Robots in Society

A robot working in a sealed factory cage is an engineering challenge. A robot sharing a sidewalk with children, providing care to elderly people, or patrolling a public space is a social challenge with an engineering component. The technical capability to build a functional embodied AI system is necessary but not sufficient for its beneficial integration into society. Equally important are questions of law, trust, public acceptance, cultural meaning, and the deliberate design of how humans and robots will share space and relationship. These questions are not secondary — they determine whether embodied AI becomes a genuine social good or a source of harm and resentment.

Legal Frameworks for Robots in Public Space

Existing law was not designed for robots. Property law, tort law, and traffic law all assume that agents acting in public space are either humans or property controlled by humans. Embodied AI systems occupy an uncomfortable middle ground — they act autonomously but are not persons; they are owned by someone but make independent decisions. Self-driving vehicles have generated the most developed regulatory response. The US National Highway Traffic Safety Administration (NHTSA) has issued federal guidance and requires manufacturers to submit safety assessment letters, but as of 2025, no federal law mandates specific performance standards for AVs. Individual states have taken varying approaches: California, Arizona, and Texas have all permitted limited commercial AV deployment under different regulatory regimes. The EU has taken a more unified approach through its General Safety Regulation and the AI Act, requiring type approval processes for autonomous vehicle systems. Sidewalk delivery robots — small autonomous vehicles that navigate pedestrian environments — have prompted local ordinances in dozens of US cities, with rules ranging from speed limits and weight caps to requirements that a human operator be within a certain radius. San Francisco famously went through multiple regulatory cycles as delivery robots from companies like Starship Technologies and Kiwibot expanded. Drones operating in public airspace are regulated by the FAA in the US, with rules governing altitude, visual line-of-sight requirements, and no-fly zones near airports. Beyond-visual-line-of-sight (BVLOS) operations for commercial delivery remain a contested regulatory frontier. Social robots in healthcare and elder care occupy a different legal space. Devices that perform clinical functions are subject to FDA oversight as medical devices. Companion robots that provide social interaction but no clinical function fall into a regulatory gap — they are neither clearly medical devices nor consumer electronics in the traditional sense.

Trust, Acceptance, and the Uncanny Valley

Technical capability does not guarantee social acceptance. Research in human-robot interaction (HRI) has documented systematic patterns in how people respond to robots, and these patterns have significant design implications. The uncanny valley, first described by Japanese roboticist Masahiro Mori in 1970, refers to the phenomenon where robots that look almost but not quite human trigger feelings of unease or revulsion in observers, whereas robots that are clearly non-human or fully human do not. The effect has been replicated in numerous studies and complicates the design of social robots. A robot that looks slightly 'off' — near-human facial features that move imperfectly — can undermine trust even if its behavior is technically correct. Trust in embodied AI is multidimensional. Research distinguishes between competence trust (does the system do what it is supposed to do?), benevolence trust (is the system trying to help me, not harm me?), and integrity trust (does the system behave consistently with stated values?). All three dimensions must be established for a person to willingly share space with and depend on a robot system. Cultural attitudes toward robots vary significantly across societies. Japanese culture has historically shown higher acceptance of robots as social companions, partly shaped by Shinto beliefs that attribute spirit to non-human objects. Many Western cultures carry fictional associations of robots with danger and uprising — the cultural weight of Frankenstein, Terminator, and HAL 9000 shapes public perception in ways that pure technical communication cannot easily overcome. Designers of social robots must grapple with these cultural contexts deliberately.

Surveillance, Privacy, and Power

Embodied AI systems are, by their nature, equipped with sensors. A mobile robot navigating a building uses cameras, microphones, and lidar. Those same sensors can record everything the robot encounters. This creates a structural tension: the sensor data needed for the robot to operate is also surveillance data about every person the robot encounters. Several high-profile cases have illustrated this tension. Amazon's Astro home robot was criticized by employees during development for sending significant amounts of video data to Amazon's cloud. Boston Dynamics' Spot robot was deployed by several police departments for surveillance before public backlash prompted policy reversals. Security robots deployed in shopping malls and office buildings routinely collect footage of people who have not consented to being recorded. The power dynamics of robot surveillance matter. If robots are deployed primarily by employers to monitor workers, by landlords to monitor tenants, or by governments to monitor citizens, embodied AI becomes an instrument of surveillance asymmetry — those with power gain still more information and control over those without it. Designing against this requires both technical choices (minimal data retention, on-device processing, no cloud transmission of identifiable data) and policy choices (strong data protection law, worker consent requirements, transparency obligations).