Long-Term Evolution of Driver Visual Attention during Automated Driving in Real-Traffic: Investigating the Influence of Mental Model and Dynamic Learned TrustA calibrated trust level is essential for the safe use of automated systems. In automated driving, overtrust can reduce the driver’s monitoring behavior and delay takeover times, which poses significant safety risks. This motivates the need for continuous, objective trust assessment for real-time system adaptations. Prior research identified eye-tracking as a promising approach. Therefore, this study examines the longitudinal relationship between dynamic learned trust and visual attention. Given that mental models influence both trust and visual attention, their role in this process is also explored over time. In a longitudinal study, twenty-three participants repeatedly operated an automated vehicle in real traffic while their visual attention was recorded via the vehicle’s built-in driver monitoring camera. Findings suggest an interrelation between dynamic learned trust and mental model formation, with mental models mediating the effect of dynamic learned trust on visual attention. This work contributes to advancing trust measurement during automated driving.2025SSStephanie Seupke et al.Automated Driving Interface & Takeover DesignEye Tracking & Gaze InteractionCHI
VR-WISE: VR based work zone immersive simulation with eye tracking and biometric sensors for capturing driver’s awarenessEnsuring safety in roadway work zones remains a major challenge, as current measures have not fully eliminated risks. Accidents around work zones often stem from the unpredictable behavior of both drivers and workers, highlighting the need to better understand their interactions and perceptions in these environments. Advances in virtual reality (VR) and high-fidelity simulation platforms like CARLA make it possible to observe behaviors in hazardous scenarios without exposing participants to real-world danger. This paper introduces VR-WISE, a VR-based traffic co-simulation platform designed to capture rich driver behavior data around work zones. The system integrates full-scene customization, animated roadway workers, and real-time driver monitoring through eye-tracking and biometric sensors. VR-WISE supports rapid work zone scenario generation and enables the collection of detailed multimodal data to study how drivers perceive and respond to dynamic roadside activity. A user study across three distinct work zone scenarios demonstrates the platform’s capabilities. Drivers' perception around work zones was evaluated using gaze metrics (e.g., duration and fixation ratios), along with physiological signals such as heart rate, electrodermal activity (EDA), and blood volume pulse (BVP), allowing for a more comprehensive understanding of drivers' attention.2025SZSHUO ZHANG et al.Eye Tracking & Gaze InteractionContext-Aware ComputingSmart Cities & Urban SensingCHI
Immersive Augmented Reality (AR) Gaming in Vehicles: The Impact of Visuo-Vestibular Congruency on Motion DiscomfortThis study investigated rear-seat passengers’ motion discomfort when displaying (in)congruent visual motion while playing an immersive Augmented Reality (AR) racing game on a headrest-mounted screen. 29 players participated in two, 30-minute drives involving urban and highway roads. In the Synchronized Game (SG), the gameplay elements and video background were live-streamed from the vehicle’s cameras and sensors, creating a congruent sensory environment. In the Desynchronized Game (DG), the gameplay was pre-recorded, resulting in incongruent visuo-vestibular motion where the visual motion in the game did not always align with that of the vehicle. Motion discomfort was measured at 2-min intervals using the MIsery SCale (MISC) and a thermal camera measuring participants’ forehead temperature. The results showed that the SG condition led to significantly lower motion discomfort compared to the DG condition. These findings suggest that immersive games that incorporate real-time vehicle motion can help to mitigate motion discomfort by providing congruent visual-vestibular input.2025SDStephanie DabicMotion Sickness & Passenger ExperienceSocial & Collaborative VRCHI
Need for Trust Calibration in Takeover request Performance in Level 3 Automated vehiclesTrust is a critical human factor influencing driver interaction with autonomous vehicles (AVs), particularly during takeover requests (TORs). Despite growing interest in trust calibration and TOR performance, no comprehensive review exists that synthesizes findings across these domains. This paper aims to address this gap by systematically reviewing the relationship between trust and TORs in Level 3 AVs. We examine how factors such as TOR timing, warning modalities, environmental conditions, driver traits, and system malfunctions impact trust dynamics and takeover performance. Additionally, we explore the role of trust calibration, its formation, miscalibration (overtrust or undertrust), and recovery mechanisms, in shaping effective human-automation interaction. By integrating insights from existing studies, we identify research gaps, including the need for adaptive TOR strategies based on real-time trust monitoring and individual differences. This review provides actionable recommendations for designing AV systems that optimize trust calibration, enhance safety, and improve user acceptance of automated driving technologies.2025JJJulakha Jahan Jui et al.Automated Driving Interface & Takeover DesignCHI
