Alan Turing's article "Computing Machinery and Intelligence" (which you can read online) is commonly interpreted as introducing the Turing test. Briefly, the Turing test is when you have two computer terminals, one controlled by a human in different room, and another controlled by a computer. A second human then has to determine which terminal is the human-controlled terminal and which is the computer-controlled terminal.

However, if you read the actual text of the article, you will notice something strange. The test Turing proposes at the introduction is actually a test to see if someone can identify a terminal controlled by a man from a terminal controlled by a woman:

The new form of the problem can be described in terms of a game which we call the 'imitation game." It is played with three people, a man (A), a woman (B), and an interrogator (C) who may be of either sex. The interrogator stays in a room apart front the other two. The object of the game for the interrogator is to determine which of the other two is the man and which is the woman. He knows them by labels X and Y, and at the end of the game he says either "X is A and Y is B" or "X is B and Y is A." The interrogator is allowed to put questions to A and B thus:

It's only at the end of the article that Turing brings computers into play. He asks what would happen if a machine replaced the man, and whether the game would change. Turing reasons that if the third human identified the human man from the human woman at the same rate the third human identifies the machine pretending to be a man from the human woman, then the machine is indistinguishable from being a human.

We now ask the question, "What will happen when a machine takes the part of [the man] in this game?" Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman? These questions replace our original, "Can machines think?"

Note that gender continues to play a role in the test. The Turing test, as Turing proposes it, is whether a human can identify a machine from a terminal controlled by a woman. Never in his paper does Turing propose a test where the only thing a human has to identify a terminal controlled by a human whose gender is unspecified from a computer controlled terminal.

My question is, why talk about gender at all? In particular, why talk about gender at the very beginning of the paper, and only later discuss the (presumably) real topic of the paper, computers? Note that it's extremely unlikely that this is "random" (whatever "random" means in the context of literature).

This answer has gotten a lot of answers arguing that gender has nothing to do with the test. I suppose this is a good thing: it's good to get a lot of different viewpoints. However, a possible answer that no one has considered yet is that Turing's paper may be related to various tropes about AI and gender. An answer arguing along that lines is something I would really be interested in reading.

If you're interested in writing such an answer but worried that it will get buried under the other answers here, don't worry! I'll accept the first good answer I see, which will move the answer to the top of the answers.

If you're unconvinced that such an answer exists, I would encourage you to take a closer look at Tariq Ali's excellent answer, which cites a paper dedicated to rebutting interpretations of Turing's paper that focus on gender. I'm interested in an answer that explains the arguments that Tariq Ali's answer rebuts.

  • 2
    @Randal'Thor I disagree that the title is misleading. The final version of the Turing test isn't "can you tell a computer from a human," it's "can you tell a computer from a woman". (maybe I should edit the question to make this clear.)
    – user111
    May 1, 2017 at 23:40
  • 2
    When you say "Turing's paper may be related to various tropes about AI and gender", are you talking about the trope of AIs being portrayed as female that we see in films like Metropolis or Ex Machina? Because that could indeed be an interesting approach; one (who had more knowledge of the topic than me) could draw a parallel between the mixture of human and "other" inherent to an AI and the way women in some male-authored fiction have been portrayed as a mixture of human and "other".
    – Torisuda
    May 2, 2017 at 3:30
  • 4
    @MattThrower meta discussion: literature.meta.stackexchange.com/q/717/111
    – user111
    May 2, 2017 at 14:29
  • 3
    (this was reopened by 5 people agreeing that it should be open - the three VTROers, Hamlet, and me)
    – Mithical
    May 2, 2017 at 20:47
  • 3
    Removed my downvote, but I'm still not happy with the title. I don't agree that discussion of gender is necessarily "political subtext". (Sorry to keep badgering you with comments about this question, but evidently I'm not the only one who has issues with it :-) )
    – Rand al'Thor
    May 2, 2017 at 22:51

7 Answers 7


I dispute your claim that the Turing test, or Turing's article, is about gender. Extraordinary claims require extraordinary evidence; I see no hint in Turing's article that the test is about gender. If you think that it is, it is up to you to come up with evidence why that is a reasonable conclusion. Therefore, the only thing that is answerable in your question is: why does Turing use gender, to explain or make a point that is unrelated to gender?

