Perceived Safety
- Also called: subjective safety
- Category: Control › Human in the loop
- Tags: human-robot-interaction, safety, user-studies, trust, evaluation
How safe, calm and unthreatened a person feels when interacting with or near a robot, as distinct from how safe the robot objectively is.
Definition
Perceived safety is a person's subjective sense of being safe around a robot: the absence of fear, stress or a sense of threat during actual or imagined interaction. Its opposite, perceived danger, is the anticipation that the interaction could cause harm. Perceived safety is separate from physical safety. A robot can be engineered to stay within strict force and speed limits and still make people tense, and a robot can look harmless while carrying real risks.
Many factors shape perceived safety. Motion properties are among the strongest: speed, acceleration, how close the robot comes, the direction of approach, and whether its motion is smooth and predictable. Robot size, appearance, sound, and whether it signals its intentions through lights, gaze or legible motion also matter, as do the person's experience with robots, their control over the interaction and their trust in the system. Perceived safety tends to change with exposure, which makes it sensitive to novelty effects.
It is measured in three main ways. Questionnaires ask people to rate their feelings, for example the perceived safety subscale of the Godspeed questionnaire or newer dedicated scales for perceived danger. Physiological measures such as skin conductance and heart rate capture arousal during the interaction. Behavioral measures look at what people do: how far away they stand, whether they hesitate, flinch or slow down their own work. Each has weaknesses, so studies often combine them.
Why it matters
People who feel unsafe near a robot work more slowly, keep their distance, avoid using it or refuse to accept it at all, even if the robot is physically safe. Designing for perceived safety alongside physical safety is therefore necessary for collaborative, service and assistive robots to be useful in practice.
Key applications
- Tuning speed, distance and approach paths of collaborative and mobile robots
- Evaluating user experience in human-robot interaction studies
- Designing intention signals such as lights, sounds and legible motion
- Acceptance studies for assistive, care and service robots
- Comparing robot designs or behaviors before deployment
In practice
A team deploying a mobile manipulator in a hospital corridor tests three approach behaviors: a direct path at normal speed, a slower path, and a path that curves to approach from the front and pauses before reaching out. Each participant experiences all three in random order while wearing a wristband that logs skin conductance, and afterward rates each on a perceived safety questionnaire. The team also measures how far participants step back. The curved, slower approach with a pause gets the best ratings and smallest step-backs, so it becomes the default behavior near people, while faster motion is kept for empty corridors.
Common misconceptions
- If a robot meets safety standards, people will feel safe — compliance with physical safety requirements does not guarantee that users feel comfortable
- Perceived safety only depends on the robot — the person's experience, sense of control and context strongly affect it
- One questionnaire item is enough to measure it — short scales can be unreliable, so studies often combine validated scales with behavioral and physiological data
- Perceived safety stays constant — it often changes as people become familiar with a robot
Who should know this
- HRI researchers designing user studies
- Engineers tuning motion near people
- Product designers of service and assistive robots
- Safety engineers working on collaborative workstations
Related technologies
- Legible Motion Planning
- Proxemics
- Galvanic Skin Response Sensing
- Speed and Separation Monitoring
- Likert and Semantic Differential Scales
Related terms
- Godspeed Questionnaire
- Physical Human-Robot Interaction
- Cobot
- Novelty Effect
- Safety Filter
- Trajectory Planning
Related projects
- evo: De-facto standard tool for evaluating and plotting odometry/SLAM trajectories
- Reachy Mini: Open-source desktop robot from Pollen and Hugging Face with an expressive 6-DoF head
- RoboDojo: Manipulation benchmark that runs in Isaac Sim and on real arms
- Inspect Robots: Define a robotics benchmark once, run any policy on any robot
- SenseNova-RoboRSI: Open runtime that runs, scores and improves robot agents driven by language models
- RAI: Vendor-agnostic agent framework for ROS 2 robots and physical AI