Evaluating Interfaces for Non-Driving Related Tasks While Operating an E-scooterMicromobility vehicles, such as e-scooters, provide ecological and financial benefits over automotive transportation. However, as with car drivers, micromobility users often perform non-driving related tasks (NDRTs), interacting with stereo controls or navigation tasks, which can lead to accidents. It remains unclear what control interfaces are appropriate and safe for micromobility. We evaluated six interface modalities for NDRTs and conducted a within-subjects study with 35 participants (yielding n=210 observations) in an e-scooter simulator to compare modality safety and preferences. Our results align with existing work on gaze and tactility in the automotive NDRTs context. However, unique to e-scooters, interfaces that required users to alter their grip on the handlebars were less preferred as they compromised stability. Social comfort also emerged as a critical factor due to concerns about public visibility. This work aims to encourage the design of safer, more socially acceptable interfaces for e-scooters and other emerging micromobility vehicles.2025KTKenshikimyo Terao et al.In-Vehicle Haptic, Audio & Multimodal FeedbackMicromobility (E-bike, E-scooter) InteractionCHI
SOH Illusion: Misunderstandings of EV Battery State of Health and Methods to Promote UnderstandingLegislation in the USA will soon require that electric vehicles (EVs) display battery degradation in the instrument cluster as "state of health" (SOH), the percentage of the battery's original capacity. However, the extent to which consumers understand SOH degradation patterns is not known. In an initial study with vehicle owners, we find preliminary evidence for a 'SOH illusion', wherein people expect linear rates of EV battery degradation over time even though batteries degrade non-linearly. Additionally, a third of participants incorrectly conflated SOH with a battery's remaining usable life, demonstrating some misunderstanding of SOH among vehicle owners. In a follow-up study we find that framing SOH information with different chart types and legends reduces linear degradation assumptions and aligns people's expectations. We discuss implications for the design of SOH representations in user interfaces that vehicle UI designers could employ to promote better EV battery understanding.2025KLKylie R. Lin et al.EV Charging & Eco-Driving InterfacesInteractive Data VisualizationCHI
Haptic-Augmented AV Experiences: Potentials for Blind and Low-Vision UsersHaptic technology has diverse applications in automated vehicles (AVs), yet research lacks a holistic view of its role in user experiences and its inclusive potential for blind and low-vision (BLV) users. This paper reviews state-of-the-art haptic interfaces in AVs, examining technological foundations, applications, user experience considerations, and adaptability for BLV accessibility. We found that existing haptic interfaces are primarily designed for drivers during automation transitions, emphasizing effectiveness over hedonic experience. There is a knowledge gap in how such interfaces can improve BLV user experience in fully automated vehicles. We propose a shift from haptic-supported automated driving to Haptic-Augmented AV Experiences, advocating for more inclusive and adaptive haptic interactions beyond traditional driver-centric paradigms.2025ZMZhengtao Ma et al.Automated Driving Interface & Takeover DesignIn-Vehicle Haptic, Audio & Multimodal FeedbackVisual Impairment Technologies (Screen Readers, Tactile Graphics, Braille)CHI
Examining Cross-Cultural Differences in Intelligent Vehicle Agents: Repair Strategies after Their FailuresAnthropomorphic design in intelligent vehicle agents (IVAs) is crucial for driving safety and user experience. Cultural background may shape user preferences, as evidenced by Chinese car manufacturers offering more anthropomorphic IVAs (e.g., physical robots, human-like virtual agents) than their Western counterparts. This suggests that a universal approach to anthropomorphic design may not be feasible. While prior academic research has examined cross-cultural differences in visual anthropomorphism, behavioral anthropomorphism remains understudied. In this study, we developed a taxonomy of user requests (N = 60), evaluated the performance and responses of eight IVAs in premium-level cars in the Chinese market (five from Chinese brands, three from Western brands), and analyzed their verbal repair behaviors (e.g., apology, promise) following request failures. Overall, the five Chinese-brand IVAs and three Western-brand IVAs did not differ in their corrective responses to user requests or their likelihood of employing verbal repair strategies. However, our in-depth analysis revealed that Chinese-brand IVAs were more likely to use combined repair strategies rather than single ones and to incorporate intimacy expressions in their verbal repair behaviors compared to their Western-brand counterparts. This suggests potential cross-cultural differences in the design of social strategies for IVAs. We also observed IVA-level variations within both Chinese-brand and Western-brand groups. Future cross-cultural research is needed to inform evidence-based anthropomorphic design.2025LLLan Lan et al.External HMI (eHMI) — Communication with Pedestrians & CyclistsVoice User Interface (VUI) DesignMultilingual & Cross-Cultural Voice InteractionCHI
