Where we are: To talk about how sci-fi AI attributes correlate, we first have to understand how their attributes are distributed. In the first distribution post, I presented the foundational distributions for sex and gender presentation across sci-fi AI. Today we’ll discuss embodiment.
Another simple measurement is how the AIs are embodied. That is, how to they manifest in the world of the story (or diegesis): Are they walking around, appearing as a screen on a wall, or as pulsing stars in the cosmos?
The categories that emerged from the survey were as follows:
Virtual, where a character only had, for example, a body or face that was generated for presentation to other characters on a screen or via volumetric projection. Joi from Blade Runner 2049 is virtual.
Disembodied, if the AI doesn’t have a particular, or an ad-hoc embodiment. The Machine from Person of Interest is disembodied.
Edgar from Electric Dreams is a Personal computer. In this regard, Edgar is a sui generis, or a category containing only one example.
Architectural: Some AIs are stuck to the walls of a building. HAL 9000 from 2001: A Space Odyssey is architectural.
Vehicular, where a character is embodied in a vehicle of some sort. K.I.T.T. from Knight Rider is vehicular.
Zoomorphic robot, where the robot is built to look something like an animal. Often these characters do not have voice. Muffit from the original Battlestar Galactica television series is an example.
Mechanical robot, where the robot is mechanical (and more mechanical looking than humanoid looking). WALL·E is mechanical.
Anthropomorphic robot, where the robot is proportioned like a human, and has most all the surface features of a human, but is readily identifiable as a robot. The Iron Giant is anthropomorphic.
Indistinguishable from human, where the robot can “pass” as a human. Only detailed or violent inspection will reveal it to be non-human. Aida from Agents of S.H.I.E.L.D. is indistinguishable from humans.
Here’s what that looks like in a bar chart.
Sometimes the details are tricksy
Sci-fi can make these things tricky. For example, the virtual crewmembers of the U.S.S. Callister might be considered indistinguishable from humans—as long as they are wearing clothes. Their unfortunate captain (and captor) had them created in virtual space such that they had no genitals. They are listed as bodily male and bodily female (rather than biologically) even though they are also indistinguishable from human.
Similarly, David from Prometheus has a fingerprint with a subtle Weyland-Yutani logo maker’s mark built into it (see the image below), but since this would only be apparent to someone who knew exactly where to look and for what, David is also listed as indistinguishable from human.
Why so human?
My conjecture to explain the high number of AIs that indistinguishable from human is threefold.
First, it is a matter of production convenience—that is, it is much easier and cheaper to insert a line of dialogue that establishes a character as a human-looking robot, rather than any of the other ways of signaling robotic-ness:
Create a costume like Robbie the Robot
Make a puppet like Teddy from A.I. Artificial Intelligence
Do prosthetic makeup like The Terminator
Create a set piece that syncs with audio like Alphy from Barbarella
Produce special effects, like Ava from Ex Machina
There’s also a fit-to-mediaargument which notes that people are much better and more comfortable at reading the emotional states of people than they are of machines. If catharsis, or the emotional journey, is part of what the art is about, humans work as a medium. (This lack of emotional information in interfaces was played to great effect in 2001: A Space Odyssey, unnerving us with the psychopathy of HAL’s unblinking eye.) Actors, too (I highly suspect) enjoy using their bodies, voices, and faces to do their jobs without the additional layers of prosthetics or puppetry. So we would expect an overweighting of indistinguishable from humans because they are often the best tools for the narrative job, from both the audience’s and the actor’s perspective.
There’s another argument—a genre-and-narrative argument—that people are mostly interested in stories about people, and most sci-fi is a speculation about social effects rather than actual technology, and so indistinguishable robots are the best embodiment of what we’re interested in, anyway. Humans, just with different rules.
In the first post of this series, I explained what I was out to learn, what I looked at, and how I tagged it. Ultimately, we want to look at the data and be able to answer questions like “Are female AIs more subservient than male AIs?” And in order to do that, we first have to understand what the distributions are for sex and subservience. So let’s talk distributions.
Distribution is a fancy term for how many of each value we see for a given attribute. For example, if we wanted to look at the distribution of eye color across the world, we would count how many browns, blues, hazels, ambers, green, gray, and reds that we see, (finding a way to deal with heterochromia, etc.) and compare them in a bar chart.
Of course eye color is not of interest in this case. For Gendered AI, we are interested in comparing other attributes to gender presentation. We’ll look at the other attributes in later posts, but we’re going to begin with sex ratio, and that will fill up a post all its own.
