts away technical complexity, letting researchers and designers focus on their research question, study design, and the in-cabin interaction they want to evaluate. A preliminary evaluation with participants from mixed disciplinary backgrounds provided initial validation and surfaced key areas for improvement, motivating this workshop as the next step in an iterative development process. Participants will be introduced to the platform architecture, guided through designing their own HMI widget, and supported in configuring a fully functional user study scenario within the simulator. The workshop invites the broader AutomotiveUI community to engage hands-on with SCARline and collectively inform its future development.

The field of the future of mobility provides researchers, practitioners, regulators and students a brand-new canvas for envisioning emerging technologies, human behaviours, styling designs, customer services and many more details that need rethinking, reconfiguration and re-design in transportation, vehicles and related services. Emerging mobility trends such as autonomous/connected/electric vehicles, micro- and urban-air- mobility, behavioural changes including acceptance of sharing, seeking more equal/inclusive/flexible services, and responding to emotional needs, are transforming innovations. We propose a human-centred interactive workshop facilitated by a playful toolkit and learning process catering for future of mobility themes, developed for professional audiences to go through an ideation process that considers types of people, technology trends and circumstances when creating new vehicles, mobility systems and services. Outcomes will include disruptive ideas around vehicles, systems and services; brand new research agendas for future mobility; emergent mobility innovation methods, all produced through team collaboration with balanced, interdisciplinary expertise.

Passenger comfort is emerging as a key challenge for the successful deployment and public acceptance of highly automated vehicles. Existing research has examined individual comfort determinants, including vehicle dynamics, driving style, trust, perceived safety, and human-machine interaction (HMI), although these factors have been largely investigated in isolation. Consequently, a comprehensive understanding of how they jointly influence passenger comfort remains limited. This workshop aims to bring together researchers and practitioners from the AutomotiveUI community to develop a holistic perspective on comfort in automated driving. Through interactive brainstorming, small-group discussions, and collaborative framework building, participants will identify key comfort determinants, explore the relationships between vehicle behavior, user characteristics, contextual factors, and HMI designs, and discuss methodological challenges in measuring and improving passenger comfort. The workshop will culminate in a community-driven research agenda and an initial conceptual framework to guide future interdisciplinary research on adaptive, human-centered automated mobility systems.

Remote assistance is emerging as an important approach for supporting automated vehicles in complex, ambiguous, or unforeseen situations. While remote assistance differs from traditional take-over in terms of operator location, control authority, intervention level, and scalability, both concern how humans should be involved when automated driving systems. Building on the well-established body of take-over request (TOR) research, this workshop examines how human intervention can be conceptualised and structured in remote assistance contexts. Participants will compare TOR and remote assistance, identify key dimensions of intervention, and discuss how these dimensions translate into user interface design implications. The workshop will produce an initial framework, interface design considerations, and a research agenda for supporting effective, timely, and scalable human intervention in remote assistance.

Technological advances are making automotive human-machine interfaces (HMIs) a critical component of future mobility in academia, industry, and regulatory sectors. Despite substantial research activity, translating HMI research findings into automotive product development remains challenging, highlighting the need for stronger research-practice translation. Our preliminary expert interviews identified five themes related to current collaboration modes, barriers to research translation, and opportunities for stronger academia-industry partnerships. Leveraging the unique composition of the AutoUI community, which brings together researchers, practitioners, and policymakers worldwide, this workshop aims to discuss these findings and identify strategies for empowering translational research in automotive HMI. We will present insights from expert interviews, organize a mini seminar highlighting successful collaboration frameworks, and facilitate group discussions to generate actionable recommendations for bridging research and practice. Expected outcomes include refined themes, collaboration pathways, and practical guidelines to support the translation of research discoveries into product development and regulatory decision-making.

