International Journal of Social Robotics最新文献

Robots in multi-user environments require adaptation to produce personalized interactions. In these scenarios, the user’s feedback leads the robots to learn from experiences and use this knowledge to generate adapted activities to the user’s preferences. However, preferences are user-specific and may suffer variations, so learning is required to personalize the robot’s actions to each user. Robots can obtain feedback in Human–Robot Interaction by asking users their opinion about the activity (explicit feedback) or estimating it from the interaction (implicit feedback). This paper presents a Reinforcement Learning framework for social robots to personalize activity selection using the preferences and feedback obtained from the users. This paper also studies the role of user feedback in learning, and it asks whether combining explicit and implicit user feedback produces better robot adaptive behavior than considering them separately. We evaluated the system with 24 participants in a long-term experiment where they were divided into three conditions: (i) adapting the activity selection using the explicit feedback that was obtained from asking the user how much they liked the activities; (ii) using the implicit feedback obtained from interaction metrics of each activity generated from the user’s actions; and (iii) combining explicit and implicit feedback. As we hypothesized, the results show that combining both feedback produces better adaptive values when correlating initial and final activity scores, overcoming the use of individual explicit and implicit feedback. We also found that the kind of user feedback does not affect the user’s engagement or the number of activities carried out during the experiment.

Annotation tools serve a critical role in the generation of datasets that fuel machine learning applications. With the advent of Foundation Models, particularly those based on Transformer architectures and expansive language models, the capacity for training on comprehensive, multimodal datasets has been substantially enhanced. This not only facilitates robust generalization across diverse data categories and knowledge domains but also necessitates a novel form of annotation—prompt engineering—for qualitative model fine-tuning. This advancement creates new avenues for machine intelligence to more precisely identify, forecast, and replicate human behavior, addressing historical limitations that contribute to algorithmic inequities. Nevertheless, the voluminous and intricate nature of the data essential for training multimodal models poses significant engineering challenges, particularly with regard to bias. No consensus has yet emerged on optimal procedures for conducting this annotation work in a manner that is ethically responsible, secure, and efficient. This historical literature review traces advancements in these technologies from 2018 onward, underscores significant contributions, and identifies existing knowledge gaps and avenues for future research pertinent to the development of Transformer-based multimodal Foundation Models. An initial survey of over 724 articles yielded 156 studies that met the criteria for historical analysis; these were further narrowed down to 46 key papers spanning the years 2018–2022. The review offers valuable perspectives on the evolution of best practices, pinpoints current knowledge deficiencies, and suggests potential directions for future research. The paper includes six figures and delves into the transformation of research landscapes in the realm of machine-assisted behavioral annotation, focusing on critical issues such as bias.

Abstract

As China has become the largest user of industrial robots, the need to understand how workers perceive robot-human substitution and how their perceptions influence their job behaviors is becoming increasingly crucial. This paper examined whether workers’ fear of being replaced by robots (FRR) is correlated with one aspect of job behavior: turnover intention, which refers to the extent to which an individual intends to change their job within a specific time period. Using a dataset covering 1512 manufacturing workers in Guangdong province of China, we found that workers who fear losing their jobs to robots report significantly higher turnover intention. We also found that the positive effect of FRR on turnover intention increased when robots were already utilised in the workplace. This effect was also found to be increase when workers perceived that their wages did not increase with the rise in productivity due to robotisation. Based on these findings, we provide practical recommendations to organizations on effectively addressing the turnover intention arising from the FRR.

As social robots are being built with the aim of employing them in our social environments, it is crucial to understand whether we are inclined to include them in our social ingroups. Social inclusion might depend on various factors. To understand if people have the tendency to treat robots as their in-group members, we adapted a classical social psychology paradigm, namely the “Cyberball game”, to a 3-D experimental protocol involving an embodied humanoid robot. In our experiment, participants played the ball-tossing game with the iCub robot and another human confederate. In our version, the human confederate was instructed to exclude the robot from the game. This was done to investigate whether participants would re-include the robot in the game. In addition, we examined if acquired technical knowledge about robots would affect social inclusion. To this aim, participants performed the Cyberball twice, namely before and after a familiarization phase when they were provided with technical knowledge about the mechanics and software related to the functionality of the robot. Results showed that participants socially re-included the robot during the task, equally before and after the familiarization session. The familiarization phase did not affect the frequency of social inclusion, suggesting that humans tend to socially include robots, independent of the knowledge they have about their inner functioning.

Abstract

This study proposes to improve the reliability, robustness and human-like nature of Human–Robot Collaboration (HRC). For that, the classical Perception–Action cycle is extended to a Perception–Intention–Action (PIA) cycle, which includes an Intention stage at the same level as the Perception one, being in charge of obtaining both the implicit and the explicit intention of the human, opposing to classical approaches based on inferring everything from perception. This complete cycle is exposed theoretically including its use of the concept of Situation Awareness, which is shown as a key element for the correct understanding of the current situation and future action prediction. This enables the assignment of roles to the agents involved in a collaborative task and the building of collaborative plans. To visualize the cycle, a collaborative transportation task is used as a use-case. A force-based model is designed to combine the robot’s perception of its environment with the force exerted by the human and other factors in an illustrative way. Finally, a total of 58 volunteers participate in two rounds of experiments. In these, it is shown that the human agrees to explicitly state their intention without undue extra effort and that the human understands that this helps to minimize robot errors or misunderstandings. It is also shown that a system that correctly combines inference with explicit elicitation of the human’s intention is the best rated by the human on multiple parameters related to effective Human–Robot Interaction (HRI), such as perceived safety or trust in the robot.

