[Research Poster] When Robots Spill the Beans: Exploring Transparency Declarations in Human-Robot Interaction

When Robots Spill the Beans: Exploring Transparency Declarations in HumanRobot Interaction Helpful André Helgert, Lukas Erle, Andre Dittmann, Sabrina C. Eimler, Carolin Straßmann [email protected] University of Applied Sciences Ruhr West, Germany Take-Home Message Theory Research Goal Method Results & Interpretation Take-Home Message Data transparency is vital: Social robots harvest personal data, raising privacy & security concerns (Lutz et al., 2019);disclosure allows informed choices (Wang et al., 2021; Selkowitz et al., 2016) HRI is cognitively demanding: Robot interactions trigger social schemas (Rosenthal-von der Pütten et al., 2013), higher load than browsing a website (Guznov et al., 2020). Multimodal transparency may cause information overload → potential backfiring effect. Unimodal transparency conveys clearer messages → better understanding & stronger automation perception. More is not always better: Multimodality doesn't automatically enhance clarity or acceptance. Although unimodal transparency was perceived as more effective, multimodal strategies should not be dismissed. They offer valuable potential in contexts requiring accessibility or tailored communication for diverse user needs. How do robots deliver transparency? Research tackles what to explain but pays far less attention to how (Schött et al., 2023) - that is, the communication channel (voice, screen, gesture, poster, etc.) a robot should use to convey transparency in the first place. Compare different transparency declaration types: •Unimodal (voice / screen / poster) •Multimodal (voice + screen, voice + poster) •No transparency declaration regarding users’ perception, understanding, trust, and performance. (Convenience) Sample: N = 95; M= 24.64, SD = 9.40 VR as research instrument Task: Register for a library card → enter personal data and take a photo 6 Conditions: 1x control, 3 × unimodal, 2 × multimodal Understanding Participants who received a unimodal transparency channel—voice, tablet, or poster—understood the data-handling process significantly better than those in the control and multimodal group Automation Unimodal delivery also made the robot feel more autonomous; adding a second channel or giving no explanation at all both scored lower. Sociability The same “less-is-more” pattern emerged: unimodal delivery left the strongest social impression, whereas multimodal explanations dampened it. Although unimodal transparency was perceived as more effective, multimodal strategies should not be dismissed. They offer valuable potential in contexts requiring accessibility or tailored communication for diverse user needs. Multimodal transparency may cause information overload → potential backfiring effect. Unimodal transparency conveys clearer messages → better understanding & stronger automation perception. More is not always better: Multimodality doesn't automatically enhance clarity or acceptance. Paper References Procedure Pre-Questionnaire VR-Tutorial VR-Interaction Post-Questionnaire Demographie, Robotic-/VR-Experience Robot explains control system in VR Disclosure of private data to robot Scales: understanding (self-developed), trust in automated systems, SERVQUAL, RAS, HRIES, TAM3