News Ticker

Jamie Todd Rubin’s Three Criteria for Robots

John W. Campbell wrote an editorial in the October 1939 Astounding in which he talked about his future vision of robots. By late 1939, robots were already a growing trope of science fiction and within the next year, Isaac Asimov would begin revolutionizing that trope with the first of his Robot stories, giving us the vision of robots that we have today. Campbell, in this editorial, argued,

The world is already full of thousands of robots, working busily and intelligently on their assigned tasks. Naturally, they don’t appear manlike because a man makes a very poor instrument for any given task.

He gave as an example, “telephone switching robots” which don’t look manlike at all. I thought about telephone switches today and something inside me rebelled at the idea that a telephone switch could be considered a robot. The question Campbell was getting at, even if not explicitly stated, was: how does a robot differ from any other machine?

I batted around ideas with a friend whose office is next to mine. Obviously, a robot is a human-labor-saving device. But a computer is also a human-labor-saving device and we don’t consider our laptops, iPads and iPhones robots. I decided, therefore, that there had to be a mechanical element to a robot: moving parts made up of one or more various machines (gears, belts, levels, wheels, planes, etc.) And this mechanical element has to translate into some kind of controlled motion. Thus, an aircraft autopilot system which is controlled by computer, is still robotic because the yoke and throttle and other surface controls move.

“But what about automatic locks in a car?” my friend argued. That, too, seems to meet the criteria I’ve defined, but automatic locks don’t feel robotic. Clearly, something is lacking in my definition, but what? I considered this for several days and decided that while a robot might be a computer-controlled mechanical system involving controlled motion, there was one other element, key, which I’d left out: the influence of science fiction stories.

The earliest modern image of a robot (and the the word itself) probably comes from Karel Čapek‘s 1920 play, “R.U.R” (Rossum’s Universal Robots). The December 1938 issue of Astounding contained Lester del Rey’s classic story, “Helen O’Loy” about two men who alter an ordinary robot to have emotions. One of the men falls in love with the resulting robot and spends the rest of his life with her, and of course, she asks to be destroyed when he dies. In both the play and the story, robots are portrayed very much like humans.

del Rey’s story was unusual in part because it portrayed robots as benign. In the early days of science fiction, robots were often portrayed as evil or beserk. It is Isaac Asimov, in his Robot stories, who introduced computers into robots with his “positronic” brains. And he and Campbell corralled what a robot could do with the famous Law of Robotics, paving the way for robots that could be a boon to human society. (Of course, his novel The Caves of Steel investigated the dark side of this boon.)

Asimov’s robots stories (and later, the novels) spanned decades. They started out as simple puzzle stories with the puzzles involving the Laws of Robotics. But they evolved into rather touching pieces that danced on the borderlines of what makes a robot human. The climax of these stories was probably Asimov’s Hugo and Nebula-award winning “Bicentennial Man” (later expanded into The Positronic Man by Robert Silverberg, and made into a fairly bleh movie starring Robin Williams.)

These weren’t the only robot stories in science fiction, but Asimov’s robot stories certainly had the most influence. Many computer scientists turned to their profession because of Asimov’s stories with the result that what we see in robots today are clearly influenced by Asimov’s vision. Certainly robots like the floor cleaning Roomba is one example, with the pinnacle (so far) being Honda’s Asimo. (The name is no coincidence.) Motion pictures took these images a step further, allowing us to see on the big screen Philip K. Dick’s “replicants”; George Lucas’s “droids”; and most recently, the re-imagined Cylon’s of Battlestar Galactica.

It is this influence–this image of robots as depicted in stories like “Helen O’Loy” and “Runaround” and “Liar!” and “Bicentennial Man”–that I’d argue is the missing element to our modern definition of a robot. Thus, if asked to define what makes something a robot (as opposed to an ordinary machine or computer), I’d answer with what I’ll call:

Jamie’s Three Criteria for a Robot
  1. A robot must be labor-saving, performing a function that might otherwise be performed by a person and do so with better efficiency.
  2. A robot must be mechanical in nature and show motion; it cannot be just a computer or computer function.
  3. A robot must reflect the spirit of the ideas first imagined in science fiction stories; this reflection can be in form, function, or both.

The third item is perhaps the most important because it carries with it what science fiction writers have envisioned over the course of the last 90 years. It eliminates things like automatic door locks and telephone switches–which don’t seem to belong to the science-fictional realm of robots, but does include things like automatic pilot systems, house-cleaning robots, and even car assembly line robots.

