Guy Hoffman

Eye contact between strangers, even fleeting, can spark interaction and foster connection, happiness, and belonging. Yet in public spaces, such encounters are often suppressed by “civil inattention,” with many people absorbed in their phones. We explore how reconfiguring the ambient environment with MirrorBot, a mobile robot with adaptive mirrors, can encourage social encounters by subtly redirecting glances. By shifting reflections between self- and mutual recognition, MirrorBot invites serendipitous eye contact, shared awareness, and low-stakes engagement. In a controlled 2×2 between-subjects study with 90 participants (45 dyads) across four conditions (MirrorBot, Bot-only, Mirror-only, and Control), we found that MirrorBot led participants to initiate conversation more often, feel greater closeness and togetherness, and have more enjoyable interactions. Our findings position robots not only as social agents but as socio-spatial interfaces that choreograph sight lines and shared attention in physical space, opening new possibilities for technologies that cultivate human connection in public life.

Brief eye contact with strangers can foster connection, belonging, and positive affect, yet such moments are often scarce in public spaces. This paper investigates how a spatially situated robot can reshape the visual field of a shared space to influence how strangers notice and respond to one another. We present MirrorBot, a mobile robot equipped with two actuated mirrors that dynamically redirect reflections to reshape sightlines between people. In a study with 32 strangers in 16 pairs in a waiting-room setting, MirrorBot elicited patterns such as low-stakes icebreaking, nonverbal synchrony, joint sensemaking, asymmetric engagement, and avoidance. Participants also attributed multiple roles to the robot, such as mediator, observer, magnifier, or disrupter, revealing that its social meaning was fluid and co-constructed. Our work extends HRI by showing that robots can act not only as conversational partners but also as spatial mediators, curating opportunities for human–human connection through the reconfiguration of spatial relationships

The question “what is a robot?” has long been contested as automated embodied systems encompass many forms. We examine this fundamental question in Human-Robot Interaction through the case of Major League Baseball’s “robot umpire,” officially known as the Automated Ball-Strike System (ABS). Drawing on the concept of “fluid technology,” we analyze how the robot umpire is not a fixed technological artifact but a fluid sociotechnical assemblage whose definition and function are continuously negotiated. Through ethnographic fieldwork and interviews with stakeholders across the baseball ecosystem, we demonstrate that the robot umpire’s physical boundaries, operational parameters, and authorship remain contested and evolving, shaped by ongoing interactions between technology developers, league officials, umpires, players, and fans. Our findings reveal that treating robots as fluid technologies—rather than as discrete objects—opens new possibilities for understanding human-robot relationships. We contribute both theoretical insights regarding the ontological flexibility of “robots” and methodological approaches for studying and designing robots as sociotechnical assemblages.

Robots fail, potentially leading to a loss in the robot's perceived reliability (PR), a measure correlated with trustworthiness. In this study we examine how various kinds of failures affect the PR of the robot differently, and how this measure recovers without explicit social repair actions by the robot. In a preregistered and controlled online video study, participants were asked to predict a robot's success in a pick-and-place task. We examined manipulation failures (slips), freezing (lapses), and three types of incorrect picked objects or place goals (mistakes). Participants were shown one of 11 videos -- one of five types of failure, one of five types of failure followed by a successful execution in the same video, or a successful execution video. This was followed by two additional successful execution videos. Participants bet money either on the robot or on a coin toss after each video. People's betting patterns along with a qualitative analysis of their survey responses highlight that mistakes are less damaging to PR than slips or lapses, and some mistakes are even perceived as successes. We also see that successes immediately following a failure have the same effect on PR as successes without a preceding failure. Finally, we show that successful executions recover PR after a failure. Our findings highlight which robot failures are in higher need of repair in a human-robot interaction, and how trust could be recovered by robot successes.

Robots fail, potentially leading to a loss in the robot's perceived reliability (PR), a measure correlated with trustworthiness. In this study we examine how various kinds of failures affect the PR of the robot differently, and how this measure recovers without explicit social repair actions by the robot. In a preregistered and controlled online video study, participants were asked to predict a robot's success in a pick-and-place task. We examined manipulation failures (slips), freezing (lapses), and three types of incorrect picked objects or place goals (mistakes). Participants were shown one of 11 videos -- one of five types of failure, one of five types of failure followed by a successful execution in the same video, or a successful execution video. This was followed by two additional successful execution videos. Participants bet money either on the robot or on a coin toss after each video. People's betting patterns along with a qualitative analysis of their survey responses highlight that mistakes are less damaging to PR than slips or lapses, and some mistakes are even perceived as successes. We also see that successes immediately following a failure have the same effect on PR as successes without a preceding failure. Finally, we show that successful executions recover PR after a failure. Our findings highlight which robot failures are in higher need of repair in a human-robot interaction, and how trust could be recovered by robot successes.

This paper presents a virtual reality (VR) study that examines how multiple road users jointly interact with an autonomous vehicle (AV) in complex traffic scenarios. Moving beyond dyadic studies (e.g., AV-pedestrian or AV-passenger), our multi-user setup simulates ambiguous all-way stop intersections involving a pedestrian, a human driver in a conventional vehicle, and a passenger in an AV, all interacting simultaneously with the AV. We investigated how users perceive and respond to two distinct types of AV behaviors: an efficient AV that proceeds as soon as it is safe to do so, and a prosocial AV that yields to others before entering the intersection. Sixteen groups of three participants (N=48) took part in the study, with each group interacting with a single AV type across four ambiguous traffic scenarios. Our findings show that even simple AV behavior logics can meaningfully shape crossing negotiation dynamics and highlight how trust and perception can vary across different user roles. We conclude by discussing how our methods and insights can inform the research and design of AV interactions in complex multi-agent traffic environments.

We present an interactive virtual agent that employs politeness strategies to suggest compromises during a joint decision-making task with a human partner. In our approach, politeness is expressed not only through language, but also through the agent’s willingness to compromise by proposing options that better align with the human partner’s preferences. We conducted a within-participant experiment (n = 97) in which users interacted with two types of agents: a performance-optimizing agent and a socially considerate agent. The former prioritized decision quality, while the latter occasionally made socially motivated quality concessions to the user as a form of face-work to defer to the human partner. Our results show that participants reported higher satisfaction with the final decision and consistent intention to collaborate with the socially considerate agent on both tasks. However, the social agent was less effective than the performance-optimizing agent at improving the users’ initial decisions. These findings highlight trade-offs in human-agent interaction between optimizing task performance and managing the social dynamics of collaboration. We conclude by discussing the ethical and cultural considerations that may be involved in navigating this trade-off.

Eye contact, even momentarily between strangers, plays a pivotal role in fostering human connection, promoting happiness, and enhancing belonging. Yet, the physical rigidity of public spaces, such as airport terminals, often limits opportunities for meaningful interactions between strangers who often remain absorbed in their personal activities. This paper introduces Mirrorbot, a robotic mirror system that transforms static environments into dynamic, socio-spatial interfaces. Through autonomous navigation and adaptive mirror control, Mirrorbot facilitates serendipitous, non-verbal interactions by dynamically transitioning reflections from self-focused to mutual recognition, sparking eye contact, shared awareness, and playful engagement. By integrating mirrors—a familiar and accessible architectural element—Mirrorbot disrupts conventional isolation in public spaces, enabling embodied, accessible interactions that go beyond screen-based solutions. This work demonstrates the potential of interactive mirrors to enrich public spaces, fostering spontaneous connections in shared environments.

This study investigates the application of real-world Reinforcement Learning (RL) in facilitating synchronized movement and fostering social connections during initial interactions between strangers. We introduced a novel, projection-based system where an RL agent is projected onto a table to interact with participants through its movements. Our research involved the development and testing of a Deep Q-Network (DQN) integrated with a Long Short-Term Memory (LSTM) model that identifies synchronized movements and determines the reward for the DQN. The preliminary results showcase the effectiveness of the LSTM model, revealing that synchronized movement can be a potent reward mechanism, enabling RL agents to learn effective policies. The initial findings indicate the feasibility of the RL approach in enhancing social interactions, providing partial support for our designed actions.