I see no evidence that this is anything but the straightforward didactic device it appears to be.

Note that Turing uses the word sex, you use the word gender. I'll use the word sex, following Turing; the distinction is actually somewhat relevant, but not philosophically so.

In this answer will not distinguish between thinking, consciousness, sentience, soul possession and other similar but not identical concepts. Turing uses the word “thinking” and links it to some variations on these other ideas but does not make a clear distinction between them. I will do likewise, although much could be said on the topic.

Also, my perspective is that of a (former) computer scientist and occasional teacher of computer science. I am familiar with the concept of the Turing test and the historical context in which it was formulated but have not researched Turing's writings or the precise context in which he wrote this text.

Let us start with the historical context. In Turing's time, the concept of a machine that might think in a way that's similar to a human is rather novel, daring and prima facie rejected. The concept is not wholly new — numerous human creation myths and golem and other similar ideas have existed for longer than we have recorded history. What was new in Turing's time was the idea that the creation of a thinking mind could be understood in technological terms and not just as magic: a thinking mind would be constructed by engineers, who would use techniques that are in principle reproducible (see §3) rather than a magical insufflation of life by dint of opaque biological processes or divine grant. It was a rather shocking idea, which is why about than half of the article consists of a refutation of counterarguments.

So we have a philosophically daring idea which needs to be explained in a way that won't outright shock the audience and cause immediate rejection. An abrupt presentation of the idea would risk being emotionally rejected without serious intellectual consideration. This calls for a gradual presentation that first hooks the audience into a reasonable-sounding series of arguments, and then shows that if this idea is reasonable then so should be the more daring one.

The daring idea is that a human and a (sufficiently advanced) thinking machine cannot be distinguished through the way they think, and the way they communicate via language in the abstract. (Turing neglects the language aspect, assuming that a thinking machine would be able to master human language. I will likewise consider mastery of language as part of the requirements of thinking.) The human and the machine may be distinguished by physical characteristics — they may have different size and shape, they may have different means of concrete communication (typing vs holding a pen, different timbers of voice, etc.). The point of the setup of the Turing test is to argue that these are only superficial differences that do not matter to the crucial question of whether the machine thinks.

So we need to explain that there's a test for which a thinking machine and a human will be indistinguishable. But since this claim would immediately be rejected, we need to bring it up gradually.

Then let's start where we have two different humans who need to be distinguished. Nothing shocking about that. Human A is trying to pass off as human B, human B is trying to refute this, and the goal of the interrogator C is to determine which is which. With two specific humans, C would have an easy job, by asking something that only B would know. So the test doesn't really work with specific humans. It would work a lot better with categories of humans. This way, A could justify ignorance of a fact, or variation from a stereotype, because not every member of the category would have to know the fact or follow the stereotype.

We need to find a way to distinguish between two categories of humans who are not easily distinguished by their knowledge or mental ability. Yet these categories should be unambiguously distinct. Preferably those categories should have physical distinctions, to make the requirements for the setup of the test more evident.

Things like nationality (which would lead to knowledge and language differences) and age (not well-defined and could also have noticeable intrinsic differences) wouldn't work. On the other hand, sex perfectly fits the requirement. Note that this does not require to accept that there is no intrinsic mental difference between men and women. While Turing's society did make some distinctions in this respect, his audience would not automatically have jumped to the conclusion that the test would obviously reveal which is which. And in any case what matters is not that the test could not reveal which is which under any circumstance, but that it is plausible that the interrogator would be unable to decide.

Hence, first make the test be about sex. Then, once the audience has accepted the framework of the test as reasonable, switch the test to be about human vs machine. If the test is good enough to determine whether men and women can be distinguished as a category, then it's good enough to determine whether humans and machines can be distinguished as a category.