I am satisfied with this definition of a robot, particularly because it pays homage to the vision and imagination of generations of science fiction writers, whose work is sometimes marginalized by our “literary” brothers and sisters. To me, it is just another example of the powerful influence that science fiction has had in the evolution of science and technology; how it has provided examples for possible avenues of investigation.

And, of course, I forgive Campbell his definition of a robot, since he was writing his comments at a time when the science-fictional notion of a robot was still in its infancy. The writers he developed, and many others had yet to add their visions and opinions to the mix.

So what is your opinion? How would you define a robot?

About Jamie Todd Rubin (31 Articles)
Jamie Todd Rubin is a science fiction writer and blogger. His fiction has appeared in Analog, Intergalactic Medicine Show, Apex Magazine, and most recently through 40K Books. He writes the Wayward Time Traveler column for SF Signal and vacations frequently in the Golden Age of science fiction.
Contact: Website

15 Comments on Jamie Todd Rubin’s Three Criteria for Robots

  1. I think it really depends from what someone calls “a robot”. The definition varies. Someone might call a computer with AI a robot. Someone else might only call robot a machine that moves (and, perhaps, is anthropomorphic) with some kind of intelligence.

    Also, “The word robota means literally “work”, “labor” or “corvee”, “serf labor”, and figuratively “drudgery” or “hard work” in Czech and many Slavic languages.” So, in that sense, a robot must be something that performs manual labour of some kind. But, as the word is being used today, I think it depend from the definition you use.

  2. DV, that gets to my point, which is you can define a robot in the same way you define a plant or an animal, through some kind of agreed upon taxonomy. I offer a very simple one above, but you might imagine there being a more complex structure to the taxonomy. For instance, human-like robots might have other attributes (asymmetry, for instance, or AI). “Helper” robots like the Roomba may not by asymmetrical or have any kind of AI. Assembly line robots might differ in some ways as well. But all of them are called robots because they have some broad elements in common that make them robots. In my opinion, those broad elements are the three I offer above.

  3. Whoops, I should have typed, KV. My bad. 🙁

  4. Very interesting question, and to give it a short answer, I don’t know. Expand that a little and I agree with KV that it depends on the defintition of the term robot.

    To look at it more closely I want to consider is HAL (2001: Space Odessy) a robot? It has intelligence, it is there to do the work of keeping the crew alive (at least thats what it was designed to do), When you consider that HAL is built into the ship it has mechanical features as well as well the fact that it does move. I would say that for me HAL fails your 3rd law, because it is called an AI in the book rather than a robot even if it does meet the rest of your criteria.

    I suppose that actually I think only your 3rd rule matters to me, the other 2 are dependant on so many different intreptations: A telephone exchange meets the 1st rule, and its an old mechanical one then it meets the 2nd (I once worked in one of the major telephone exchanges in London fitting the new digitial exchanges and used to go and visit the old switching room and must say it was an impressive experience) but it fails the 3rd rule, its ‘just’ a machine, it has nothing that would make me think of it as a robot.

  5. Andy, the question becomes a little more intuitive when you work your way up. I think Campbell was wrong to consider a telephone switch a robot. I wouldn’t consider a toaster a robot. I wouldn’t consider software a robot, even though Excel might be labor-saving as compared to a roll of paper and a slide rule. As for AI, that is hard to say. I would go along with you with HAL if you consider him part of the entire spacecraft. To that end, the AI/spaceship combinations in Jack McDevitt’s novels might also be considered robotic. Maybe. Isaac Asimov once wrote an essay on what makes something living versus nonliving and I think you need the same approach with robot vs. nonrobot. That is where a taxonomy can help get you partway.

    For that other part, that intangible spark, I depend on the vision of science fiction writers.

  6. Ironically, Rossum’s Universal Robots were actually androids, not robots.

  7. Michael, maybe we should extend biology’s binomial nomenclature to include robots and their variants. Androids would probably fall into the mix somewhere closer to the “what is the difference between a robot and a man” end of the spectrum: something Asimov addressed early on with “Reason!” and much later with “The Bicentennial Man”.

    The question here, of course, is what is the difference between a robot and a machine.

  8. In some cases, defining a Robot is like defining SF–“I know it when I see one”


    I do think that anything considered a robot is on that human-android spectrum, but on the far end of it.