Mental workload is an important construct for understanding human cognitive states in conventional, automated, and highly automated ‎vehicles. As vehicle automation progresses, the role of the human changes from active driving to supervision or passive use, which leads ‎to different types of cognitive demands across different levels of automation. Existing approaches for assessing mental workload in ‎automotive contexts have mainly relied on subjective questionnaires and, to a lesser extent, performance-based measures. While these ‎methods have been widely used in conventional and partially automated driving, their suitability for highly automated environments is ‎still under discussion, as the nature of human–vehicle interaction is changing. This workshop focuses on mental workload assessment in ‎conventional, automated, and highly automated vehicles. It brings together researchers and practitioners to critically discuss current ‎measurement approaches and their limitations, and to explore a broader set of assessment methods, including questionnaires, behavioral ‎measures, and neurophysiological techniques. The aim is to create a shared understanding of how mental workload can be assessed across ‎different levels of vehicle automation and to identify directions for improving existing evaluation methods and tools used in this field.

Urban Air Mobility (UAM), enabled by electric vertical take-off and landing vehicles (eVTOLs), is increasingly envisioned as a future layer of urban transportation. Existing UAM research has focused mainly on technical feasibility, airspace management, infrastructure, and vehicle operation, while less attention has been paid to how UAM may reshape everyday urban life, public space, human–vehicle interaction, social acceptance, and trust in highly automated aerial systems.

This workshop invites AutomotiveUI researchers, designers, practitioners, students, and policymakers to explore human-centered and trustworthy UAM futures. It combines The Urban Air Experience Probe, a lightweight eVTOL rooftop-landing simulation centered on the final approach and touchdown phase, with the Urban Air Futures Canvas that prompts participants to examine UAM from the perspectives of passengers, pilots, and bystanders. Participants will identify context-specific tensions in touchdown scenarios and translate them into societal and trust-related challenges. Drawing on AutomotiveUI work on automated driving, the workshop then supports participants in developing trust-oriented HMI concepts and governance response strategies.

Previous takeover support has mainly relied on scene-related information, such as detected risk and remaining lead time, but these signals reflect the vehicle’s view rather than driver situation awareness, which is important for determining whether support is needed. While previous research shows the promise of using multimodal physiological data to evaluate driver situation awareness, it is still unclear how these can be integrated into real-time support systems. This demonstration then introduces Driver-Adaptive Support Tree (DAST), a real-time CARLA-based testbed that uses Lab Streaming Layer to synchronize simulator events with physiological signals and maps these data streams onto a support decision tree representing driver states. This mapping translates synchronized data into concrete support choices by linking each takeover moment to the driver’s state that requires assistance. Overall, our DAST testbed provides researchers with a repeatable platform for comparing the effects of adaptive policies in takeover scenarios.

In this study, we propose an automotive user interface (UI) that leverages tactile-audio cross-modal perception to simulate everyday tactile experiences, such as crushing a plastic bottle. Based on inputs from capacitive touch sensors, the system provides synchronized vibrotactile and auditory feedback via a vibration motor and a speaker. We developed an interactive prototype for continuous adjustment tasks in which the feedback dynamically reflects the simulated degree of physical deformation. This UI enables drivers to perceive parameter changes and interface states haptically without relying on a graphical user interface. By enabling eyes-free operation and reducing visual dependence, this system is expected to mitigate driving distractions and foster a novel user experience.

We propose +Walk, a concept that uses spare time before reaching a destination as an opportunity for exploratory walking under arrive-by constraints. To realize this concept, we built the +Walk system, a proof-of-concept implementation that combines time-budgeted routing with handheld pseudo-force haptic guidance. The routing method supports small detours while keeping the arrival time close to the planned time, and the haptic device provides turn-by-turn cues through pseudo-force sensations. +Walk opens a promising design space for time-aware, exploratory, eyes-up walking experiences.