Simple sex ratio
Author’s request: With that section title I know some hackles are already raised. Please know this is very tough space to write for. Despite having paid for a number of paid content reviews, I may have made some missteps. I am a n00b writer on these topics, and I respond best to friendly engagement rather than a digital pillory.
The very simple explanation of sex ratio is women-to-men. But of course that’s waaaaay too simple for either the real world or our purposes. At the very (very) least, AI might have no gender, so we need a “none” or “other” category. Let’s start with these very oversimplified numbers and move to more detailed later.
The chart shown below shows the data from the survey focusing on simple categories of female, other, and male. The chart shows that AI characters are strongly overweighted male, with a rough ratio of 2 male : 1 female : 0.75 other. The 2:1 M:F ratio is eerily in line with USC Signal Analysis and Interpretation Laboratory’s finding where speaking roles in 1000 scripts they studied, men’s dialogue, and even the number of characters was double (or over) that for women. This is greatly different than the real-world sex ratios of 1:1 as reported in the Wikipedia article about world sex ratios.
I would talk about the weird discrepancies of just this distribution, but any ranting at this point would be overshadowed by the ranting that happens next. Deep breath.
Having an “other” category isn’t enough. After all, characters in one of these bars can be as different as HAL and Gigolo Joe, and that doesn’t seem right. So, let’s break this oversimplification down into more refined bits.
More detailed gender presentation ratios
First, of course, we should note that characters rarely discuss gender directly, and—at least in this sample—discuss gender dysphoria all of never. Also we can’t reach out to ask any of them directly since they’re fictional. So when I speak of gender, it should be read as “gender presentation,” and unfortunately at this point you are stuck with nothing more scientific than my reading of the following four variables.
Primary sex characteristics, or biological presentation: The presence of masculine or feminine sexual organs. None of the titles I reviewed were pornographic, and full-frontal nudity is pretty rare up until Westworld, so this often comes down to implication. Gigolo Joe, for instance, could not do what must be a key part of his primary function without male sex organs (with all the important caveats that penetrative sex is just one kind of sex), so he is listed as “Masculine” here.
Secondary sex characteristics, or body presentation: These are much more directly observable, and include those other markers of sex, like facial hair and shoulder-to-hip ratio.
Voice presentation: This is my hearing of whether the voice has a lower, masculine register, or a higher, feminine register. (In a few cases I checked on the actor listing in IMDB and did web searches for evidence of self-identification.)
Pronoun presentation: How other characters refer to the AI character with pronouns. R2D2, for instance, has absolutely no sex characteristics, and no voice, but is still referred to as a “he” throughout the Star Wars franchise.
A note on labeling: I’m aware that there are tricky nuances in the labels. After all, how is body not part of one’s biology? But the shorthand proves useful so we can use the shorthand “BIO” and know what it means instead of always having to use the longer phrase “implicit or explicit primary sex characteristics.”
For each AI character, I tagged each of these variables as either Masculine, Fluid, Neutral, Feminine, Unknown, Multiple, Many, or N/A. (The “n/a” may seem weird, but for instance, HAL doesn’t have a body, so primary and secondary sex characteristics are not applicable.)
Combining voice and pronouns into “social”
There are plenty of characters with no voice or non-human voices, and a few characters that are not referred to by pronoun. Since these two indicate a social performance of gender, I treated them in the algorithms as an “OR” when considering stacking. That means if either variable was present, and they didn’t contradict, I counted it the presenting aspect. Compare these two examples…
R2D2: N/A Primary, N/A Secondary, neutral voice, male pronoun = alsosocially male
The main thing to note about how these three variables (counting voice and pronouns as “socially”) played out is that they overwhelmingly stacked. That’s not a term of art, so let me explain. It means that if a character has masculine primary sex characteristics, that invariably meant that he also had masculine secondary sex characteristics, and voice/pronouns. If a character had no evidence of primary sex characteristics, but had feminine secondary sex characteristics, she invariably had feminine voice/pronouns.
It makes more sense if I show you. So, here are six representative examples from the survey of how this monosex stacking looks.
I suspect this is an effect of binary concepts of gender on the part of the markers of the sci-fi, implemented as increasingly detailed costumes for the AI. But when you consider these variables, these 6 are a pale semblance of what could be. Include “fluid” or “nonbinary” as a possibility, and don’t bother with stacking, and there are 58 more possible combinations of these variables.
