It is argued that prompt design should be treated as a socio-technical problem rather than a minor implementation detail, requiring explicit capability boundaries, transparent behavioural assumptions, and context-sensitive safeguards to support reliable and interpretable HRI.
Abstract
Large language models (LLMs) are increasingly used for verbal interaction in social robots, yet prompt design in human-robot interaction (HRI) remains underspecified. As a result, robots may present hallucinated capabilities, unclear behavioural boundaries, and misleading personas. This paper develops a framework for prompt design in LLM-based robots and introduces a structured prompt template comprising eight functional components through which robot behaviour can be specified, bounded, and adapted. The framework is grounded in a review of prior LLM-based HRI work and complemented by survey and discussion data from HRI experts gathered at the Robo-Identity workshop at IEEE RO-MAN 2025 (N=27). The qualitative findings highlight limited legibility of robot personality, the need for user adaptation, and strong ethical concerns about safety, deception, and governance. Based on these findings, we present prompting guidelines accompanied by proof-of-concept template as a structured design and reporting aid for HRI research. We argue that prompt design should be treated as a socio-technical problem rather than a minor implementation detail, requiring explicit capability boundaries, transparent behavioural assumptions, and context-sensitive safeguards to support reliable and interpretable HRI.
While collaborative robots (cobots) are designed to augment industrial workflows, their integration is often hindered by not accounting for the individual personality traits that influence human-to-human collaboration. Although personality is a primary driver of trust and coordination, it remains underexplored in multi-agent settings in which a robot serves as an agent moderating human-to-human collaboration. To investigate this, we designed an augmented reality (AR) interface to facilitate human-to-human collaboration moderated by a cobot. Using a gamified pick-and-place task, we examined how personality influences workers' behavior and conflict resolution in a shared workspace. Results from a study with 16 participants, split into 8 pairs of workers, indicate that extraversion/introversion, in particular, was associated with differences in interaction strategies and perceptions of workload. Our findings suggest that future AR-enabled multiagent workflows should consider personality variability to better support collaborative settings.
Sebastian Heidrich, Christine Saeedi-Givi, John Liu et al.· International Conference on...· 0 citations
Humanoid social robots are becoming accessible to university laboratories beyond well-funded robotics groups, yet structured guidance for deploying them in student-led usability research remains scarce. This paper presents a reusable deployment framework built around Mads, the pib humanoid robot operated at the university's Usability Engineering and Verification Lab. The framework consists of two components: a Standard Operating Procedure (SOP) governing the full study lifecycle, and a browser-based Wizard-of-Oz (WoZ) control setup that enables students to conduct live user studies without requiring robotics or programming expertise. We designed this framework for low-cost platforms and describe a planned proof-of-concept study in which Mads acts as an active pause guide, leading participants through short movement breaks and delivering spoken feedback in real time. The framework and its protocol are offered as a documented artefact for the HRI and HCI education communities, with empirical validation planned as explicit future work.
HEART, a healthcare-specific evaluative architecture comprising Human-Centred Communication, Ethical and Trustworthy Deployment, Ethical and Trustworthy Deployment, Adaptive and Embodied Intelligence, Relationship Continuity, and Translational Healthcare Value is proposed.
Robots that follow natural-language instructions in everyday indoor environments must act on incomplete human utterances. Instructions often omit essential information, such as the identity of an out-of-view object, an intended destination, or the user's goal. Existing datasets contain little real-world situated dialogue and provide few practice-grounded criteria for deciding when a robot should act, confirm, clarify, or refuse. We retrospectively analyze a pilot Wizard-of-Oz study in which five participants performed everyday indoor tasks, including door opening, drawer opening, feeding, drinking, and cleaning, with a wheelchair-mounted mobile manipulator while the wizard responded without a formal communication policy. This preserved authentic user behavior but produced inconsistent robot-side decisions, motivating an explicit decision scheme. From 40 episodes, we derived a hierarchical taxonomy of six response modes (ANSWER, REPORT_DONE, REFUSE, CONFIRM, CLARIFY, ACT) and four ambiguity types (intent, referential, spatial, intelligibility). Two human annotators and an AI annotator applied the scheme to the pilot data. Clean-label rates were 91% and 89%, and Cohen's ranged from 0.72 to 0.95 across decision-point, mode, and ambiguity levels for both human-human and human-AI comparisons. Fine-tuning LLaVA-1.6-7B on taxonomy-derived labels for ACT and CLARIFY indicates the feasibility of training vision-language models using annotations from our taxonomy. Remaining boundary cases in decision-point identification and REPORT_DONE motivate a constrained protocol for more consistent dialogue collection.
Guang-Ping Liu, Nicholas Hawkins, Tipu Sultan et al.· 0 citations
Existing research on robot abuse and mind attribution has largely employed experimental methods focusing on individual responses, with less attention to how such responses continuously diffuse into public meaning through real platform interactions. This study takes the stability test video of Unitree's G1 humanoid robot on Douyin as a focal case, and uses large language model‑assisted inductive qualitative coding to analyze highly‑liked comments and their secondary replies. The findings reveal that although the video itself presents a technical performance test, the meaning‑making in the comment section almost completely bypasses the technical dimension. Commenters, through self‑reference, associate the robot with their own childhood, family, and other life experiences, endowing it with subjective experiential capacity beyond its technical attributes; among these, the interaction chain framed by childhood experiences is the largest and most influential. Concurrently, existing public discourse resources on the platform, such as school bullying, elder abuse, and animal protection, are repeatedly invoked and grafted onto the robot, elevating individual emotions into a public victim narrative. Based on these findings, this study proposes a three‑layer analytical framework of "video presentation—comment interaction—public discourse," revealing that the robot's victim identity does not originate from its own properties or the video content, but is collectively produced by commenters through self‑experience projection and reuse of existing discourses in ongoing interactions.
Ranran Zeng· Social sciences and humaniti...· 0 citations
Social robots have been widely explored as tools for autism intervention, yet this literature has focused predominantly on children and has rarely involved autistic adults as active contributors to design. This creates a mismatch between existing systems and the social-cognitive challenges autistic adults actually face in everyday life, including navigating ambiguous interpersonal contexts, managing conversational timing, and interpreting implied emotional meaning. To address this gap, we conducted an online focus group and co-design session with five autistic adults to explore what a social robot for social-cognition training should do, how it should interact, and under what conditions it would be genuinely useful. The 90-minute session combined open discussion with structured co-design activities on a shared digital whiteboard, and the resulting verbal and visual data were analysed using reflexive thematic analysis. The analysis yielded seven themes that define core design requirements: the robot should function as a scaffold rather than a substitute, prioritise authenticity over comfort, provide personalised and user-controlled feedback, accommodate emotional self-awareness gaps, respect privacy and contextual boundaries, support rehearsal for real-world social situations, and remain configurable in identity, form, and expression. Together, the findings suggest that autistic adults envision the robot not as a companion or live social assistant, but as a private, configurable rehearsal partner designed to support independence over time.
Yuval Zohar, Mordi Benhamou, Guy Laban· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