In the context of recent demographic changes and related societal challenges, socially assistive robots (SARs) are considered having the potential to support independence and care of older adults. However, little is known about the preferred SAR-features of older adults with mild cognitive impairment (MCI) residing in assisted living and their caregivers. Semi-structured interviews were conducted with two stakeholder groups: older adults with MCI and their (in)formal caregivers. Inductive thematic analysis was used to analyse the data. Forty individual semi-structured interviews were conducted with older adults with MCI (N = 30) and (in)formal caregivers (N = 10). Data revealed seven common role-expectations regarding SARs for both the older adults and caregivers: (1) companion, (2) health assistant, (3) household assistant, (4) physical assistant, (5) cognitive assistant, (6) coach, (7) leisure buddy. One additional, eighth role was identified for the caregivers, i.e. job assistant. The results of this study provide a better knowledge of the features to consider during the development process of SARs in order to maximize the perceived usefulness and hence the intention to use and actual adoption. Additionally, a feasibility analysis showed which features should have the primary focus during the further software development of an existing SAR called James^® within the ReMIND-project.

Robots are a source of evaluative conflict and thus elicit ambivalence. In fact, psychological research has shown across domains that people simultaneously report strong positive and strong negative evaluations about one and the same attitude object. This is defined as ambivalence. In the current research, we extended existing ambivalence research by measuring ambivalence towards various robot-related stimuli using explicit (i.e., self-report) and implicit measures. Concretely, we used a mouse tracking approach to gain insights into the experience and resolution of evaluative conflict elicited by robots. We conducted an extended replication across four experiments with N = 411 overall. This featured a mixed-methods approach and included a single paper meta-analysis. Thereby, we showed that the amount of reported conflicting thoughts and feelings (i.e., objective ambivalence) and self-reported experienced conflict (i.e., subjective ambivalence) were consistently higher towards robot-related stimuli compared to stimuli evoking univalent responses. Further, implicit measures of ambivalence revealed that response times were higher when evaluating robot-related stimuli compared to univalent stimuli, however results concerning behavioral indicators of ambivalence in mouse trajectories were inconsistent. This might indicate that behavioral indicators of ambivalence apparently depend on the respective robot-related stimulus. We could not obtain evidence of systematic information processing as a cognitive indicator of ambivalence, however, qualitative data suggested that participants might focus on especially strong arguments to compensate their experienced conflict. Furthermore, interindividual differences did not seem to substantially influence ambivalence towards robots. Taken together, the current work successfully applied the implicit and explicit measurement of ambivalent attitudes to the domain of social robotics, while at the same time identifying potential boundaries for its application.

The rapid development of artificial intelligence technology allows robots to have social functions. In the case of human individuals interacting directly with a robot with artificial intelligence, if the individual can perceive the same or similar feelings as they have when interacting with a real human, the robot can be considered to have social presence. Trust is an important factor that affects human–robot collaboration. This research explores the influence of the character and information disclosure of robots on trust in human–robot collaboration as well as the mediating role of social presence. This study uses the Columbia Card Task to design a human–robot cooperative experiment platform. During the experiment, robots provide different levels of character (introversion vs. extroversion) and information disclosure (high disclosure vs. low disclosure). The results show that the character of robots has a significant impact on emotional trust: the higher the level of extroversion is, the stronger the level of human emotional trust. Furthermore, the level of information disclosure by robots has a significant impact on cognitive trust: the higher the level of information disclosure is, the stronger the level of cognitive trust. Social presence has a mediating role in the effect of character on emotional trust and the impact of information disclosure on cognitive trust. The research results can provide suggestions for improving the acceptance of social robots in human–robot collaboration and improving the quality and efficiency of collaborative human–robot task decision-making. Research on robots’ character and information disclosure can provide a theoretical basis for related researchers and developers.

Our world is being increasingly pervaded by intelligent robots with varying degrees of autonomy. To seamlessly integrate themselves in our society, these machines should possess the ability to navigate the complexities of our daily routines even in the absence of a human’s direct input. In other words, we want these robots to understand the intentions of their partners with the purpose of predicting the best way to help them. In this paper, we present the initial iteration of cognitive architecture for social perception and engagement in robots: a symbolic cognitive architecture that uses qualitative spatial reasoning to anticipate the pursued goal of another agent and to calculate the best collaborative behavior. This is performed through an ensemble of parallel processes that model a low-level action recognition and a high-level goal understanding, both of which are formally verified. We have tested this architecture in a simulated kitchen environment and the results we have collected show that the robot is able to both recognize an ongoing goal and to properly collaborate towards its achievement. This demonstrates a new use of qualitative spatial relations applied to the problem of intention reading in the domain of human–robot interaction.

Developing an authentic understanding of potential users’ needs and translating these into usable categories as an input to research and development is an open problem. It is generally accepted that genuine knowledge of user needs is essential for the creation of any new technology. For assistive robots, however, this knowledge is even more important for two key reasons. First, because the form and function of these technologies is still in the process of negotiation, and second, because assistive robots are ultimately intended for a vulnerable population. In this paper, we describe a number of existing strategies to address this challenge and discuss some of their shortcomings, including a loss of data richness and context, the stereotyping of users and a lack of transparency and traceability. The primary contribution of this paper is a novel Authentic Citations process for capturing needs which aims to address these shortcomings. This process involves a thematic analysis of complex qualitative data to derive robotics needs for older people, which emphasises the retention of the original situated description, or ‘authentic citation’, for ongoing sensitising and grounding at all stages of the research and development cycle, and by various stakeholders. The Authentic Citations process adds additional rigour to a process that can be tacit and opaque and can be used by robotics researchers to analyse and translate qualitative research into usable categories. An additional contribution of this paper is an initial outline of a taxonomy of assistive robotics needs for older people, which contributes to improving the understanding of the user as a situated and complex person and can be used as an input to design.