Mind Over Matter - Investigating the Influence of Driver's Perception in the Misuse of Automated VehiclesAs vehicles with several levels of automation become increasingly common, there is an increase in incidents involving the misuse of Driving Automation Systems (DAS). The manner in which drivers interact with DAS indicates that the problem extends beyond UI design. We investigate how drivers' perceptions and expectations affect the understanding and consequent usage of DAS. The study employed a Wizard-of-Oz approach to simulate a vehicle with a Level 2 and Level 3 DAS on a public highway. Sixteen participants were exposed to the two driving modes and two distinct UIs. Observations, think-aloud protocols, and in-depth interviews documented their interaction with the different DAS. Irrespective of the UI, various errors were detected, including omission, commission, mode confusion. Deeper investigation into the sources led to the conclusion that drivers' preconceptions of the DAS were a major contributor, resulting in misuse. This highlights the need to look beyond UI design as a sole solution to address driver-vehicle interaction.2025FNFjollë Novakazi et al.Head-Up Display (HUD) & Advanced Driver Assistance Systems (ADAS)AI-Assisted Decision-Making & AutomationCHI
Non-Emergency Notification Timing for Drivers Doing Non-Driving-Related Tasks in Autonomous Vehicles: An Interruptibility StudyFuture high-level autonomous vehicles (AVs) will enable drivers to engage in non-driving-related tasks (NDRTs) during autopilot. Occasionally, an in-vehicle agent may need to notify drivers of important, yet not urgent, information. Through a four-session interruptibility study on a desktop autonomous driving simulator, we investigated how drivers assess their availability to receive notifications by rating moments as good or bad for interruption. Our results suggest drivers fall into four notification availability groups: always available, prioritizing NDRTs, task-content dependent, and mental-state dependent. Using multimodal behavioral data of the participants and vehicle data from the simulation, we trained a proof-of-concept classification model to determine the appropriate timing to send non-emergency notifications to drivers doing NDRTs. Head pose and gaze direction data from the eye tracker were crucial in the predictions. Based on our quantitative modeling and qualitative observation, we discuss the feasibility of notification timing prediction in the real world and design considerations from individual, task, and context perspectives.2025HWHongyu Howie Wang et al.In-Vehicle Haptic, Audio & Multimodal FeedbackEye Tracking & Gaze InteractionNotification & Interruption ManagementCHI
Drivers’ Attention to Dash-Based Human-Machine Interfaces: The Effect of Partial Automation and Cognitive LoadA vehicle’s dash-based Human-Machine Interface (HMI) provides critical information to drivers. However, the location of these displays can take drivers' visual attention away from the forward view and compromise safety. As vehicle automation becomes increasingly common, its impact on drivers’ visual attention to dash-based HMI remains under-explored. Moreover, drivers tend to engage more frequently in non-driving-related tasks during automation, but how the cognitive load imposed by these tasks affects drivers’ inspection of HMI displays is unclear. This driving simulator study examined how partial automation and cognitive load (imposed by a 2-back task) influence drivers’ visual attention to dash-based HMI containing speed and automation status information (N=46). Results showed that increased levels of automation and cognitive load additively reduced drivers’ visual attention to the dash area. Drivers prioritized inspecting the speedometer over the automation status information across all conditions. Our findings provide important implications for HMI design in automated vehicles.2025HQHao Qin et al.Head-Up Display (HUD) & Advanced Driver Assistance Systems (ADAS)CHI
Seeing Beyond the Leading Vehicle: Designing V2V Braking Visualizations to Support Novice DriversFor novice drivers, highway driving often makes it challenging to anticipate sudden braking events and adapt to the dynamic nature of traffic conditions. Vehicle-to-vehicle (V2V) communication technologies offer a promising solution by providing beyond-line-of-sight (BLOS) braking information, enhancing situational awareness and reducing collision risks. This study evaluates the effectiveness of two visualization strategies for V2V-based braking information: (1) braking cues from the indirect vehicle (ILV-HMI), and (2) upstream traffic flow braking information (UTF-HMI). Both were compared to a baseline condition (NO HMI) without additional information. A controlled driving simulator experiment with 24 novice drivers assessed driving safety, driving performance, cognitive load, and usability. Results show that ILV-HMI significantly improves safety margins, reduces cognitive load, and enhances usability compared to NO HMI and UTF-HMI. These findings highlight the advantages of V2V-based braking visualization in improving novice drivers’ decision-making, easing cognitive load, and enhancing traffic safety.2025HSHongling Sheng et al.V2X (Vehicle-to-Everything) Communication DesignCHI
Designing With Motion: Exploring Vestibular Cues as a Subtle Awareness Nudge Modality in Automated VehiclesAutomated driving systems, particularly at SAE Level 3, present new challenges in managing driver attention to ensure smooth transitions from automated to manual control. This paper reports on a qualitative investigation of vestibular cues—implemented via subtle deceleration events—as a form of a dynamic Human-Machine Interface (dHMI) that subtly "nudges" a user's attention away from a non-driving related task (NDRT) and towards the driving environment. Conducted as a test-track study (N=25), we explore awareness, acceptance, and design considerations related to these cues. Findings reveal that while participants showed positive attitudes toward vestibular nudges as safety features, they were unable to differentiate nudges from necessary vehicle deceleration during automated driving. The study reveals how drivers interpret the implicit interaction with the dHMI during realistic NDRT and potential limitations. The study highlights the need for multimodality HMI approaches and customisation to optimise user experience in conditional automated vehicles.2025YYYueteng Yu et al.Automated Driving Interface & Takeover DesignIn-Vehicle Haptic, Audio & Multimodal FeedbackCHI
Enhancing Cyclist Safety in the EU: A Study on Lateral Overtaking Distance Across Seven Scenarios Using Lab and Crowdsourced MethodsCyclists face significant risks from vehicles that overtake too closely. Through crowdsourcing (N = 200) and driving simulator (N = 20) experiments, this study examines driver behaviour in seven scenarios: laser projection, road sign, road marking, car projection, centre line and side line markings (baseline), cycle lane and no road markings. Crowdsourced participants consistently underestimated overtaking distances, particularly at wider gaps, despite feeling safer with greater distances. The simulation results showed that drivers maintained an average passing distance of 3.4~m when not constrained by traffic, exceeding the 1.5~m law of the European Union. However, interventions varied in effectiveness: while laser projection was preferred, it did not significantly increase passing distance. In contrast, a dedicated cycle lane and a solid centreline led to the greatest improvements. These findings highlight the discrepancies between perceived and actual safety and provide insight for policy interventions to enhance cyclist protection in the EU.2025GSGiovanni Sapienza et al.External HMI (eHMI) — Communication with Pedestrians & CyclistsV2X (Vehicle-to-Everything) Communication DesignPedestrian & Cyclist SafetyCHI
Visual Sampling Behavior Does not Explain Risk Perception: A Data-Driven xAI InvestigationHow do drivers perceive risk? Understanding what situations and factors cause drivers to perceive situations as critical can improve our understanding of road user behavior and inform automated driving technology. To investigate the factors that shape drivers’ risk perception, we conducted an eye-tracking study with 27 participants who watched dashcam videos and continuously rated the perceived risk of various driving situations. Using the resulting dataset, we developed a computer vision-based machine learning approach that generates explainable predictions of perceived risk from video and eye-tracking data. Our SHAP analysis reveals that the proximity of objects and number of cars in a scene are the most significant contributors to perceived criticality. Most interestingly, while people tend to sample similar objects in critical situations, their risk perception remains highly personal making visual sampling behavior a weak predictor of perceived risk. Overall, our explanations reveal non-linear insights beyond previous work, suggesting that risk perception is not only shaped by visual input, but primarily by cognitive processes which is in line with theoretical models of Situation Awareness. The dataset, source code, and a comprehensive usage guide are publicly available: https://osf.io/cwd6h/?view_only=31a8173570de4b0383f55d52dc784492.2025MLMartin Lorenz et al.Eye Tracking & Gaze InteractionExplainable AI (XAI)AI-Assisted Decision-Making & AutomationCHI
Pedestrian Planet: What YouTube Driving from 233 Countries and Territories Teaches Us About the WorldPedestrian crossing behaviour varies globally. This study analyses dashcam footage from the PYT dataset, covering 133 countries, to examine decision time to cross, crossing speed, and contextual variables, including detected vehicles, traffic mortality, GDP, and Gini. Bulgaria had the longest decision time (10.50 s), while San Marino exhibited the fastest crossing speed (1.14m/s). A global negative correlation between speed and decision time (r = -0.54) suggests that more cautious or uncertain pedestrians cross more slowly. Regional differences reveal stronger inverse correlations in Europe and North America, likely due to varying infrastructure, regulation, and cultures. Pedestrian decision time is positively correlated with the presence of other road users, especially bicycles (r = 0.35). Similar crossing times in countries with different infrastructures, such as Belgium and India, underscore the complex interaction between infrastructure and behavioural adaptation. These findings emphasise the importance of culturally aware road design and the development of adaptive interfaces for vehicles.2025MAMd Shadab Alam et al.External HMI (eHMI) — Communication with Pedestrians & CyclistsV2X (Vehicle-to-Everything) Communication DesignPedestrian & Cyclist SafetyCHI