As autonomous vehicle technologies advance, increasing portions of driving control are shifting from human drivers to automated systems. Takeover situations, where control must be safely returned to the driver, remain a critical focus for safety evaluation. These handovers depend on effective takeover requests, yet existing studies are difficult to compare because they use different simulator setups, scenario designs, and performance measures. This fragmentation creates a need for a consistent workflow for designing, running, and reporting takeover experiments. In addition, developing and testing human-machine interface designs can be resource-intensive, limiting rapid comparison of alternative alert presentations. To address these limitations, we introduce TakeoverBench, an open-source benchmarking platform for autonomous vehicle takeover research. The platform integrates scenario execution, takeover alert triggering, human-machine interface testing, telemetry capture, hardware documentation, and automated report generation. Our demonstration shows how these components support consistent experimental conditions, reduce manual data-processing effort, and improve comparison across research setups.

Driving simulators support human factors and automotive user interface research, allowing repeated testing of unsafe or future-oriented scenarios within controlled and low-risk settings. However, researchers often invest additional effort because commercial platforms represent scenarios through vehicle behavior and timing, whereas researchers define trials through experimental conditions, critical events, and cues. Human factors studies also involve physiological sensors and post-trial questionnaires, which are often managed separately, risking misalignment among simulator events, sensor data, and responses. To address these challenges, Narrate-to-Sim uses deterministic, rule-based parsing to translate supported natural-language prompts into executable scenarios and synchronize simulator logs with physiological data via Lab Streaming Layer (LSL). It features a survey module for post-trial questionnaires and allows researchers to organize trials, sensors, and questionnaires into experiment playlists for data collection. Overall, this demo presents a lightweight driving-simulator platform that reduces engineering effort and allows researchers to focus on study design.

The advent of automated vehicles (AVs) prompted AutomotiveUI researchers to investigate how AVs can safely share the road with Human Road Users (HRUs). Researchers developed interfaces to facilitate AV-HRU interaction, e.g., on-vehicle eHMIs. However, these were mostly evaluated through online surveys with stationary participants, or virtual reality (VR) setups in small indoor settings, limiting ecological validity. This video submission showcases ARena, an open-source augmented reality (AR) simulator that projects life-sized virtual AVs, interfaces, and traffic infrastructure into outdoor environments, allowing participants to move freely. ARena enables the evaluation of diverse AV behaviours, including hazardous non-yielding scenarios that are unsafe to study using real vehicles. The platform logs key behavioural measures, including speed, cadence and gaze. ARena is extensible: Researchers can create custom traffic scenario scenes without changing logging or backend logic. The video shows how ARena supports realism for AV-HRU interaction research and advances the large-scale adoption of AVs

Adolescence is a key stage when independent travel becomes more common. However, road crossing can remain challenging for autistic adolescents and those with other special educational needs (SEN), particularly where crossing support is limited or ambiguous. This video submission presents an ongoing study evaluating road-crossing infrastructure for inclusive mobility conducted in a CAVE-based pedestrian simulator that replicates the roads outside SEN schools in Leeds, UK. The study compares a no-interface baseline with three infrastructure designs: a zebra crossing, a pelican crossing, and a projection-enhanced pelican crossing. Sixty-four adolescents aged 11-13 years were recruited, including 32 autistic and 32 typically developing participants, with groups balanced by gender. Participants completed crossing trials across two environments: a two-lane road and a two-lane road with a cycle lane, simulating two real school environments in Leeds. The study aims to provide evidence-based recommendations for inclusive school crossing infrastructure and to support local traffic engineers in designing safer road environments for children with diverse needs.

As personal mobility (PM) ridership surges, contemporary urban planning focuses heavily on dedicating isolated lanes, often overlooking the continuity of the seamless transit experience. To address these fragmented transfer points between public transit and micro-mobility, this paper introduces ‘Surf Road’, a novel paradigm in urban transportation. While modern cities are physically connected, a rider’s journey is often abruptly disrupted at the boundaries of macro-infrastructure. Accompanied by a video demonstration, we present a symbiotic PM system that organically docks with moving tram networks, transforming a stagnant commute into a playful, continuous “urban surfing” experience. Without requiring prohibitive infrastructure overhauls, Surf Road seamlessly coexists with existing transit grids to deliver a high-continuity user experience (UX), offering a futuristic framework for human-infrastructure symbiosis.

