Social Robots Nudging in Cruise Terminals: A Large-Scale Field Study

Social Robots Nudging in Cruise Terminals: A Large-Scale Field Study Lucrezia Grassi1, Anna Novaresio2, Giampaolo Vitali2, Monica Cariola2, Alessio Tei4, Manuel Catalano5, Alberto Minoia3, and Antonio Sgorbissa1 Abstract—We present a field study investigating whether social robots can nudge cruise passengers to complete a digital satisfaction survey. Conducted across 15 non-consecutive days at the Stazione Marittima (Ponte dei Mille, Genoa, Italy), the study involved approximately 11,000 passengers under three conditions: (1) static QR-code posters, (2) robots autonomously promoting the code, and (3) robots with human staff support. Robots increased scans and survey submissions by more than ten times compared to posters, while staff support added only marginal gains. The study suggests that social robots can serve as effective nudging agents even in crowded, non-cooperative environments. Index Terms—Social Robots, Human–Robot Interaction, Nudging, Field Study, Persuasive Technology I. INTRODUCTION Social robots are increasingly deployed in public spaces such as shopping malls, airports, hospitals, and museums, where they must operate autonomously and interact naturally with heterogeneous and often untrained users. One promising application is nudging: encouraging people to perform actions beneficial to themselves or the community through subtle social cues rather than explicit commands. Research has shown that robots can foster hand hygiene compliance in hospitals [1], encourage correct recycling behavior in public spaces [2], and increase questionnaire participation in cultural settings such as museums [3]. These examples illustrate the persuasive potential of social robots, but most deployments occurred in semi-controlled environments, with cooperative users and longer interaction times. High-throughput transport hubs are especially challenging: passengers move quickly in crowded spaces, focused on immediate goals (e.g., check-in). Prior airport studies showed the feasibility of socially aware navigation and guidance, but evidence remains limited on whether robots can influence behaviors that travelers might otherwise neglect [4]. In this context, we investigate whether social robots can act as effective nudging agents in a particularly demanding environment: the check-in area of the Stazione Marittima (Ponte dei Mille, Genoa, Italy), where cruise passengers transit 1DIBRIS Department, University of Genoa, Via all’Opera Pia 13, 16145 Genoa, Italy. 2IRCrES, National Research Council (CNR), Strada delle Cacce 73, 10135 Turin, Italy. 3Stazioni Marittime S.p.A., Ponte dei Mille, 16123 Genoa, Italy. 4DIEC Department, University of Genoa, Via Vivaldi 5, 16126 Genoa, Italy. 5Soft Robotics for Human Cooperation and Rehabilitation, Italian Institute of Technology (IIT), Via San Quirico 19D, 16163 Genoa, Italy. Corresponding author’s email: [email protected] Fig. 1. Pepper and AlterEgo deployed in the check-in area of the Stazione Marittima during the experiments. with the sole objective of boarding. In such a setting, persuading travelers to engage in a non-essential and unscheduled task—completing a digital customer satisfaction questionnaire—is inherently difficult. Our deployment involved about 11,000 passengers across 15 days, constituting one of the largest real-world studies of persuasive social robots. We test two hypotheses: (H1) social robots significantly increase engagement (QR-code scans and survey completions) compared to static posters; and (H2) the persuasive effect of robots would be further enhanced by human staff support. By combining a bilingual, diversity-aware conversational system with stakeholder co-design, we evaluate whether autonomous robots can provide a measurable behavioral impact in a noncooperative, time-pressured environment. II. METHODOLOGY The deployment was based on the CAIR architecture [5], a modular client–server framework supporting bilingual, diversity-aware human–robot interaction. The server hosted three core services: (1) the Hub for encrypted communication with multiple clients, (2) the Dialogue Manager combining OWL2 ontologies and large language models (LLMs) to generate utterances that were flexible yet constrained to deploymentrelevant topics, and (3) the Plan Manager for coordinating social actions. Speech recognition relied on Microsoft Azure APIs, while OpenAI GPT-4o produced natural language sentences guided by ontology constraints. Two robot platforms were employed. Pepper, a commercial humanoid, ran a Kotlin-based Android client handling audio capture, server communication, multimodal rendering (speech 2025 I-RIM Conference October 17-19, Rome, Italy ISBN: 9788894580570 10.5281/zenodo.17629738 111 with synchronized gestures), and displayed the QR code directly on its tablet. AlterEgo, a research prototype from the Italian Institute of Technology (IIT), ran a ROS-based client across two onboard computers: one for perception and dialogue, one for motor control. Lacking a screen, AlterEgo carried a printed QR code on its chest. In both cases, the CAIR clients ensured consistent conversational behavior across heterogeneous hardware. Initial field trials revealed that passive robots were largely ignored in the crowded check-in hall. To address this, we introduced programmed interventions: scheduled greetings, reminders, or gestures triggered every few minutes when