    Also, what about autonomy? Every robot that I can think of, even with restrictions on what they can do, MUST be able to undertake autonomous actions. 

  9. Paul, when I was hashing out my criteria, I originally wrote the following for the second item:

    2. A robot must be mechanical in nature and show autonomous motion; it cannot be just a computer or computer function.

    I took out “autonomous” because I felt it limited the definition too much. A Roomba is not really autonomous, for instance, it just moves in a direction until it encounters an obstacle and then changes direction. Car assembly line robots are autonomous only in that they follow very specific manufacturing instructions. They don’t make decisions for themselves.

    Does a robot need to be able to make autonomous decisions? I could be convinced, I think.


  10. @ Jamie – Thinking about the old mechanical telephone exchanges I still can’t decide…. walking into the room and hearing one going full belt was a truly impressive and even awe inspiring experience, there certainly seemed to be something special/ otherworldly going on…. it felt ‘alive’…. but then it was quite simply to break the machine down and to understand what was actually happening as it made the connections, as the cogs clicked and whirred.

    I suppose my point is that its us humans who define what is and isn’t a robot, and that we do it depending on how much we can anthropomorphise the ‘machine’, its the giving of human characteristics that elevates it to the height of being a robot. Further to this there are also other scifi classifications that come into play… AI and Cyborg immediately spring to mind…

    Not that I am trying to rain on your parade as I think its a fascinating question and am enjoying thinking about it.

  11. Paul,

    and that we do it depending on how much we can anthropomorphise the ‘machine’, its the giving of human characteristics that elevates it to the height of being a robot

    Yes, and I think we are on the same page here. This is the point the Campbell was arguing against in his editorial–that humans were badly designed for many tasks and so designing a robot after a human was silly… but that something completely unhumanlike in nature could still be a robot. This is where he mentioned a telephone switch. As another example, would an elevator be considered a robot? It meets the first two criteria, certainly.

    It’s the third criteria that gets to your point,

    its us humans who define what is and isn’t a robot

    Most of us can probably agree on the first two, but what it is that science fiction has given us that imprints on our minds an image of what a robot should be. Or, as you put it early, “How is it that we know one when we see one?”

  12. Arg, I wrote “Paul,” and I meant, “Andy.” Doing too much at the same time. Sorry about that. :-/

  13. @Frank 😉

    I don’t think something needs to be ‘humanlike’ for it to be anthromorphised, many people name their cars and consider them to have personality. Working as a PC engineer for years I regard broken computers as tempremental women… you know those high maintenece women who do your head in, but when you get them onside they can be so much fun… until they have another tantrum.

    Not exactly sure where I am going with this yet, but I supose my point is a duality between an acknoweldgement that a robot is simply a machine that meets a couple of arbitary criteria… which is meaningless because there will always be exceptions that point out how ridiculous whatever criteria you create are. And the otherhand the acceptance that a robot is something special, its something more than just a machine. Often in ways that can’t be easily defined, because those definitions fall prey to the fact that they are case specific and as soon as you try to make general rules…

  14. I’m fond of the definition Douglas Adams put into the Hitchhiker’s Guide: “a mechanical apparatus designed to do the work of a man.”  Succinct and hard to beat, it incorporates the “must have moving parts” clause that distinguishes a robot from software and implies the multi-functional capability that I think distinguishes a robot from a simple device like an automatic door lock.  The only change I’d make is to change “man” to “person.”  I don’t think there’s any need to incorporate science fiction into the definition of “robot;” that could lead to circular logic when attempting to answer questions like “is Moya a ‘robot’?”

  15. David, that’s a great line and I’d forgotten all about it (it’s been 20 years at least since I’ve read HHG). It is more clear than Campbell’s 1939 definition. That said, I still take issue with Campbell’s definition of a telephone switch as a robot. Of course, I’m imaging the digital switches we have today not the massive switches he was speaking of in 1939. Your modified version of Adams’ definition might exclude it too, though I suppose the real question is: was a phone switch designed to the work of a person? certainly in old movies and TV shows there are switchboard operators, but I’m not sure that’s the same thing.

    I understand your point about circular logic with regard to including the influence of science fiction in the criteria, but I still feel that what we call robots today were influenced by what science fiction writers imagined robots to be, whether it be in fuction or design.

Comments are closed.

%d bloggers like this: