The integration of social robots into educational contexts is steadily increasing, as these agents are used to foster engagement, support personalized learning, and create interactive experiences. In these contexts, empathy towards a social robot is a crucial mechanism to build trust, motivation, and social connections. Narrative framing has been shown to support the elicitation of empathy. However, little is known about how children’s empathy towards robots is affected by different narrative framing and how individual factors shape this relationship. Across two preregistered experiments, we investigated 7–15-year-olds’ empathy towards a social robot in focus groups (Experiment 1; n = 19) and the effect of narrative framing (sad vs. neutral), baseline empathy, age, and gender thereon in an experimental study (Experiment 2; n = 73). Experiment 1 showed that robot perception, personal experiences, and social norms affected children’s empathy towards a robot. Experiment 2 showed no significant effects of narrative framing, age, or baseline empathy on children’s empathy towards the robot. However, varying the narrative framing of the robot resulted in gender differences in elicited empathy, with girls showing higher empathy than boys in the neutral narrative condition. Our findings indicate that contextual and relational cues might exert a stronger influence on children’s empathic responses towards robots than developmental factors or dispositional traits specific to individual children. The two experiments that compose this study offer valuable insights into how empathy might be elicited through social robots. These insights hold promise for informing how best to design robotic agents that children can connect with in meaningful and effective ways in education and learning contexts.
We present a detailed and unusual case study of ensuring trustworthiness in social robots in practice—the artwork Cat Royale in which a robot played with a family of three cats. We reveal how delivering Cat Royale involved tackling diverse aspects of trustworthiness beyond the immediate interaction between cats and robot, from the control room, to animal welfare expertise, extensive ethical review, and carefully planned public engagement. We contribute a framework for describing ecosystems of trust comprising five layers: personal workspace, orchestration, wider workspace, organisational culture, and the public sphere.
Robotic systems are increasingly integrated into healthcare settings, providing physical assistance, social interaction, remote diagnostics, and data-intensive services. Although these systems collect sensitive multimodal data, such as audio, video, physiological signals, and contextual metadata, existing security frameworks often fail to adequately address the combined challenges of privacy, cybersecurity, and human-robot interaction (HRI). To understand further, we conducted a systematic review of 62 peer-reviewed studies from an initial pool of 393 articles. Using our proposed PRoSec-HRI (Privacy and Robotic Security in Healthcare Robotics Interaction) framework, we identified leading technical strategies, including differential privacy for sensor data anonymization, federated learning for decentralized model training, blockchain-based authentication for auditable processes, and formal verification for privacy-compliant behavior. Despite majority of studies discuss privacy-preserving ( 68% ) and cybersecurity ( 76% ) techniques, we found only 10% conducted in-situ evaluations, and just 13% considered firmware or lifecycle security. Moreover, only 14% demonstrated regulatory compliance (e.g., GDPR, HIPAA), and a mere 9% addressed real-time threat mitigation. While 61% papers featured trust-building mechanisms, such as symbolic gestures or consent dashboards, only 11% of those accounted for cultural sensitivity or user education. Our findings expose significant gaps and provide a foundation for developing privacy-aware, secure, and user-centered healthcare robots.
The Syrian refugee crisis is among the largest globally. We are developing a social robot tailored to the needs of displaced Syrians hosted in Scotland. As part of a mixed-methods study to understand the needs of this population and the possible use cases of the robot, we conducted two focus groups with Syrian refugees and asylum seekers residing in Glasgow. Using thematic analysis, we identified these participants’ unmet needs and existing gaps in access to services. Participants observed an Arabic-speaking social robot, and together we explored its potential as a solution to help navigate bureaucratic processes and access services. The participants expressed curiosity and enthusiasm about the robot. As they shared experiences of homelessness and displacement, they also highlighted bureaucracy and the English language as key barriers to accessing services. This study identifies key design requirements for developing a multilingual support robot for refugees and asylum seekers.