We provide a multi-sensor dataset containing RGB-D and motion tracking data from sequential human-to-human object handovers. We recorded 12 pairs of participants executing shelving and un-shelving tasks involving 30 object handovers, resulting in 1440 handovers. Each recording includes the position trajectories of 27 markers placed on the upper bodies of both the giver and the receiver, recorded at 120 Hz, as well as the position and orientation trajectories of 13 upper-body bones, which are estimated from the marker data. The recordings also include two RGB-D data streams at 30Hz. We also provide four anthropometric measurements of the participants: height, waistline height, arm span, and weight. The dataset is valuable for investigating the body movements, grasps, and coordination strategies utilized by humans while performing tasks such as shelving which involve multiple sequential object handovers. Additionally, the dataset can be used to teach robots perform tasks involving object handovers with people, as well as self-handovers to adjust grasps.

This chapter critically addresses a blind spot related to material selection in personal robot design. Robots are usually designed with materials chosen for convenience and engineering functionality, resulting in machines overwhelmingly made up of plastic and metal. When it comes to personal and home robots, this choice does not fit with the rest of the designed environment we live in, where manufactured objects feature diverse and thoughtful material choices that build on millennia of craft and design traditions. This chapter argues that, when faced with material choices, designers of personal robots should go beyond mere pragmatism, embracing cultural and affective facets that contribute to producing a holistic relationship between humans and their artifacts. It lists four ways in which materials affect the interaction between users and manufactured objects, with lessons for personal robot designers. To provide a concrete example, Blossom is a robot designed with the intentional use of traditional craft materials, like wool and wood, embodying a speculative suggestion to expand the scope of material imagination in personal robot design.

Home robots are envisioned to provide in-home assistance for older adults and other people who may need help with daily tasks. To gather information for inferring user status, robots typically require cameras to detect human subjects, track their positions, and recognize their activities or poses. However, having cameras in personal spaces, such as homes, could pose privacy concerns and risks due to the potential misuse or compromise of personal image data. It can also lead to psychological unease and feelings of insecurity, stemming from the fear of being watched and recorded. To address this issue, this paper proposes a method for preserving privacy based on physically obstructing the robot’s camera image and computer vision methods for detection and tracking of humans in these obstructed images. We present a hardware platform that includes a semi-transparent physical layer in front of the robot’s cameras to obtain privacy-preserving shadow images, and a software framework that uses a pre-trained EfficientNet, retrained with a newly-collected dataset of human shadow images for detecting and tracking human subjects. The testing results reveal that the network achieves reliable accuracy in detecting humans from various distances and angles, and it can be applied to a new subject that it has never seen before. Finally, the algorithm is implemented in a gaze-based human-robot interaction scenario, demonstrating its ability to track humans in real time while preserving privacy.

Designing plant-driven actuators presents an opportunity to create new types of devices that grow, age, and decay, such as robots that embody these qualities in their physical structure. Plant-robot hybrids that grow and decay incorporate unpredictable and gradual transformations inherent across living organisms and suggest an alternative to the design principles of immediacy, responsiveness, control, accuracy, and durability commonly found in robotic design. To explore this, we present a design space of primitives for plant-driven robotic actuators. Proof-of-concept prototypes illustrate how concepts like slow change, slow movement, decay, and destruction can be incorporated into robotic forms. We describe the design considerations required for building plant-driven actuators for robots, including experimental findings regarding the mechanical properties of plant forces. Finally, we speculate on the potential benefits of plant-robot hybrids to interactive domains such as robotics.

Sculpture offers a centuries-long tradition of techniques for expressing emotion and movement in a static form. Insights from this field present an opportunity to design robots that express not only through movement, but also via dynamic cues in their static positions. Such cues can suggest motion potential, emotion, and character. This paper presents three principles identified in sculpture techniques that can be applied to robot design: (a) depicting exposure and protection of emotional pivot points in the body, (b) weight distribution, and (c) the revelation of movement mechanisms and tension through flexible skins. We employ the first two of these principles in an interactive design and motion control environment to demonstrate the potential for application to the design of social collaborative robots. We illustrate the third principle via a robot design that uses a flexible fabric skin stretched over rigid and elastic actuation elements. Using insights from sculpture can promote the design of robots from a transdisciplinary perspective by increasing the readability of robot intent and affect even when the robot is not actively moving.

As the potential for human-AI design collaboration increases, understanding the role of artificial intelligence (AI) in the design process becomes more important. How does AI currently support the design process and how could it do so in the future? To answer this question, we categorized existing AI design support systems (DSS) according to the Double-Diamond design process model, and found that they are mostly used in the later stages of the design process, focusing on generating design solutions. In contrast, very few systems focus on the early stages of the process, which include discovering and defining design problems. To explore this finding’s alignment with designers’ expectations in real-world design, we present a case study involving emerging AI technologies such as ChatGPT and robots. This study proposes that AI agents can potentially assist designers by providing inspirations, defining design problems with constraints, offering grounded metaphors, and exploring design materials.