Once the test has been established in section 1, Turing does not come back to the idea of testing for sex (gender). The paper is not about a test for gender.

  • 2
    I disagree with your description of the test as being "about human vs machine". The final version of the test that Turing proposes isn't "can one tell a human from a machine," it's "is the rate that a human can successfully differentiate between a man and a woman the same rate that a human can successfully differentiate between a machine and a woman." While it's a plausible interpretation that the test starts by using gender as a way to make the test more accessible (which is why I upvoted this answer), you might want to edit the answer to take this into account.
    – user111
    May 1, 2017 at 23:46
  • 1
    It may have been remiss of Turing to not say so explicitly, but I've always taken the spirit of the paper as being "replace one of the humans with a machine". As strictly written, the original test is not about "telling the difference" but whether a man can pretend to be a woman ("A's object in the game to try and cause C to make the wrong identification" and "The object of the game for the third player (B) is to help the interrogator"). Again, although not explicitly said, to me the spirit of the game could have A and B reversed so the woman is trying to fool the interrogator.
    – TripeHound
    May 2, 2017 at 7:23
  • 3
    @Hamlet The purpose of the sentence you quote beginning "Is the rate..." seems clear in the context of scientific experimentation. Turing knows that the computer will not 'win' every time, so here he is setting a baseline for measuring the statistical significance of its performance. In fact, that may be the answer to your overall question.
    – sdenham
    May 2, 2017 at 12:03
  • @sdenham I would upvote an answer arguing along that lines. (Although I don't think it's the only answer to this question.)
    – user111
    May 2, 2017 at 18:39
  • @Hamlet As this would be an addendum to Gilles' answer, I would prefer it if she added it to that answer, assuming she agrees with what I wrote. If that does not happen, I will make the point in a separate answer.
    – sdenham
    May 2, 2017 at 19:02

Actually, Alan Turing had expressed his ideas about the Turing Test before without using gender...at least, according to Wikipedia. In 1948, Alan Turing wrote in an unpublished manuscript:

It is not difficult to devise a paper machine which will play a not very bad game of chess. Now get three men as subjects for the experiment. A, B and C. A and C are to be rather poor chess players, B is the operator who works the paper machine. ... Two rooms are used with some arrangement for communicating moves, and a game is played between C and either A or the paper machine. C may find it quite difficult to tell which he is playing.

That manuscript would later be published in 1968.

Due to this quote, I think Lauren Ipsum's claim that Turing was "talking about gender purely to make it easier to discuss A and B" is probably correct. For whatever reason, Turing didn't like using the chess example, and instead switched it with the party game example...but both paragraphs are essentially talking about the same "game". The point is that A (a computer) is playing against B (a non-computer) and is trying to trick C (the judge).

However, there is an alternative viewpoint about the Turing Test, called the "literal reading" by Gualtiero Piccinini in the article Turing’s Rules for the Imitation Game.

Some authors have read Turing’s passage in a ... literal way, suggesting that the goal of the machine is to simulate a man imitating a woman, while the interrogator – unaware of the real purpose of the test – is still attempting to determine which of the two players is the woman and which is the man. I will call this the literal reading. Supporters of the literal reading disagree over which of the machine’s capacities are being uncovered by Turing’s game. Some argue that his point is testing the machine’s ability to utilize language like a person; the blindness of the interrogator and the gender impersonation are introduced for methodological reasons – in order to make the test unbiased. Others suggest that Turing’s point was not to test the machine’s ability to utilize language like a human, but literally to test the machine’s competence at replicating the abilities of a human male whois attempting to imitate a human female.

In some footnotes, Gualtiero mentions a proponent of the "literal reading" that appears to grapple with the specific gender nature of the test:

Footnote 2: According to Genova, the literal reading accounts for Turing’s replacement proposal because Turing held the view that thinking is imitating; thus, a machine successful at imitating must be thinking (Genova, 1994, pp. 315–322). According to Genova, the request that the machine specifically simulate a human male imitating a human femaleis explained by what she takes to be Turing’s views on sexual identity, due to his own experience as a homosexual (ib., esp. pp. 314–315).