Most navigation systems tell users how to arrive fastest if they leave now. Yet in flexible multi-stop journeys, the best strategy may be counterintuitive: waiting briefly, leaving later, or reordering stops can reduce exposure to predicted congestion and improve the overall journey. This video presents a navigation concept for strategic departure timing. Through a narrated in-vehicle scenario, it demonstrates how a navigation system compares immediate departure with alternative timing strategies. A proof-of-concept prototype supports the scenario by using temporal activity patterns as time-dependent delay signals and a Genetic Network Programming-based planner to explore route order, departure timing, and dwell-time decisions. Finally, we highlight future design opportunities for in-vehicle intelligent agents to transparently communicate the rationale, benefits and alternatives behind recommendations.

Takeover warnings in conditionally automated vehicles are typically triggered when the automated driving system detects that it is approaching its operational limits. However, without evidence of what drivers have perceived, interpreted, and anticipated about traffic events, such warnings may not provide precise support. While physiological modeling can provide non-intrusive, real-time evidence of driver situation awareness (SA), previous studies using physiological signals often represent driver SA using a single score, obscuring whether support should address spatial risk location or temporal hazard progression. This video presents target-specific SA modeling across four SA targets organized by these two event dimensions: forward-path and adjacent-lane SA for spatial risk location, and materialized- and unmaterialized-hazard SA for temporal hazard progression. In the video, we present an adaptive takeover support strategy based on target-specific SA modeling and our preliminary video-based takeover study. Overall, we demonstrate how fine-grained driver SA assessment can guide adaptive takeover support.

Current AI-based conversational agents (CAs) largely lack situational awareness, limiting their ability to understand driving context, driver state, and immediate driver needs. Consequently, the use of CA may increase drivers’ cognitive load and impair safety and user experience. This study investigates the role of situationally aware CA in long-haul trucks and how such CA should be designed to support drivers in an efficient way. The insights were gathered through a Wizard-of-Oz driving simulator study exploring two CA concepts. The results show that CA are well-suited for strategic tasks, could be useful for some tactical tasks depending on the cognitive load, and are generally inappropriate at the operational level, where they interfere with core driving actions. Effective CA design must prioritize driving as the primary task, adapt interaction timing to drivers’ cognitive load, and ensure transparency and predictability to build trust.

Drivers in conditionally automated vehicles must rebuild situation awareness (SA) when automation reaches its limits. Under the attention-SA model, rebuilding requires sampling driving-scene cues after takeover requests. Non-driving-related tasks can divert attention, requiring assessments to identify which cues drivers sampled. Prior scene- and object-level measures omit cue-level information about what drivers perceived, interpreted, and anticipated: the former provide only scene-wide judgments; the latter report detection alone. To address this gap, 33 drivers completed a video-based study assessing takeover-relevant cue-level SA along two dimensions: Spatial Direction (longitudinal versus lateral) and Event State (precursor versus materialized). A generalized linear mixed-effects model analyzed probe correctness. Results showed that drivers answered lateral-cue probes more accurately than longitudinal-cue probes and materialized-hazard probes more accurately than hazardous-precursor probes. The materialized-versus-precursor difference appeared only for lateral probes; longitudinal-probe correctness was similar across levels. These findings may inform development of cue-aware takeover support for next-generation automated vehicles.