robots were idle, to regain attention and repeatedly promote the survey. A lightweight navigation support layer was also added, allowing hidden operators to reposition robots to maintain visibility during peak boarding phases, while avoiding the risks of full autonomous navigation in dense crowds. The system and experimental setup were refined through a co-design process with Stazioni Marittime, involving iterative meetings and in-situ tests with management, logistics, and communication staff. Stakeholders emphasized the importance of bilingual interaction (Italian/English), explicit transparency in the robot’s role (displaying the QR code and explaining the survey), and minimal disruption to passenger flow. III. EXPERIMENTS AND RESULTS The study followed a between-subjects design involving a total of N= 10,754 passengers across 15 boarding days, constituting one of the largest real-world deployments of persuasive social robots in a transport hub. The study involved three conditions: (C1) static QR-code posters placed at the center of the hall (baseline); (C2) Pepper and AlterEgo autonomously promoting the QR code through speech, gestures, and programmed interventions; (C3): same as C2, with two human staff members who intervened only after the robot had attracted initial attention, clarifying questions or gently encouraging passengers. Each condition was tested on five non-consecutive days, balanced across ship types and dates. Boarding passengers were naturally exposed to the setup without prior recruitment. Engagement was measured at three levels: (i) QR-code scans, (ii) completion of the core satisfaction questionnaire about terminal services, and (iii) completion of a robot questionnaire on perceptions of the robots. Table I reports aggregate results across conditions. Robots substantially outperformed the static baseline: QRcode scans rose from 0.32% of passengers in the poster condition to 4.27% with robots, and 5.03% with robots plus staff. A similar pattern emerged for questionnaire completions: from almost zero in the baseline (0.02%) to 1.12% (robots) and 1.69% (robots plus staff). The robot questionnaire, offered only after completing the core one, was filled in by 0.24% and 0.40% of passengers in the robot conditions, while none completed it in the baseline. The questionnaire was accessible in all conditions as it addressed passengers’ overall attitude toward robotic technologies. TABLE I PASSENGER ENGAGEMENT AGGREGATED BY CONDITION. PERCENTAGES ARE RELATIVE TO BOARDED PASSENGERS. CORE Q. = CORE QUESTIONNAIRE; ROBOT Q. = ROBOT QUESTIONNAIRE. Condition Boarded QR Scans Core Q. Robot Q. C1 4,096 13 (0.32%) 1 (0.02%) 0 (0%) C2 3,395 145 (4.27%) 38 (1.12%) 8 (0.24%) C3 3,263 164 (5.03%) 55 (1.69%) 13 (0.40%) Statistical analyses (chi-square and Fisher’s exact tests) confirmed that both robot conditions significantly increased engagement relative to posters, while the difference between robots-only and robots-plus-staff did not reach significance. IV. DISCUSSION The study clearly confirmed H1: passengers were about 14 times more likely to scan the QR code and 46 times more likely to complete the core questionnaire when robots were present. H2 was not supported: adding staff produced only modest, non-significant improvements. Engagement decreased with effort, from scans to core and robot questionnaires, which is expected in a time-pressured boarding context. Importantly, transparency was ensured: robots explicitly presented the QR code and explained the survey, making their persuasive intent clear. Overall, the deployment demonstrates that autonomous robots can function as effective nudging agents even in crowded, non-cooperative transport hubs. The CAIR architecture enabled adaptive, bilingual, diversity-aware interaction tailored to the terminal context, while programmed interventions and lightweight teleoperation contributed to success. Future work will analyze questionnaire responses and explore long-term integration of robots into port operations. ACKNOWLEDGMENT This work was carried out within the framework of the project “RAISE – Robotics and AI for Socio-economic Empowerment,” as part of the activities of Spoke 2 and Spoke 4, and has been supported by the European Union – NextGenerationEU. REFERENCES [1] T. Worlikar, S. Goudar, S. Chalke, and et al., “Is it feasible to use a humanoid robot to promote hand hygiene adherence in a hospital setting?” American Journal of Infection Control, vol. 49, no. 12, pp. 1525–1529, 2021. [2] S. Y. Lo, Y. Zhou, C. M. Y. Cheung, and et al., “Robots as persuaders to promote recycling,” International Journal of Social Robotics, vol. 14, no. 2, pp. 491–503, 2022. [3] T. Natori, T. Iio, Y. Yoshikawa, and et al., “Impact of table-top robots on questionnaire response rates at a science museum,” International Journal of Social Robotics, vol. 17, pp. 729–742, 2025. [4] R. Triebel, K. Arras, R. Alami, and et al., “SPENCER: A socially aware service robot for passenger guidance and help in busy airports,” in Field and Service Robotics: Results of the 10th International Conference, D. S. Wettergreen and T. D. Barfoot, Eds. Cham: Springer International Publishing, 2016, pp. 607–622. [5] L. Grassi, C. T. Recchiuto, and A. Sgorbissa, “Enhancing LLM-based human-robot interaction with nuances for diversity awareness,” in 2024 33rd IEEE International Conference on Robot and Human Interactive Communication (RO-MAN). IEEE, 2024, pp. 2287–2294. 112