Footnote 7: Genova’s account in terms of Turing’s alleged view that thought is imitation is ... problematic. First, such a view is no reason to restrict a test for thought to the simulation of a human male imitating a human female, rather than allowing for a wider range of simulations. Second, and more importantly, Genova provides no textual evidence to warrant her attribution to Turing of the view that thought is imitation. In studying his work, I have found no evidence that Turing held such a view.

In his paper, Gualtiero makes several arguments against the "literal reading". Instead, he is in favor of the "standard reading", which is the one that is most commonly accepted in society.

According to the traditional interpretation of this passage, when a machine and a human being are playing the game, the goal of the interrogator is to discover which is the human being and which is the machine,while the goal of the machine is to be mistaken for a human being. I will refer to this as the standard reading.

His evidence in favor of the "standard reading" is the 1948 quote, other passages within the text which more naturally suggests an "standard reading" interpretation, and a lack of a good explanation for how the "literal reading" served Alan Turing's purpose of replacing the question "Can Machines Think?":

Any reading of Turing’s rules must explain how the imitation game fulfills his goal of replacing the question of whether machines can think. As I said, the standard reading’s account is straightforward: if a machine can demonstrate mastery of human language, knowledge, and inferential capacities to the point that it is mistaken for a human being, most people would consider it intelligent – or so they should according to Turing. With respect to this replacement goal, though, the literal reading generates more questions than answers. ... The proponents of the literal reading owe us not only an answer to these questions, but an explanation for why Turing did not address them at all.

He, however, does not fully dismiss the literal reading, indicating in a footnote:

These questions have actually been answered at length by Sterrett (2000), who argues that the test defined by the literal reading makes a better test for intelligence than the test defined by the standard reading. Of course, Sterrett does not attribute her arguments to Turing. The issue of what is the best test for machine intelligence is irrelevant to the topic of the present paper. Here, I concentrate on what Turing said, and didn’t say.

  • Comments are not for extended discussion; this conversation has been moved to chat.
    – user111
    May 2, 2017 at 4:05

Turing is talking about gender purely to make it easier to discuss A and B, I think. He was a gay man who didn't suffer fools gladly; gender wasn't a point of discrimination for him beyond what living in the early part of the 20th century would make any white cis male. He was interested in thinking, in puzzles, in how the machines worked, in creating a computer (literally a computational device, not "computer" as we think of the word) which could answer any question or problem you put to it without being specifically programmed for that sole question.

Andrew Hodges wrote a pretty exhaustive bio of Turing, Alan Turing: The Enigma, and there is no misogyny or gender bias highlighted that I recall. He was engaged for a time to Joan Clarke, another boffin, and ended things amicably over his homosexuality. But (as shown in the film The Imitation Game) he respected her mind greatly. There's no historical evidence that Turing would think a computer "only" had to clear the hurdle of "being as good as a mere woman" to fool a man. He was using contemporary parlance of "man" to mean sentient human.