Autonomous vehicles (AV’s) can improve transportation accessibility and safety, yet public adoption remains a challenge for carmakers. Visual connection to nature is known to reduce anxiety, improve mood, restore mental bandwidth, and create more desirable environments for humans; Edward Wilson’s Biophilia Hypothesis holds that an essential aspect of human well-being is a connection to the natural world. Building on prior research on natural elements in interior spaces, this study hypothesizes that a navigational user interface (NUI) incorporating nature-based visuals would increase AV passenger comfort in terms of positive emotion, perceived restoration, and connection to nature. In the study, 12 participants viewed a virtual reality simulation to evaluate the effect of three nature visuals (Water, Wildlife, and Weather) within the vehicle’s NUI. Participants reported a 74% increase in positive affect (Weather), a 57% increase in perceived restoration (Weather), and a 95% increase in connection to nature (Wildlife), compared to a blank interface (Friedman tests, all p < .01). Participants most preferred a depiction of natural cycles, reproducing sunlight and environmental change, and the movement of water ripples. By incorporating natural elements into NUI’s, this study introduces one strategy to expand AV interactions into a psychologically preferable encounter.

Transitions from automated to manual control remain a critical challenge in L2/L3 automated vehicles, where sudden or poorly explained takeover requests can create confusion, anxiety, delayed reactions, and reduced trust. This student research project explores how automotive human-machine interfaces (HMIs) can support driver readiness during system-to-driver handover. In this paper, takeover refers to a system-initiated request for the human driver to resume manual control, while handover refers to the broader transition of control from automation back to the driver. Driver readiness is understood as the driver’s preparedness to reorient to the driving task, understand why intervention is needed, and begin manual control. The work originated in a graduate Human-Computer Interaction (HCI) course and followed a user-centered design process including background research, qualitative user research, persona development, user stories, ideation, low- and high-fidelity prototyping, and immersive concept visualization. Based on insights from cautious commuters and technology-aware drivers, we designed the Calm Handover System, a specific multimodal takeover interface that combines phased visual alerts, predictive takeover-zone visualization, auditory cues, and haptic feedback. The contribution of the work lies in reframing takeover as a staged, explainable, and design-led transition rather than a single sudden alert. The project is presented as an early-stage student design exploration rather than a validated experimental system.

This study evaluated how visual variables in Forward Collision Warning (FCW) following-distance designs were associated with drivers’ visual attention. Four display designs were tested in a prerecorded highway-driving scenario with two non-driving-related tasks. Eye-tracking data from 15 participants were analyzed using heatmap visualization and linear mixed-effects models. Thin-stripe was associated with a shorter time to first fixation on the FCW area of interest (AOI) than Wide-stripe and a higher overall glance frequency than the other displays. Percentage dwell time (PDT) revealed AOI-specific viewing patterns: PDT in the FCW AOI was higher for Thin-stripe and Elliptical than for Wide-stripe, whereas Radar showed greater PDT in the WINDOW AOI than the other conditions. Results suggest distinct gaze patterns across displays, potentially related to texture and orientation, but remain exploratory because display order was fixed and visual variables were not independently manipulated.

Automated vehicles (AVs) increasingly rely on connected communication systems, which make them vulnerable to cyber-attacks. During such events, drivers may not understand the situation and might need additional assistance. Although driving agents can provide warnings, large language model (LLM)-based driving agents may offer deeper explanation and follow-up support. Prior work has examined cyber-attacks in vehicles and the role of in-vehicle intelligent agents, but limited research has investigated how different agents specifically LLM-based driving agents affect driver response during AV cyber-attacks. This ongoing study proposes a driving simulator experiment comparing low-critical and high-critical cyber-attacks in three agent conditions: no agent, a rule-based driving agent, and an LLM-based driving agent. The rule-based driving agent provides short cyber-attack warnings and action instructions, while the LLM-based driving agent provides conversational explanations and follow-up question-answer support. Behavioral and subjective measures will be collected. The expected contribution has implications in the field of Human-computer interaction for the future design of driving agents according to the risk of cyber-attacks.