Socially Adaptive Autonomous Vehicles: Effects of Contingent Driving Behavior on Drivers' ExperiencesSocial scientists have argued that autonomous vehicles (AVs) need to act as effective social agents; they have to respond implicitly to other drivers' behaviors as human drivers would. In this paper, we investigate how contingent driving behavior in AVs influences human drivers' experiences. We compared three algorithmic driving models: one trained on human driving data that responds to interactions (a familiar contingent behavior) and two artificial models that intend to either always-yield or never-yield regardless of how the interaction unfolds (non-contingent behaviors). Results show that a familiar contingent behavior significantly reduces drivers' hesitance and stress when interacting with AVs. The direct relationship between familiar contingency and positive experience indicates that AVs should incorporate socially familiar driving patterns through contextually-adaptive algorithms to improve the chances of successful deployment and acceptance in mixed human-AV traffic environments.2025CYChishang "Mario" Yang et al.Automated Driving Interface & Takeover DesignAI-Assisted Decision-Making & AutomationCHI
Designing Adaptive AV Interfaces: Linking Acceptance Profiles to Design Preferences for Enhanced AdoptionTechnology Acceptance Models (TAMs) offer valuable insights into AV user acceptance, yet little research translates these factors into design requirements for partial and full autonomy (PAV/FAV). SOM clustering of 284 surveys revealed distinct acceptor and rejector profiles, with notable differences in performance expectancy, self-efficacy, and anxiety. Rejectors exhibited “autonomy sensitivity,” with increased demands for customization, redundancy, and experientiality in FAVs, while Acceptors maintained stable preferences. These findings inform our proposed Profile–Context Interaction (PCI) framework for dual-adaptive interfaces. The PCI framework recommends four design quadrants, Acceptor–PAV, Acceptor–FAV, Rejector–PAV, and Rejector–FAV to tailor interface features to both user profiles and autonomy levels, thereby bridging the gap between acceptance theory and actionable design.2025BCBenjamin T Cham et al.Automated Driving Interface & Takeover DesignAI-Assisted Decision-Making & AutomationCHI
Quantifying Customer Preferences for Active Haptic Feedback in Automotive Steering Wheel Control ButtonsControl elements with active haptic feedback have become established in modern automotive user interfaces, but numerous media reports and studies indicate customer dissatisfaction with the current design. To investigate the customer preference that has not been well-characterized so far and to deduce whether differences in preference can be linked to a driving task or customer attributes, a Preliminary-Study with adjective rating, an Expert-Study with pairwise comparison, and a Customer-Study with an additional real driving task were conducted. The results show different customer preference groups, with the vast majority favoring short haptic feedback within 13.6 ms and 20.6 ms length. A driving task does not influence preference, nor is the preference affected by attributes such as gender, age, or thumb size. These findings can be used to optimize active haptic feedback according to customer preferences. As a result, well-designed haptics can increase customer satisfaction and the perceived value of automotive controls.2025MSMax Stölzle et al.In-Vehicle Haptic, Audio & Multimodal FeedbackVibrotactile Feedback & Skin StimulationCHI
Simulating Multiple Road User Perspectives on Autonomous Vehicle BehaviorsThis work presents a system and a study in which we have multiple road users interact simultaneously with an autonomous vehicle (AV) in a virtual reality (VR) environment. We go beyond studying dyadic interactions (e.g., AV-pedestrian or AV-driver) to involve a pedestrian, a human driver, and an AV passenger all jointly interacting with an AV in the same VR scenario. We probed multiple user perspectives with two different prototypes of AV behavior strategies in ambiguous stop-sign intersections. An efficient AV attempts to enter the intersection as soon as it can without collision, while a prosocial AV waits for other road users to pass before proceeding. We recruited 16 three-person groups (N=48), where half interacted with the first AV type and the other half interacted with the second AV type in four different traffic configurations. Our investigation demonstrates that road users in different roles can have diverging preferences and trust levels in the same AV behavior when making joint decisions. Finally, we discuss how our methods and findings can be used to guide further explorations for AV interaction research with multiple agents in different roles.2025JJJiHyun Jeong et al.Automated Driving Interface & Takeover DesignExternal HMI (eHMI) — Communication with Pedestrians & CyclistsTeleoperated DrivingCHI