Shape-changing skin is an exciting modality due to its accessible and engaging nature. Its softness and flexibility make it adaptable to different interactive devices that children with and without visual impairments can share. Although their potential as an emotionally expressive medium has been shown for sighted adults, their potential as an inclusive modality remains unexplored. This work explores the shape-emotional mappings in children with and without visual impairment. We conducted a user study with 50 children (26 with visual impairment) to investigate their emotional associations with five skin shapes and two movement conditions. Results show that shape-emotional mappings are dependent on visual abilities. Our study raises awareness of the influence of visual experiences on tactile vocabulary and emotional mapping among sighted, low-vision, and blind children. We finish discussing the causal associations between tactile stimuli and emotions and suggest inclusive design recommendations for shape-changing devices.

As artificial intelligence (AI) is playing an increasingly important role in our society and global economy, AI education and literacy have become necessary components in college and K-12 education to prepare students for an AI-powered society. However, current AI curricula have not yet been made accessible and engaging enough for students and schools from all socio-economic backgrounds with different educational goals. In this work, we developed an open-source learning module for college and high school students, which allows students to build their own robot companion from the ground up. This open platform can be used to provide hands-on experience and introductory knowledge about various aspects of AI, including robotics, machine learning (ML), software engineering, and mechanical engineering. Because of the social and personal nature of a socially assistive robot companion, this module also puts a special emphasis on human-centered AI, enabling students to develop a better understanding of human-AI interaction and AI ethics through hands-on learning activities. With open-source documentation, assembling manuals and affordable materials, students from different socio-economic backgrounds can personalize their learning experience based on their individual educational goals. To evaluate the student-perceived quality of our module, we conducted a usability testing workshop with 15 college students recruited from a minority-serving institution. Our results indicate that our AI module is effective, easy-to-follow, and engaging, and it increases student interest in studying AI/ML and robotics in the future. We hope that this work will contribute toward accessible and engaging AI education in human-AI interaction for college and high school students.

Children with visual impairments often struggle to fully participate in group activities due to limited access to visual cues. They have difficulty perceiving what is happening, when, and how to act-leading to children with and without visual impairments being frustrated with the group activity, reducing mutual interactions. To address this, we created Touchibo, a tactile storyteller robot acting in a multisensory setting, encouraging touch-based interactions. Touchibo provides an inclusive space for group interaction as touch is a highly accessible modality in a mixed-visual ability context. In a study involving 107 children (37 with visual impairments), we compared Touchibo to an audio-only storyteller. Results indicate that Touchibo significantly improved children's individual and group participation perception, sparking touch-based interactions and the storyteller was more likable and helpful. Our study highlights touch-based robots' potential to enrich children's social interactions by prompting interpersonal touch, particularly in mixed-visual ability settings.

We investigate what it might look like for a robot to work with a human on a need-finding design task using an affinity diagram. While some recent projects have examined how human–robot teams might explore solutions to design problems, human–robot collaboration in the sensemaking aspects of the design process has not been studied. Designers use affinity diagrams to make sense of unstructured information by clustering paper notes on a work surface. To explore human–robot collaboration on a sensemaking design activity, we developed HIRO, an autonomous robot that constructs affinity diagrams with humans. In a within-user study, 56 participants affinity-diagrammed themes to characterize needs in quotes taken from real-world user data, once alone and once with HIRO. Users spent more time on the task with HIRO than alone, without strong evidence for corresponding effects on cognitive load. In addition, a majority of participants said they preferred to work with HIRO. From post-interaction interviews, we identified eight themes leading to four guidelines for robots that collaborate with humans on sensemaking design tasks: (1) account for the robot’s speed, (2) pursue mutual understanding rather than just correctness, (3) identify opportunities for constructive disagreements, and (4) use other modes of communication in addition to physical materials.

