Defining a Robot
Last updated: Sep 17, 2023
In this article, we attempt to define a robot by examining the underlying conditions that distinguish robotic systems. It is impossible to satisfy all opinions with one definition, but being objectively right is not the goal. Instead, the aim is to better understand the nuances surrounding robots.
The word robot has literary origins (specifically, Czech) meaning forced labor, vaguely implying that a robot is a machine-worker. Curiously though, in the initial literary context, robots were described as artificial biological organisms, rather than the popular mechanical interpretation (e.g., automaton, device, machine).
The definition of a robot changes across cultures and has changed throughout time. Some machines considered robots by Japanese standards are not by American standards . But we can try to find overlap and force a consensus.
Nowadays, we can agree that a robot is a system, comprised of various components and subsystems, that has some objective, some awareness of itself or surroundings, some ability to physically interact with its environment (move, touch, grab, drill, shoot, …), and potentially some level of autonomy (in the engineering sense; automatic, not driven; no detailed direct human operation). It is difficult to reach an agreeable and precise definition as opinions often quickly disperse.
Consider the example of teleoperated systems, still considered by some to be robots, but also excluded by others. Nevertheless, many roboticists actively working on teleoperation have made it more intelligent.
Another example is in wearable systems (e.g., exoskeletons, exosuits, prosthetics). Again, whether a particular system should be called a wearable robot, or rather, simply a wearable system, is up to interpretation.
There is no satisfying definition for robotics as the field is too vast, somewhat fragmented, and not mature enough. The need for a clearcut and comprehensive definition is certainly appealing. But, to settle on one today would be premature and reductive.
Still, it doesn’t hurt to try, while striving to be less incorrect. Here is a proposed definition, in both long- then short-form.
A robot is any goal-oriented and human-made system, typically a machine but not necessarily, endowed with some capabilities for sensing, perception, cognition, planning, control, and acting in the physical world; while subject to human authority, in the form of high-level command or low-level control.
A robot is any goal-oriented and human-made system under human authority that can sense, plan, and act in the physical world.
Discussion & Implications
The above definitions may be dissected and discussed for depth and clarity.
A robot is a system.
We may emphasize the word system rather than machine . There is no reason why a robot cannot be grown as an organism, rather than built as a machine. If it can be designed for a goal, commanded or controlled, with some level of autonomy, then the choice of the corpus, body, or mechanism is, in principle, arbitrary. Machines are much better understood, and can be designed to be more predictable, and thus today are almost always preferred to organisms.
This systems-view also alludes to robotic swarms. In some swarms, each individual machine can be complex enough to be a robot in its own right. But, in principle, individual machines that are too simple to be considered robots, can be collectively considered a robotic system, provided that the swarm can collectively meet the definition.
A robot directly interacts with the physical world.
In other words, if it cannot mechanically interact with the physical world, it is less robot and more device, or sometimes even a computer. If it cannot sense nor plan, then it lacks the ability to adapt, which is closer to traditional machines than to ‘intelligent’ robots. Many consumer “robot assistants” or “robot companions” get branded as robots, and those that interact physically indeed are; but those who cannot are closer to computers than to robots, even if they have speech features and sophisticated natural language processing.
A robot is made by humans, for humans.
Despite the musings of popular culture and the entertainment industry, robots are a human invention: designed by people, built by people, and ruled by people. Bluntly put, a robot is an artificial creation indentured to serve a specific goal or purpose defined by its human creators, owners, or programmers.
One implication here is that a living system found in nature cannot be converted into a robot. A machine modeled after a horse can be a robot, but a live horse controlled using implants is closer to a rigidly-controlled animal. A robot, in this regard, may be thought of as “artificial life” with “human masters”. One curious scenario is: would a dead horse, i.e., a cadaver, taxidermied and repurposed into an obedient ‘biomachine’ (cyborg?) be considered a robot? Potentially yes, it is somewhat artificial life, that is used to serve humans.
Another implication is about authority. Since robots are built by people to serve people, then it must remain under human authority to be considered a robot. If that subordination is breached, then the system is no longer a robot, but a rogue or independent machine. My interpretation of this is as follows:
Robots may be given some forms of intelligence, but by design, they lack free will. Otherwise, we create, at best, independent machines, at worst, rogue machines.
Many fictional movie characters we call ‘robots’, especially ‘killer robots’, such as the famous Terminator, are not under human authority and so are not robots in the normative sense. There are other tricky scenarios: Consider a robot, once working for people, now fallen under the authority of a hypothetical and hostile non-human entity. It is a ‘robot’ vis-à-vis the entity, but perhaps just a killer machine to people. What about robots commanded by one group of people, instructed against another group of people? It gets tricky, both nominally and ethically.
It is in this context that many laws of robotics have been proposed or imagined, the most famous of which are Asimov’s in literary contexts.
When a robot exhibits behavior that may be perceived as (supposedly) free-willed, then it should, in principle, cease being considered a robot (a bound system), but rather a free entity. But such scenarios are, so far, only in fiction.
Consider these systems, and argue whether each is closer to a robot, or not.
In everyday practice, it is perfectly fine to call robot-esque machines robots, for concision. This article was merely a pedantic attempt to feed thought and debate. Though eventually, legislation should take the nomenclature and classification of automation more seriously (in Japan, there is a subjective albeit interesting classification of robots into six levels of “generality” or autonomy). Clarity would help to design sensible laws and stimulate less fuzzy discussions surrounding automation regarding moral agency, economic policy, and so on.
Certainly, the definition remains fuzzy. But in many ways, the failure to cement the meaning of “robot” nor reach a singular interpretation of it, is that which brings much richness and diversity to the field.
 Saeed B. Niku, §1.3, Introduction to Robotics: Analysis, Control, Applications, 2019, Wiley.