Schedule

Abstract: By now we are all aware that robots have the potential to serve as tools to improve human mental wellbeing. However, research in the following areas is still limited: 1) exploration of the expectations and perceptions of prospective users of robotic wellbeing coaches, and the professionals who currently deliver these interventions; 2) real-time AI-driven affective adaptation mechanisms; and 3) longitudinal HRI studies and deployment. In this talk, I will present the recent explorations of the Cambridge Affective Intelligence and Robotics Lab in these areas with insights for short- and long-term adaptation.

Abstract: This paper describes current work on deploying our social robot ARI in two different use-cases at day-care centres for the elderly: one, focused on providing support of cognitive activities, while the second one, focused on promoting engagement in the routines carried out at the day-care. In both cases, a user-centric design approach has proven to be essential to properly tackle the needs of each use-case. While in the first case we are presenting the current development in preparation for a pilot test taking place in the following months, in the second case, we present the prototype development process and initial validation and feedback gathered from a 2-day pilot study. We conclude by suggesting similarities and differences between the needs of both pilots and future challenges.

Abstract: Initial findings show that people may blame or punish robots for their (in)actions in normative situations. Yet, specifics on which robot behaviors are perceived as morally sensitive or how people morally judge robots for norm violating behavior remains unknown. We have investigated participants’ judgements of norm violating behaviors in an online vignette study (n = 107) using a mixed methods design with agent type (young man, mechanical robot, and humanoid robot) as between subject factor and behavior type (cutting the line, littering in the park, pushing all elevator buttons, and invading personal space) as within subject factor. Participants’ judgments of norm violating behaviors are mostly affected by their moral judgment of the norm violation and the type of agent that violates the norm. Our exploration of people’s psychological understanding of robots as intentional moral agents generates a deeper understanding of human-robot cohabitation and may help drafting guidelines for correct and effective implementation of norm capacity into robot systems.

Abstract: The talk addresses mental state attribution - i.e. the interpretation of others’ behavior in terms of underlying Intentions, beliefs, emotions, etc. – as well as its crucial role in social robotics, and why this is not unproblematic.

Abstract: Social robots are found to be effective in language learning and teaching. The present study proposes a novel Reinforcement Learning-based alphabet learning activity for the acquisition of the newly introduced Kazakh Latin alphabet. We conducted a between-subject design experiment with 60 Kazakh children aged 6-8 years old from a local public school to compare their learning performance across the three conditions: CoWriting Kazakh robot (CW), Reinforcement Learning (RL) robot, and a human tutor (HT), and find the most cognitively demanding yet most engaging way of alphabet acquisition. The results show that children learned significantly more letters with the HT followed by the CW robot, while the least effective condition for alphabet learning was the RL robot. However, the two robot conditions received significantly higher likability scores than the human tutor. These results are yet exploratory and need further investigation.

Abstract: In this paper, we evaluate how dialogues generated by large language models, in this context GPT-3, exhibit two distinct types of inappropriate responses. We then suggest a framework for how embodied dialogue systems could adapt to the appropriateness of the system’s dialogue in the context where the embodied system exists. We also present a plan for how this framework will be used for a future data collection.

Abstract: The talk will start by discussing the results of the CARESSES project, aiming at developing robots that reconfigure their verbal and non-verbal behaviour depending on the cultural identity of the person with whom they interact. Then, starting from the lesson learned in the project, we will discuss ongoing work to explore how diversity awareness can contribute to developing robots that respect and value people's diversity to ensure improved performance and acceptability.

Abstract: Adhering to individual, unsupervised practice of repetitive exercises can significantly improve a stroke survivor’s recovery over the long-term. However, motivational struggles when a physiotherapist is not present are one of the biggest contributing factors to lack of adherence when the rehabilitation is carried out at home. In this paper, we propose a robotic coach capable of motivating users to conduct rehabilitation using both high-level personalisation to users with particular traits and in particular training contexts, and low-level adaption to individuals based on their interaction with the robot.

Abstract: Imitation is a paramount step in child development and is thought to contribute to the development of theory of mind. Child-robot imitation is a widespread approach in robot-assisted therapy, notably with children with autism spectrum disorders (ASD).Although imitation deficits in ASD are still controversial, it has been observed that children with ASD may present developmental coordination disorders with motor imitation deficits. In this paper, we propose a novel method for motor imitation using Central Pattern Generators (CPG) and employ it in a feasibility study involving children with ASD and the Pepper robot.

Abstract: In robot-assisted therapy for individuals with Autism Spectrum Disorder, the workload of therapists during a therapeutic session is increased if they have to control the robot manually. To allow therapists to focus on the interaction with the person instead, the robot should be more autonomous, namely it should be able to interpret the person’s state and continuously adapt its actions according to their behaviour. In this paper, we develop a personalised robot behaviour model that can be used in the robot decision-making process during an activity; this behaviour model is trained with the help of a user model that has been learned from real interaction data. We use Q-learning for this task, such that the results demonstrate that the policy requires about 10,000 iterations to converge. We thus investigate policy transfer for improving the convergence speed; we show that this is a feasible solution, but an inappropriate initial policy can lead to a suboptimal final return.

Abstract: Robots applied in therapeutic scenarios, for instance in the therapy of individuals with Autism Spectrum Disorder, are sometimes used for imitation learning activities in which a person needs to repeat motions by the robot. To simplify the task of incorporating new types of motions that a robot can perform, it is desirable that the robot has the ability to learn motions by observing demonstrations from a human, such as a therapist. In this paper, we investigate an approach for acquiring motions from skeleton observations of a human, which are collected by a robot-centric RGB-D camera. Given a sequence of observations of various joints, the joint positions are mapped to match the configuration of a robot before being executed by a PID position controller. We evaluate the method, in particular the reproduction error, by performing a study with QTrobot in which the robot acquired different upper-body dance moves from multiple participants. The results indicate the method’s overall feasibility, but also indicate that the reproduction quality is affected by noise in the skeleton observations.