What is covered in this chapter:• The role of nonverbal communication in interactions between people-how communication is enhanced by facial expressions, hand gestures, body posture, and sounds; • The importance of interpreting, using, and responding to nonverbal cues in the appropriate way, both to successful humanrobot interactions and to generate a positive perception of robots; • Nonverbal communication channels that are unique to robots, as well as channels that replicate those commonly used by humans; • How robotic sounds, lights, and colors or physical gestures with arms, legs, tails, ears, and other body parts can be effective for communicating with people. 81

Humans recognise and respond to robots as social agents, to such extent that they occasionally a empt to bully a robot. e current paper investigates whether aggressive behaviour directed towards robots is in uenced by the same social processes that guide human bullying behaviour. More speci cally, it measured the e ects of dehumanisation primes and anthropomorphic qualities of the robot on participants' verbal abuse of a virtual robotic agents. Contrary to previous ndings in human-human interaction, priming participants with power did not result in less mind a ribution. However, evidence for dehumanisation was still found, as the less mind participants a ributed to the robot, the more aggressive responses they gave. In the main study this e ect was moderated by the manipulations of power and robot anthropomorphism; the low anthropomorphic robot in the power prime condition endured signi cantly less abuse, and mind a ribution remained a signi cant predictor for verbal aggression in all conditions save the low anthropomorphic robot with no prime. It is concluded that dehumanisation occurs in human-robot interaction and that like in human-human interaction, it is linked to aggressive behaviour. Moreover, it is argued that this dehumanisation is di erent from anthropomorphism as well as human-human dehumanisation, since anthropomorphism itself did not predict aggressive behaviour and dehumanisation of robots was not in uenced by primes that have been established in human-human dehumanisation research.

Background Over the last 2 decades, virtual reality technologies (VRTs) have been proposed as a way to enhance and improve smoking cessation therapy. Objective This systematic review aims to evaluate and summarize the current knowledge on the application of VRT in various smoking cessation therapies, as well as to explore potential directions for future research and intervention development. Methods A literature review of smoking interventions using VRT was conducted. Results Not all intervention studies included an alternative therapy or a placebo condition against which the effectiveness of the intervention could be benchmarked, or a follow-up measure to ensure that the effects were lasting. Virtual reality (VR) cue exposure therapy was the most extensively studied intervention, but its effect on long-term smoking behavior was inconsistent. Behavioral therapies such as a VR approach-avoidance task or gamified interventions were less common but reported positive results. Notably, only 1 study combined Electronic Nicotine Delivery Devices with VRT. Conclusions The inclusion of a behavioral component, as is done in the VR approach-avoidance task and gamified interventions, may be an interesting avenue for future research on smoking interventions. As Electronic Nicotine Delivery Devices are still the subject of much controversy, their potential to support smoking cessation remains unclear. For future research, behavioral or multicomponent interventions are promising avenues of exploration. Future studies should improve their validity by comparing their intervention group with at least 1 alternative or placebo control group, as well as incorporating follow-up measures.

AbstractIt is not uncommon for humans to exhibit abusive behaviour towards robots. This study compares how abusive behaviour towards a human is perceived differently in comparison with identical behaviour towards a robot. We showed participants 16 video clips of unparalleled quality that depicted different levels of violence and abuse. For each video, we asked participants to rate the moral acceptability of the action, the violence depicted, the intention to harm, and how abusive the action was. The results indicate no significant difference in the perceived morality of the actions shown in the videos across the two victim agents. When the agents started to fight back, their reactive aggressive behaviour was rated differently. Humans fighting back were seen as less immoral compared with robots fighting back. A mediation analysis showed that this was predominately due to participants perceiving the robot’s response as more abusive than the human’s response.

There have been multiple incidents where humans attacked robots in a public environment (Brscić et al., in: Proceedings of the international conference on human-robot interaction, ACM/IEEE,

Mind perception is a fundamental part of anthropomorphism and has recently been suggested to be a dual process. The current research studied the influence of implicit and explicit mind perception on a robot’s right to be protected from abuse, both in terms of participants condemning abuse that befell the robot as well as in terms of participants’ tendency to humiliate the robot themselves. Results indicated that acceptability of robot abuse can be manipulated through explicit mind perception, yet are inconclusive about the influence of implicit mind perception. Interestingly, explicit attribution of mind to the robot did not make people less likely to mistreat the robot. This suggests that the relationship between a robot’s perceived mind and right to protection is far from straightforward, and has implications for researchers and engineers who want to tackle the issue of robot abuse.

With the emergence of onboard vision processing for areas such as the internet of things (IoT), edge computing and autonomous robots, there is increasing demand for computationally efficient convolutional neural network (CNN) models to perform real-time object detection on resource constraints hardware devices. Tiny-YOLO is generally considered as one of the faster object detectors for low-end devices and is the basis for our work. Our experiments on this network have shown that Tiny-YOLO can achieve 0.14 frames per second (FPS) on the Raspberry Pi 3 B, which is too slow for soccer playing autonomous humanoid robots detecting goal and ball objects. In this paper we propose an adaptation to the YOLO CNN model named xYOLO, that can achieve object detection at a speed of 9.66 FPS on the Raspberry Pi 3 B. This is achieved by trading an acceptable amount of accuracy, making the network approximately 70 times faster than Tiny-YOLO. Greater inference speed-ups were also achieved on a desktop CPU and GPU. Additionally we contribute an annotated Darknet dataset for goal and ball detection.

No abstract