  • 3
    I don't think this answers the question. My question is, in a paper about computers, why is the first thing that Turing talks about gender, and why is the category Turing testing for really the male-female binary instead of the human-computer binary. I find your analysis of authorial intent both simplistic and irrelevant to the question. It doesn't matter if Turing was a misogynist or not (not that being married and gay is evidence that someone isn't misogynist): I want to know why Turing is writing about gender in a paper about computers.
    – user111
    Apr 30, 2017 at 15:37
  • 3
    This answer is yet another example of why this site needs to move away from authorial intent, at least until people learn to distinguish between intent and meaning. I'm not asking about intent, I'm asking about meaning. Whether there is "no historical evidence that Turing would think a computer "only" had to clear the hurdle of "being as good as a mere woman" to fool a man." has nothing to do with what Turing actually said.
    – user111
    Apr 30, 2017 at 19:10
  • 5
    @Hamlet you did ask for drama. ;) I specified it as "as far as I remember" because it was a book I read six months ago, did not have in my pre-caffienated hand when I wrote the answer, and was not reading with the intent of ferreting out any evidence that Turing felt men and women had mental capacities which were significantly different. Apr 30, 2017 at 21:48
  • 4
    @Hamlet My answer is that, after reading the biography, I don't believe Turing thought that way. And I would have noticed such an attitude, because it would really have stuck out given his generally egalitarian attitudes about everything else, especially from a man living in the 1940s and 1950s. The lack of Turing commenting that men and women think differently is what was memorable to me. I'm sorry that I can't produce a quote from the book for you to support my answer. May 1, 2017 at 15:07
  • 5
    @Hamlet You say we have to move away from authorial intent, but your question is explicitly about authorial intent! Then you complain that Lauren Ipsum has misunderstood your intent as author of the question.
    – sdenham
    May 2, 2017 at 11:22

It is important to bear in mind that the Turing test is really an illustration rather than a attempt to prove anything. Bear in mind also that when this statement was made the idea that a computer could successfully imitate a human was a very distant possibility so he is very much talking about an overall concept rather than the nuances of its possible execution.

In this sense the ability to distinguish between a man and a woman is a simple and understandable way to get the overall context and scope of the task across in fact it is more about explaining the nature of the problem than having any specific bearing on the test itself.

I don't think it is a massive stretch to say that it would be possible to distinguish between a man and a woman purely by written communication, in many cases probably not easy but it would be reasonable to expect that an average person might get it right more often than not.

So what Turing is saying here is that to pass the test a computer would not have to be indistinguishable from any specific person but that it is no more different from one person than any other person ie if it is no easier to distinguish a computer from a woman as it a man from a woman then it has successfully passed for human. Whether it is the man or the woman replaced makes no difference.

Here the difference between a man and a woman is a relatable example of two people where its is plausible to be able to differentiate them but it is not trivial to explain exactly how you would do it. For example if you were trying to differentiate between a Lawyer and a Physicist there might well be trivial tricks with grammar and vocabulary that would make the task a lot easier by brute force analysis ie it is not a task which is trivial even with a large amount of computing power.

You could start taking about person A, person B and person C and then describe at length that they are quite similar but not identical but saying a man a woman and an observer makes for a much clearer example.

We also need to be clear what the Turing test actually says. It is simply the ability to fool a human being in conversation. In of itself, it is not directly about modeling human intelligence.


This answer rides on the coattails of Gilles and Tariq Ali, who have both provided informative and insightful answers, and addresses a follow-up question concerning the meaning of this paragraph from the paper:

We now ask the question, "What will happen when a machine takes the part of A in this game?" Will the interrogator decide wrongly as often when the game is played like this as he does when the game is played between a man and a woman? These questions replace our original, "Can machines think?"

Earlier, in setting up the thought experiment, Turing writes "It is A's object in the game to try and cause C to make the wrong identification." Now, Turing is establishing the criterion for evaluating the outcome, by setting up an experimental control, as is done in statistical tests, such as drug studies. He knows that the machine will not succeed every time, any more than will the human A, and here he is simply proposing that if the machine's performance over a number of tests is statistically in line with a human's performance, then that would be good grounds for regarding it to be intelligent. Turing is proposing a statistical measure as a better test of machine intelligence than "that seems intelligent to me" opinions.

Turing does assign gender roles in his formulation of the test, with the man being the subject of the test and the woman being given what is explicitly called a supporting role, but this does not seem out of line with then-prevailing mores regarding gender roles. Given those mores, it seems unlikely that anyone at the time would have misunderstood Turing's intent, even if that person objected to those mores. Pretty clearly (then as now), you can reverse the gender roles without changing the essential nature of the test. Furthermore, if Turing wanted to make a point about gender, surely he would have said "man" or "woman" throughout, rather than using the labels A and B as he does.