This paper presents preliminary findings from a six-week study evaluating the interaction between older adults and the EBO social robot across an elderly care center in Cáceres, Spain. The research examined how different robot configurations and interaction modalities influence engagement, satisfaction, and performance among 18 older adults with mild to moderate cognitive impairment (MMSE ≥ 21). Following a structured protocol approved by the UEX Bioethics and Biosafety Commission, participants engaged in twice-weekly sessions combining conversational interactions (3–5 min) and serious games, including Storytelling, Pasapalabra (word quiz game), and Simon Says (10–15 min). The study implemented a cross-over design alternating between two distinct robotic platform and interaction modes: Mode 1 (positive reinforcement, emotive expression, proximal positioning) and Mode 2 (neutral language, limited emotional expression, distant positioning). Preliminary results reveal that performance in cognitive games improved progressively across the intervention period, with 15
The rapid growth of the older adult population calls for conversational robots that foster social participation and community inclusion. We report a pilot study at Tohoku University in which eight adults (65–78 years old) each held a 20-minute semi-structured dialogue with Android, a hyper-realistic android equipped with Japanese BERT emotion recognition and GPT-3.5-based response generation. Sentiment logs and post-interaction interviews showed that participants generally felt at ease, described the robot as a “good listener,” and remained engaged throughout the exchange; only occasional affective mismatches disrupted rapport. These preliminary results suggest that real-time affect adaptation can support empathic HRI, yet also expose the need for tighter timing and context control. Our mixed-methods findings lay the groundwork for designing emotionally responsive humanoids that enrich elder-care settings.
Socially assistive robots (SARs) have shown great potential for supplementing well-being support. However, prior studies have found that existing dialogue pipelines for SARs remain limited in real-time latency, back-channeling, and personalized speech dialogue. Toward addressing these limitations, we propose using integrated end-to-end speech-language models (SLMs) with SARs. This work 1) evaluated the usability of an SLM-enabled SAR dialogue system through a small user study, and 2) identified remaining limitations through study user feedback to inform future improvements. We conducted a small within-participant user study with university students (N = 11) whose results showed that participants perceived an SLM-enabled SAR system as capable of providing empathetic feedback, natural turn-taking, back-channeling, and adaptive responses. We also found that participants reported the robot's nonverbal behaviors as lacking variability and synchronization with conversation, and the SLM's verbal feedback as generic and repetitive. These findings highlighted the need for real-time robot movement synchronized with conversation, improved prompting or fine-tuning to generate outputs better aligned with mental health practices, and more expressive, adaptive vocal generation.
This position paper explores pluriperspectivism as a core element of human creative experience and its relevance to humanrobot cocreativity We propose a layered fivedimensional model to guide the design of cocreative behaviors and the analysis of interaction dynamics This model is based on literature and results from an interview study we conducted with 10 visual artists and 8 arts educators examining how pluriperspectivism supports creative practice The findings of this study provide insight in how robots could enhance human creativity through adaptive contextsensitive behavior demonstrating the potential of pluriperspectivism This paper outlines future directions for integrating pluriperspectivism with visionlanguage models VLMs to support context sensitivity in cocreative robots
Flaky tests refer to test cases that exhibit inconsistent behaviors across multiple executions, potentially passing or failing unpredictably. They are frequently associated with suboptimal design practices that testers may utilize when crafting test cases, which undermine the quality of software testing. So, identifying the root causes of flaky tests is crucial for fixing them. Currently, inspired by the success of the Large Language Models (LLMs), researchers leverage the pre-trained language model to embed flaky test code as vectors and predict its root cause category based on vector similarity measures. However, such code embeddings generated by LLM mainly focus on capturing general semantic features but lack sufficient comprehension of the behavioral patterns involved in test scenarios, resulting in poor root cause identification. Test smells, which reflect poor coding practices or habits when writing test cases, provide complementary information in the root cause identification of test flakiness. Therefore, this paper proposes a root cause identification method for flaky tests based on test smells. Test smells are used to abstract and express behavioral patterns of test codes, and general semantic features extracted by vector embeddings to enhance the feature representation of flaky tests. Furthermore, to capture the complex nonlinear relationships between test smell features and code embeddings, a Feedforward Neural Network is constructed to categorize the root cause of test flakiness. To validate the effectiveness of our method, we performed evaluations on a dataset consisting of 451 Java flaky test cases. The experimental results indicate that our method achieves an F1-score of 80%, which is 7% higher than that of the baseline model that does not incorporate test smells.
Software systems often evolve into multiple variants to meet diverse requirements. This is usually achieved with the clone-and-own approach, where an existing variant is copied and modified. While efficient in the short term, this approach presents challenges for long-term maintenance. A suitable solution to overcome this, is to re-engineer the variants into a software product line (SPL). However, this process is labor-intensive and prone to errors. Although initial studies explore the use of large language models (LLMs) to assist in the re-engineering tasks, they do not address challenges such as hallucination and limited context windows, which restricts the applicability.In this paper, we present a novel approach to assist the transformation of cloned software variants into an SPL using an LLM. To mitigate hallucination, we propose a self-refinement feedback loop to validate the generated SPL. Additionally, we introduce a variation point filtering technique that reduces the input size, while preserving essential information. To quantify and evaluate the generated output, we propose the use of existing metrics that can be employed for the evaluation. Our evaluation demonstrates the effectiveness of the self-refinement feedback loop and variation point filtering based on an existing case study. The results, benchmarked against the proposed variability metrics, indicate that the generated SPL maintains equivalent complexity and potential for reusability, to the system it is compared against.
Modern software systems are highly configurable, offering a vast number of configuration options that can be customized to meet specific functional and non-functional requirements. To support the configuration process, several automated software approaches based on machine learning have been proposed in the literature. These approaches aim to assist developers by predicting non-functional properties based on configuration settings. A recent study demonstrated the potential of leveraging a subset of configuration options (a.k.a. features) to achieve accurate performance predictions in the Linux kernel. The promise of learning over a reduced set of features - instead of all features - is to obtain performance models that are faster to compute, simpler to interpret, and still accurate. Despite the encouraging results of the original study, several questions remain unresolved: Can the findings be generalized to other configurable systems other than Linux? Which learning algorithms deliver the most efficient results when working with a reduced number of features? What are the most effective sampling strategies for building accurate and efficient models? In this work, we extend the original study by conducting an in-depth analysis across eight configurable systems. We evaluate the impact of sampling strategies and learning algorithms on model accuracy and training efficiency. Our goal is to understand whether there is a dominant sampling strategy and learning algorithm for varying systems and performance targets. Our results reveal variability in optimal strategies across systems and advocate for tailored approaches rather than universal solutions.
Graphical user interface (GUI) prototyping helps to clarify requirements and keep stakeholders engaged in software development. While contemporary approaches retrieve GUIs relevant to a user’s query, little support exists for the actual reuse, i.e., for using an existing GUI to create a new one. To shorten the gap, we investigate GUI-centered reuse via one of the latest artificial intelligence (AI) techniques—vision-language models (VLMs). We report an empirical study involving 73 university students working on ten GUI reuse tasks. Each task is associated with different reuse directions recommended by VLMs and by a natural language (NL) method. In addition, a focused GUI element is provided to offer a starting point for making the actual changes. Our results show that VLMs significantly outperform the NL method in making reuse recommendations, but surprisingly, the focused GUI elements are not consistently modified during reuse. With the assessments made by four experienced designers, we further offer insights into the creativity of human-reuse and AI-reuse results.
DevOps enhances software production through IT automation, continuous integration, and deployment, with Docker as a key tool that packages applications and their environments into standardized images for consistent and efficient deployment. Dockerfiles, which are text-based configuration files, define the composition and runtime actions of these images. Mismanagement of dependencies between Dockerfile instructions can cause build failures, highlighting the need for accurate dependency parsing. Current methods often miss implicit dependencies due to the complex syntax and logic of Dockerfile instructions. To address this, we propose DockInsight, a novel tool that uses a rule-based approach and semantic analysis to determine Dockerfile dependencies accurately. DockInsight features a unified feature structure representation, DVector, and a dependency type table to facilitate precise dependency determination. Evaluations demonstrate that DockInsight achieves 99.44% accuracy, significantly outperforming keyword matching and large language model methods by 64.84% and 55.74%, respectively. Additionally, DockInsight maintains stable processing times across various Dockerfile lengths, proving its efficiency and scalability. Our ablation study further highlights the importance of semantic information supplementation, particularly for RUN instructions, in enhancing accuracy. DockInsight’s robust performance makes it a valuable tool for developers and DevOps engineers, contributing to more reliable and maintainable Dockerfiles.
Microservices have been a key architectural approach for over a decade, transforming system design by promoting decentralization and allowing development teams to work independently on specific microservices. While loosely coupled microservices are ideal, dependencies between them are inevitable. Often, these dependencies go unnoticed by development teams. Although syntactic dependencies can be identified, tracking semantic dependencies — when multiple microservices share similar logic — poses a greater challenge. As systems evolve, changes made to one microservice can trigger ripple effects, jeopardizing system consistency and requiring updates to dependent services, which increases maintenance and operational complexity. Effectively tracking different types of dependencies across microservices is essential for anticipating the impact of such changes. This paper introduces the Semantic Dependency Matrix as an instrument to address these challenges from a semantic perspective. We propose an automated approach to extract and represent these dependencies and demonstrate its effectiveness through a case study. This paper takes a step further by demonstrating the significance of semantic dependencies, even in cases where there are no direct dependencies between microservices. It shows that these hidden dependencies can exist independently of endpoint or data dependencies, revealing critical connections that might otherwise be overlooked.
Cloud-native technologies enable the development of scalable, secure, and resilient applications in diverse cloud environments. However, deploying and configuring these technologies can be complex and error-prone. Existing reusable Infrastructure as Code (IaC) solutions often suffer from inconsistent structures, limited customization, and potential vendor lock-in, hindering their effectiveness and usability. To mitigate this, this paper presents a framework for creating standardized and reusable IaC templates for deploying and configuring cloud-native infrastructure components. We conducted a needs analysis to identify key developer requirements and established ten design principles for creating reusable IaC templates. Additionally, we conducted a multivocal review of relevant literature to summarize best practices for implementing these principles. Leveraging these insights, we developed five IaC templates and evaluated their effectiveness and usability against existing solutions. This proposed framework integrates conceptual guidance with practical implementations of reusable templates to simplify cloud-native infrastructure setup and enhance developer productivity.
An assurance case provides a structured argument, supported by evidence, aiming to justify some key property of a system. Reference assurance cases can serve as standardised templates or examples for developing assurance cases across various industries, facilitating alignment with regulatory standards and supporting certification. They hold the potential to more efficiently develop the assurance case and ensure best practice is maintained. A key technique in developing a reference assurance case is the use of assurance patterns. These patterns, inspired by design patterns, enable the reuse of safety argument structures. In this paper we apply this concept to the assurance of autonomous inspection robots that operate in dynamic and uncertain environments. Given the inherent complexity that arises from the autonomy of these systems, a range of distinct verification methods (e.g., formal verification, simulation, physical experiments) will be required to foster confidence. This work-in-progress paper proposes a corroborative assurance approach, enabling engineers to leverage various verification and validation methods when constructing an assurance case. The main contributions of this paper are initial proposals for reusable assurance patterns based on mission patterns, and a high-level methodology for achieving a reference assurance case utilising these. An initial application of our approach is presented through a case study of road verge inspection using an autonomous robot.
Given the data-intensive nature of Machine Learning (ML) systems in general, and Large Language Models (LLM) in particular, using them in cloud based environments can become a challenge due to legislation related to privacy and security of data. Taking such aspects into consideration implies porting the LLMs to an on-premise environment, where privacy and security can be controlled. In this paper, we study this porting process of a real-life application using ChatGPT, which runs in a public cloud, to an on-premise environment. The application being ported is AIPA, a system that leverages Large Language Models (LLMs) and sophisticated data analytics to enhance the assessment of procurement call bids. The main considerations in the porting process include transparency of open source models and cost of hardware, which are central design choices of the on-premise environment. In addition to presenting the porting process, we evaluate downsides and benefits associated with porting.
The microservices architectural style has become the de facto standard for large-scale cloud applications, offering numerous benefits in scalability, maintainability, and deployment flexibility. Many organizations are pursuing the migration of legacy monolithic systems to a microservices architecture. However, this process is challenging, risky, time-intensive, and prone-to-failure while several organizations lack necessary financial resources, time, or expertise to set up this migration process. So, rather than trying to migrate a legacy system where migration is risky or not feasible, we suggest exposing it as a microservice application without without having to migrate it. In this paper, we present a reusable, automated, two-phase approach that combines evolutionary algorithms with machine learning techniques. In the first phase, we identify microservices at the method level using a multi-objective genetic algorithm that considers both structural and semantic dependencies between methods. In the second phase, we generate REST APIs for each identified microservice using a classification algorithm to assign HTTP methods and endpoints. We evaluated our approach with a case study on the Spring PetClinic application, which has both monolithic and microservices implementations that serve as ground truth for comparison. Results demonstrate that our approach successfully aligns identified microservices with those in the reference microservices implementation, highlighting its effectiveness in service identification and API generation.
Accurate active speaker detection is essential for natural verbal human-robot interaction. The available solutions have mainly focused on audio-based Voice Activity Detection (VAD), but these approaches become insufficient when audio is compromised or unavailable due to contextual factors. In such cases, the robot must infer whether someone is speaking using only video input. This paper establishes a foundation to enhance robot multimodal dialogue systems by integrating Visual Voice Activity Detection (VVAD) into the social robot Haru. Unlike prior studies that focus on detecting isolated speech segments, our method shifts toward accurately identifying speech boundaries, enabling the robot to handle turns from a visual perspective. We propose new metrics that better capture VVAD behavior in dynamic turn-taking scenarios, as well as overall speech and silence detection. Our results, which align with state-of-the-art benchmarks in isolated segments, highlight the effectiveness of VVAD in accurately identifying relevant speech instances with a simple expansion of a chunk-based algorithm to mark turns. These findings indicate the feasibility of incorporating VVAD into vision-based robotics and encourage further exploration in real-world applications to address remaining challenges in speech-based human-robot interaction, where visual detection is often the most reliable—if not the only—method for identifying speakers.
The recent rise of Large Language Models enables novel possibilities within social robotics. Human-robot interaction and verbal communication, in particular, stand to benefit significantly from how these models generate and interpret language. Recent experiments with our social robot revealed that users expect the robot to answer general knowledge questions and recall previous activities and conversations, creating bonds and engaging interactions. This paper introduces an Episodic Memory System for social robots based on offline Large Language Models designed to store and recall past experiences. The system records relevant episodes to answer user queries about previous activities and conversations and proactively suggests personalised activities based on those the user has previously enjoyed or completed. The goal is to define personalised interactions to promote human-robot bonding and engagement as humans do. We evaluated four open-source language models, comparing them in terms of success rate to answer questions about past episodes, computational requirements, and response time. We selected these models considering the robot’s hardware limitations and computational needs. We found the well-known Meta LLaMA 3 Large Language Model the best option, providing accurate responses in reasonable response times. We then integrated the model into the Mini social robot to show the system’s performance in a human-robot interaction case study.