The ANA Avatar XPRIZE was a four-year competition to develop a robotic “avatar” system to allow a human operator to sense, communicate, and act in a remote environment as though physically present. The competition featured a unique requirement that judges would operate the avatars after less than one hour of training on the human–machine interfaces, and avatar systems were judged on both objective and subjective scoring metrics. This paper presents a unified summary and analysis of the competition from technical, judging, and organizational perspectives. We study the use of telerobotics technologies and innovations pursued by the competing teams in their avatar systems, and correlate the use of these technologies with judges’ task performance and subjective survey ratings. It also summarizes perspectives from team leads, judges, and organizers about the competition’s execution and impact to inform the future development of telerobotics and telepresence.

The ANA Avatar XPRIZE was a four-year competition to develop a robotic "avatar" system to allow a human operator to sense, communicate, and act in a remote environment as though physically present. The competition featured a unique requirement that judges would operate the avatars after less than one hour of training on the human-machine interfaces, and avatar systems were judged on both objective and subjective scoring metrics. This paper presents a unified summary and analysis of the competition from technical, judging, and organizational perspectives. We study the use of telerobotics technologies and innovations pursued by the competing teams in their avatar systems, and correlate the use of these technologies with judges' task performance and subjective survey ratings. It also summarizes perspectives from team leads, judges, and organizers about the competition's execution and impact to inform the future development of telerobotics and telepresence.

As artificial intelligence (AI) is playing an increasingly important role in our society and global economy, AI education and literacy have become necessary components in college and K-12 education to prepare students for an AI-powered society. However, current AI curricula have not yet been made accessible and engaging enough for students and schools from all socio-economic backgrounds with different educational goals. In this work, we developed an open-source learning module for college and high school students, which allows students to build their own robot companion from the ground up. This open platform can be used to provide hands-on experience and introductory knowledge about various aspects of AI, including robotics, machine learning (ML), software engineering, and mechanical engineering. Because of the social and personal nature of a socially assistive robot companion, this module also puts a special emphasis on human-centered AI, enabling students to develop a better understanding of human-AI interaction and AI ethics through hands-on learning activities. With open-source documentation, assembling manuals and affordable materials, students from different socio-economic backgrounds can personalize their learning experience based on their individual educational goals. To evaluate the student-perceived quality of our module, we conducted a usability testing workshop with 15 college students recruited from a minority-serving institution. Our results indicate that our AI module is effective, easy-to-follow, and engaging, and it increases student interest in studying AI/ML and robotics in the future. We hope that this work will contribute toward accessible and engaging AI education in human-AI interaction for college and high school students.

Researchers in human–robot collaboration have extensively studied methods for inferring human intentions and predicting their actions, as this is an important precursor for robots to provide useful assistance. We review contemporary methods for intention inference and human activity prediction. Our survey finds that intentions and goals are often inferred via Bayesian posterior estimation and Markov decision processes that model internal human states as unobserved variables or represent both agents in a shared probabilistic framework. An alternative approach is to use neural networks and other supervised learning approaches to directly map observable outcomes to intentions and to make predictions about future human activity based on past observations. That said, due to the complexity of human intentions, existing work usually reasons about limited domains, makes unrealistic simplifications about intentions, and is mostly constrained to short-term predictions. This state of the art provides opportunity for future research that could include more nuanced models of intents, reason over longer horizons, and account for the human tendency to adapt.

In this work, we present a method for personalizing human-robot interaction by using emotive facial expressions to generate affective robot movements. Movement is an important medium for robots to communicate affective states, but the expertise and time required to craft new robot movements promotes a reliance on fixed preprogrammed behaviors. Enabling robots to respond to multimodal user input with newly generated movements could stave off staleness of interaction and convey a deeper degree of affective understanding than current retrieval-based methods. We use autoencoder neural networks to compress robot movement data and facial expression images into a shared latent embedding space. Then, we use a reconstruction loss to generate movements from these embeddings and triplet loss to align the embeddings by emotion classes rather than data modality. To subjectively evaluate our method, we conducted a user survey and found that generated happy and sad movements could be matched to their source face images. However, angry movements were most often mismatched to sad images. This multimodal data-driven generative method can expand an interactive agent’s behavior library and could be adopted for other multimodal affective applications.

We present a multi-sensor dataset of bimanual human-to-human object handovers. The dataset consists of 240 recordings obtained from 12 pairs of participants performing bimanual object handovers with 10 objects, and 120 recordings obtained from the same 12 pairs of participants performing unimanual handovers with 5 of those objects. Each recording includes the giver and receiver's 13 upper-body bone position and orientation trajectories, position trajectories for the 27 markers placed on their upper bodies, object position and orientation trajectories, and two RGB-D data streams. The motion trajectories are recorded at 120Hz and the RGB-D streams are recorded at 30Hz. The recordings are annotated with the three handover phases: reach, transfer, and retreat. The dataset also includes four anthropometric measurements of the participants: height, waistline height, arm span, and weight. Our dataset could help investigations of the bimanual reaching motions and grasps utilized by humans while performing handovers. Also, it can be used to train robots to perform bimanual object handovers with humans.

Designing and evaluating the affectivity of the robot body has become a frontier topic in Human-Robot Interaction (HRI) , with previous studies [ 1 , 2 ] emphasizing the importance of robot embodiment for human-robot communication. In particular, there is growing interest in how the tactile, haptic materiality of the robot influences and mediates users' affective and emotional states. Indeed, the sheer physicality of robotic systems is a crucial factor in the morphology of the robotic platform, and therefore in the robot's appearance to the user. How do the tactile properties of materials subtly influence user interaction? Why do certain morphologies prompt more empathetic interactions than others? How is nonverbal communication affected through the coordination of movements of the torso, head, and appendages to provide more naturalistic-seeming interaction? What is the role of nonverbal communication in the production of artificial empathy? And how do such factors encourage trust and foster confidence for nonexpert users to interact in the first place? This recognition of machinic corporeality has been of practical interest to designers and engineers working across a range of robot forms and functions.

We present a multi-sensor dataset of bimanual human-to-human object handovers. The dataset consists of 240 recordings obtained from 12 pairs of participants performing bimanual object handovers with 10 objects, and 120 recordings obtained from the same 12 pairs of participants performing unimanual handovers with 5 of those objects. Each recording includes the giver and receiver's 13 upper-body bone position and orientation trajectories, position trajectories for the 27 markers placed on their upper bodies, object position and orientation trajectories, and two RGB-D data streams. The motion trajectories are recorded at 120Hz and the RGB-D streams are recorded at 30Hz. The recordings are annotated with the three handover phases: reach, transfer, and retreat. The dataset also includes four anthropometric measurements of the participants: height, waistline height, arm span, and weight. Our dataset could help investigations of the bimanual reaching motions and grasps utilized by humans while performing handovers. Also, it can be used to train robots to perform bimanual object handovers with humans. More details about the dataset can be found in this paper: https://doi.org/10.1016/j.dib.2023.109277

We present a multi-sensor dataset of bimanual human-to-human object handovers. The dataset consists of 240 recordings obtained from 12 pairs of participants performing bimanual object handovers with 10 objects, and 120 recordings obtained from the same 12 pairs of participants performing unimanual handovers with 5 of those objects. Each recording includes the giver and receiver's 13 upper-body bone position and orientation trajectories, position trajectories for the 27 markers placed on their upper bodies, object position and orientation trajectories, and two RGB-D data streams. The motion trajectories are recorded at 120Hz and the RGB-D streams are recorded at 30Hz. The recordings are annotated with the three handover phases: reach, transfer, and retreat. The dataset also includes four anthropometric measurements of the participants: height, waistline height, arm span, and weight. Our dataset could help investigations of the bimanual reaching motions and grasps utilized by humans while performing handovers. Also, it can be used to train robots to perform bimanual object handovers with humans. More details about the dataset can be found in this paper: https://doi.org/10.1016/j.dib.2023.109277

Robots have the potential to assist in emergency evacuation tasks, but it is not clear how robots should behave to evacuate people who are not fully compliant, perhaps due to panic or other priorities in an emergency. In this paper, we compare two robot strategies: an actively nudging robot that initiates evacuation and pulls toward the exit and a passively waiting robot that stays around users and waits for instruction. Both strategies were automatically synthesized from a description of the desired behavior. We conduct a within participant study ( = 20) in a simulated environment to compare the evacuation effectiveness between the two robot strategies. Our results indicate an advantage of the nudging robot for effective evacuation when being exposed to the evacuation scenario for the first time. The waiting robot results in lower efficiency, higher mental load, and more physical conflicts. However, participants like the waiting robots equally or slightly more when they repeat the evacuation scenario and are more familiar with the situation. Our qualitative analysis of the participants' feedback suggests several design implications for future emergency evacuation robots.