Attempting to read the paper in the light of current mores has a hint of Whig historiography about it - it will only obfuscate one's understanding of what Turing meant, and, as pointed out above, if you think his thesis is weakened by its use of gender, it is a trivial matter to reformulate it in a gender-neutral form. Tariq Ali has given an interesting collection of attempts to read the paper literally, and I think we should acknowledge Sterrett's point (with regard to her own preference for a literal version of the test) that such readings are not Turing's work; they are derivative works by their respective authors.

Note that later in the paper, while considering various objections to the notion of artificial intelligence, Turing discusses a simplified version of the game: "The game (with the player B omitted) is frequently used in practice under the name of viva voce..." This has become what is frequently meant when someone writes of the Turing test, primarily (I imagine) because it is simpler, rather than because it is (coincidentally, I think) gender-neutral. That simplicity comes at a cost, however: there is no experimental control or baseline to measure the machine's performance - being mistaken for a human every time would be a considerable achievement, but how would half the time rate? In this one-on-one version, there is no comparable test in which a human is attempting to pass herself off as something other than what she is. As Gilles has already explained, using gender as the test was an inspired choice on Turing's part.


I will try to approch this from a different angle.

One of your comments:

I don't think this answers the question. My question is, in a paper about computers, why is the first thing that Turing talks about gender, and why is the category Turing testing for really the male-female binary instead of the human-computer binary. .... I want to know why Turing is writing about gender in a paper about computers.

It's really quite simple, there was no such thing as a computer at the time. Not as he meant it, and not as you mean it.

Turing's idea was to create a machine that could think like a human. At the time this would have been considered pure and utter non-sense. A computer can add, or subtract, but not think. This is an era where the one of the most sophisticated "computers" around was a slide rule. To suggest that a slide rule could think was beyond silly. Yes there were were readers, yes, there were "mechanical" computers. But there was not, at that time, anything, even a little bit like what we have today. Not even a tiny a bit.

A computer, was a machine that would, when supplied with an input, produce an output. Cash registers were "computers". You mashed buttons, they turned wheels, gears and, dials, and gave you a total. That's just what there was.

So like explaining a concept a child, you start with something simple, then work into your complex example.

There are three people, A man, a woman, and an interrogator. The Interrogators job is to figure out which person is the man, and which is trying to pretend to be a man.

All ready, your brain ticks on, and your starting to think about ways you can do this. Questions you could ask to find out which is the man and which is the pretend man. You might even say a few questions out loud.

<---- Right here at this point, you have a pretty fun puzzle. It's not impossible for a woman to pose as a man (under the rules of the game). You have your audience thinking.

Now suppose if you will we replace the woman with a computer. One that was built to pretend to be a man. Surely it would need to be complex, but lets say we did it... and the paper continues.

The Man/Woman "game" is just a simple game. One that everyone can play. A 5 year old can play. Men and women are different everyone knows that. But now how could you tell one from the other in the constructs of the game.

Then jump to Computer/Man.

Well computers can't communicate, well lets say they can. Well umm....

Now your thinking.

If you just start with "So I have this slide rule that talks to me", it might be time for a padded room.

As to why he decided to do man/woman instead of some other groupings of two people we can only guess, but it seems that the reason is because it's the hardest to win. Age, race, height, all are really easy to come up with questions. Man/Woman, is much more difficult.


Unfortunately I have to object outright that it's only at the end of the article that he introduces computers. He makes the transition at the end of the first section, and there are seven sections; I didn't even have to scroll to see that, and the whole article is quite long.

The title of the article is "Computer Machinery and Intelligence," and the imitation game where you distinguish between male and female is given in the context of trying to show whether machines think. The point is, can a machine trick you into thinking it's a person any better than a man can trick you into thinking he's a woman?

How do we know our John Conner-looking Terminator will pass for a real person? Well, more people think it's human than people think our woman pretending to be John Conner is actually a man.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.