A fully automated (SAE L5) vehicle can obey every traffic rule and still behave near-erroneously, and a passenger who cannot tell why loses situation awareness (SA) to the black-box driving stack, degrading the ride. Prior natural-language explanations target drivers or stop at model interpretability, not a passenger-facing account of the vehicle’s own error. This paper designs such an explainable-AI (XAI) interface and builds a simulator testbed to validate it before any human study. We frame a scenario’s error as the functional and output insufficiency (FI/OI) defined by ISO 21448 Safety of the Intended Functionality (SOTIF), and organize a Vision-Language-Action (VLA) model’s reasoning trace into Endsley’s three SA stages. The first contribution is a reusable design that maps this reasoning onto SA stages so the vehicle explains its own FI/OI. The second is a Passenger-in-the-Loop testbed coupling CARLA, a 6-DOF motion platform, and live visual and voice Human-Machine Interfaces (HMIs) over WebSocket, with six metrics as a scenario quality-assurance instrument. In a human-free dry run of 100 replays per scenario, the roundabout-deadlock scenario ran collision-free with deterministic metrics (CV  ≤  0.4%), while the aquaplaning scenario fired its terrain-transition events in every replay and is characterized by its multi-run distribution. We outline a follow-up human-subject study to test affect induction and user experience.

Driver assistance and partially automated driving can reduce active control demands on drivers, but they may become distracted or fatigued and must remain ready to supervise or re-engage. Although driver-monitoring systems have been studied extensively, direct camera-based systems are often vehicle-integrated and not uniformly available. We present a low-cost, supplementary Android smartphone interface that uses the smartphone’s front-facing camera as a supplementary driver-monitoring and alerting interface. The prototype combines on-device face analysis and TensorFlow Lite object detection to identify six states: attentive baseline, yawning, prolonged eye closure, phone-like interaction, reading, and face-not-detected. The system was evaluated in a controlled pilot study with 30 participants across 180 trials, measuring classification performance, alert latency, and post-alert recovery. The system detected multiple target behaviours, with strongest performance for prolonged eye closure and face-not-detected, while phone-like interaction and reading remained more challenging. This work contributes early HCI evidence for using smartphones as interactive driver-supervision interfaces, highlighting trade-offs between low-cost sensing, alert timing, user trust, intrusiveness, and attention recovery.

Mobility systems are reshaping how vehicles are accessed, shared, and maintained, yet vehicle development remains largely grounded in assumptions established for private ownership. As shared mobility, autonomous services, and fleet-based operations expand, vehicles increasingly operate under conditions characterized by multiple users, repeated short-term use, and operator-managed care. However, limited attention has been given to translating these changing operational contexts into vehicle design during concept development. This paper investigates how recurring operational conditions can inform early-stage vehicle design. Through a literature review and comparative practice review, the study identifies common operational conditions across emerging mobility systems and proposes a structured translation framework that interprets these conditions through interaction and formulates initial design criteria. The findings suggest that recurring operational conditions provide a stronger basis for early-stage vehicle design than mobility service categories alone. The proposed framework offers a systematic approach for incorporating emerging mobility conditions into vehicle design during concept development, contributing to closer integration between mobility transition, Automotive HCI, and automotive design.

Autonomous vehicles need to respond safely to pedestrians, yet the behavior of school-age children remains difficult to characterize and is only weakly represented in many datasets and prediction-oriented research streams. Prior work has shown that children differ from adults in attention, timing, and decision-making during street crossing, and that current pedestrian detection systems can also perform worse on children than on adults. At the same time, much of the existing literature on pedestrian intention prediction focuses on adult pedestrians, egocentric traffic scenes, or individual behavior rather than small child groups.

This work-in-progress examines how school-age child group crossings can be described from drone-based bird’s-eye-view recordings for qualitative behavioural research in automated driving contexts. Based on a controlled field study, the paper presents a structured observational account of single-directional and bidirectional child crossing scenes, focusing on waiting, scanning, initiation, following, cohesion, and lateral movement patterns. Across scenes, crossings often appeared to be socially organized, with a small number of children initiating movement while others followed with limited visible safety checks. Rather than proposing a predictive model, the paper offers an initial descriptive basis for future work on underrepresented pedestrian behaviour, trajectory analysis, and child-centered modelling questions.