Quantifying the Perceived Sense of Control in Human-Robot Interaction by Exploiting the Hand Blink Reflex

Quantifying the Perceived Sense of Control in Human-Robot Interaction by Exploiting the Hand Blink Reflex Bernardo Brogi1, Tommaso Lisini Baldi1, Gionata Salvietti1,2, and Domenico Prattichizzo1,2 Abstract—This study presents a novel approach based on the Hand Blink Reflex (HBR) to quantitatively estimate the perceived sense of control during human-robot interaction. In twenty participants, we measured HBR while a robotic arm entered their defensive peripersonal space under three control modes: autonomous, discrete, and continuous control. As expected, HBR amplitude increased by up to 57% as the robot neared the user, but crucially, increased human control confidence corresponded to reduced HBR amplitude, specifically 23% in discrete mode and 9% in continuous control. This method offers a quantitative index of human confidence in robot control, paving the way for safer, more effective collaborative interfaces and algorithms. Index Terms—Human-robot interaction, Sense of control, Hand Blink Reflex, Peripersonal space I. INTRODUCTION As robotic systems are expected to collaborate closely while sharing the same environment, understanding how humans perceive and trust robotic movements is critical. One central aspect of this trust is the Sense of Control (SoC), an individual’s perception of being the author of a given action. However, SoC in human-robot interaction (HRI) has largely been studied through subjective questionnaires, lacking a robust quantitative metric. This study introduces a novel approach to quantify SoC using the Hand Blink Reflex (HBR), a well-established neurophysiological defensive response. The HBR is an automatic eye-blink elicited by electrical stimulation of the median nerve and recorded via electromyography (EMG) from the orbicularis oculi muscle. Critically, its amplitude increases when a potential threat is perceived near the face, making it a valuable proxy for assessing unconscious threat perception [1]. The HBR is modulated by the Defensive Peripersonal Space (DPPS), a projected interface between the body and the world with a protective function for the body. When objects enter this zone, the nervous system increases vigilance to protect against possible harm. The more a person feels in control of the object approaching their DPPS, the lower the perceived threat, and the weaker the HBR. By leveraging this mechanism, we propose to measure changes in HBR amplitude under varying degrees of robot control to infer the user’s implicit confidence and sense of agency in shared environments [2]. II. METHODOLOGY To evaluate the hypothesis that the HBR amplitude is influenced by the proximity of the robotic arms to the peripersonal space, and that increased human confidence in robot control may reduce the amplitude, we designed a comprehensive experimental evaluation with twenty participants that performed the following four experiments in separate sessions (Fig. 1b): 1 are with the Department of Information Engineering and Mathematics, University of Siena, Italy. {surname}@diism.unisi.it 2 are with the department of Humanoids and Human Centered Mechatronics, Istituto Italiano di Tecnologia, Genova, Italy. (a) Exp. 3 (Discrete Control) Far Near Far Near Exp. 2 (Autonomous) Far Near Exp. 4 (Continuous Control) NearFar Exp. 1 (Hand) (b) Fig. 1: Experimental setup. In a) a representative user during Exp.2. in b) a graphical representation of the four different experiments. In Exp. 1, the hand was stimulated in “far” and “near” positions relative to the face. In Exp. 2, an autonomous robotic arm moved near or far from the face. Exp. 3 used a button for discrete control, while Exp. 4 allowed continuous speed control via joystick. Red circles indicate the robot’s endeffector positions, blue ones show the control interfaces. • Exp. 1 (Hand): Baseline validation. HBR was measured when participant’s hand moved ”near” and ”far” from their face. • Exp. 2 (Autonomous): A robotic arm entered the DPPS autonomously. HBR was measured when robot moved ”near” and ”far” from their face. • Exp. 3 (Discrete Control): The robot was controlled via an on/off button. HBR was measured when robot moved ”near” and ”far” from their face. • Exp. 4 (Continuous Control): Participants used a joystick to control the robot’s speed and motion continuously. HBR was measured when robot moved ”near” and ”far” from their face. All setups included “near” ( ≈6 cm) and “far” ( ≈60 cm) target zones relative to the subject’s face. At the debriefing, a post-experience single-item questionnaire in the form of a Visual Analog Scale (VAS) was administered, asking participants to rate their subjective experiences related to the SoC of the robot (i.e., the perceived agency). Electrical stimuli were delivered to the right median nerve at the wrist using surface electrodes and spaced at intervals over 30 seconds to prevent habituation. EMG activity was recorded from 2025 I-RIM Conference October 17-19, Rome, Italy ISBN: 9788894580570 10.5281/zenodo.17629642 59 Exp. 1 (Hand) Far Near Exp. 2 (Autonomous) Near Far Exp. 3 (Discrete Control) Near Far 0 50 100 150 ms 0 10 20 30 40 50 60 70 80 90 uV Exp. 4 (Continuous Control) Near Far 0 50 100 150 ms 0 10 20 30 40 50 60 70 80 90 uV Fig. 2: Results of the four experiments. Panels show averaged EMG waveforms for “near” (red) and “far” (blue) conditions with standard deviation bands. HBR increased when the hand or robot approached the face, with smaller differences under higher user control, especially in Exp. 4, where full control minimized the response. the orbicularis oculi muscle and sampled at 10 kHz, as visible in Fig. 1a. Signals were filtered, rectified, and analyzed to extract reflex amplitude (as area under the curve), onset latency, and duration. III. RESULTS The study confirmed that the amplitude of the HBR increases when a potential threat, such as a hand or a robotic arm, enters the subject’s DPPS (Fig. 2). In Experiment 1, when participants moved their own hand near the face, the HBR amplitude significantly increased compared to the far condition, replicating prior findings [1] and validating the experimental setup. In Experiment 2, an autonomously moving robotic arm triggered the strongest HBR response, with a 57% increase in amplitude when it approached the face, indicating a heightened perception of threat in the absence of user control. In contrast, Experiment 3 showed that, when participants could start and stop the robot’s motion via a button (discrete control), the HBR increase was reduced to approximately 23%. Finally, in Experiment 4, when participants continuously controlled the robot’s motion and speed with a joystick, the HBR increase dropped to just 9%, suggesting minimal perceived threat. Across all conditions, onset latency and response duration were also modulated by proximity, supporting the amplitude results. As visible in Fig. 3, a statistically significant negative correlation was observed between perceived Sense of Control (as rated on the VAS) and the percentage increase in HBR amplitude. In other words, this confirms that greater user control over the robot corresponds to a reduced reflexive response, revealing higher implicit confidence in the interaction. IV. DISCUSSION The main result proposed with this work is the demonstration that the HBR is a reliable physiological measure of the user’s perceived control during human-robot interaction. When the robot 0 40 80 0 25 50 75 100 Perceived Sense of Control AUC HBR percentage increase [%] Autonomous Continuous ControlDiscrete Control R = -0.48, p< 0.001 Fig. 3: The correlations between perceived SOC and HBR amplitude percentage increase. The plot reports data obtained for all the subjects in Exp. 2, 3 and 4. Spearman’s rankorder correlation test revealed a statistically significant negative correlation between the perceived sense of control and HBR percentage increase between near and far conditions. operated autonomously, participants exhibited the strongest HBR responses, indicating a greater defensive reaction due to the absence of control. As user involvement is increased, from discrete to continuous control, the reflex response was progressively attenuated, reflecting a growing sense of safety. Importantly, even partial control (as in the discrete condition) reduced the HBR amplitude compared to full autonomy, suggesting that any degree of agency can modulate unconscious threat responses. However, continuous control nearly abolished the defensive reaction, and thus positioning the robot as a less threatening extension of the self. The negative correlation between HBR amplitude and subjective ratings of Sense of Control supports the idea that HBR modulation reflects implicit trust in the robot’s behavior. These results suggest that neurophysiological measures like the HBR can complement or even substitute subjective questionnaires when evaluating the effectiveness and acceptability of robotic interfaces. V. CONCLUSIONS This study presents a novel method for quantifying the perceived Sense of Control in human-robot interaction. The results show that HBR amplitude is modulated not only by the proximity of a robotic agent to the face but also by the degree of control the human user has over the robot. A higher user control correlates with lower reflex responses, indicating increased confidence and reduced perceived threat. Unlike traditional subjective assessments, this approach offers an implicit and reproducible measure of user experience, potentially enabling real-time evaluation of interface effectiveness. Future research will explore how additional factors, such as haptic feedback and sensory augmentation, influence SoC and HBR modulation, with the goal of optimizing human-robot collaboration in dynamic, shared environments. REFERENCES [1] C. F. Sambo et al., “Defensive peripersonal space: the blink reflex evoked by hand stimulation is increased when the hand is near the face,” Journal of neurophysiology, vol. 107, no. 3, pp. 880–889, 2012. [2] T. L. Baldi et al., “Quantifying the sense of control through the hand blink reflex in human-robot interaction,” IEEE Transactions on Affective Computing, pp. 1–11, 2025. 60