Hey, want to feel both hyper-reductive and overwhelmed at the complexity of gender? Try writing a categorization algorithm for analysis.
Anyway, if they hadn’t stacked like they did, I would have had to describe their genders with a four-part-code that would result in 64 genders. But, because they do stack, that meant there were these 6, plus “multiple,” “genderfluid,” “neutral,” and “none,” for a total 9. Note that online lists of genders vary from the 58 available to Facebook’s users to the 229 found on this more creative list (my favorite is “Schrodigender – A gender which you can both feel and not feel” giving a clue to how serious that particular list is.) So while 9 can feel heavy, it does not compare to the complexity of the real world.
OK, given those descriptions of the subcategories, here’s how the numbers played out in the much more detailed analysis of gender presentation in sci-fi AI.
Detailed gender presentation
I’ve noted that we’re here for the correlations, not distributions, here, but in and of itself, this is remarkable. The subcategories provide a deeper (and more troubling) look into the data, and is necessary because these categories have to be thought of differently. Observe, for example, that the biologically-gendered characters are nearly at parity, while the bodily- and socially-gendered characters skew male. There is a frustrating 2:1 ratio for bodily male:bodily female and an infuriating 5:1 for socially male:socially female.
These ratios bear…discussion.
1 biologically male : 1 biologically female
A harsh interpretation of this stat would read a kind of heterosexual panic, where—when sex or procreation is involved—Hollywood needs to assert loudly over a hastily-ordered beer that whoa whoa whoa: Only AI chicks and AI dudes get it on. Or if they do get it on with people it’s with the right gender.
Or, more charitably I suppose, humans are largely heterosexual, and since there is a rough 1:1 sex ratio in humans, there should be a 1:1 sex ratio in them. (?) It’s a hard thing to second-guess.
It gets darker in the other categories where the sci-fi AI has a body but no biological apparatus. The ratios still skew heavily male. As if, when it comes to just being a person, a total sausagefest is the norm.
2 bodily male : 1 bodily female
Recall from above that this category is reserved for those AI characters that present a gendered body but do not have gendered reproductive or sexual capabilities. We will discuss the germane-ness and embodiment of these AIs in a later post, but for now we can note that this category of AI character, with its 2:1 ratio is roughly in the middle between the biologically and socially gendered categories, and in-line with the oversimplified distribution seen above.
5 socially male : 1 socially female
This is the category where the only markers of gender are voice and pronouns. In other words, characters for whom a gender seems like an arbitrary choice. WTF is up with a 5:1 ratio? Why are all these “arbitrarily” gendered AI characters guys? We’ll talk about germaneness to the story later, but I want to see if there is some extradiegetic reason first.
Is it the available voice talent?
We have to acknowledge that filmmakers must hire someone to voice their speaking AI characters, even if there are no other markers. Despite the fact that…
…it’s fair to say that most available voice talent is recognizably gendered, and the AI character may just inherit the presentation of its actor. Then you might expect the roles to match the sex ratios in the available talent pool. I couldn’t find any formal studies of this, so I created a throwaway account on voice.com—a major job site for voice actors—and performed separate searches for male and female talent. There I found 42,786 males, and 24,347 female non-union voice actors, around 2:1. (Union actors were closer to 1:1, with 3,079 male and 2,336 female. n.b. The site gives only those two gender options in its search.) Though that’s more anecdotal than I’d like, even the worse ratio of 2:1 still pales compared the 5:1 of socially gendered AI, so no, that’s not it. You might think that explains the “simply” gendered characters, but my suspicion is that the genders of the characters are set in the script and pass down through the process, unquestioned after that.
Is it what sci-fi audiences want?
Might the ratio be some sales rationale, some presumption that sci-fi audiences are mostly men and therefore might only be more interested in male characters? No, of course numbers vary by show and genre, but this article by Victoria McNally shows that there is only a slight majority of men in these audiences (hovering around 60% male and 40% female, rather than 73% male and 17% female, which the 5:1 socially gendered ratio would have you believe.)
Plus the 2018 annual Hollywood Diversity Report by UCLA shows that “new evidence from 2015–16 suggests that America’s increasingly diverse audiences prefer diverse film and television content,” so we would have to greatly exaggerate the connection between the sex ratio of the audience and those we see here.
There has to be some other reason, and I suspect it’s the dark patriarchal notion that “male” is somehow the default gender. Even though it is, literally, not.
Is it that Hollywood itself is mostly white and male?
The 2018 Hollywood Diversity Report shows that gatekeepers, writers, directors, and (points at self) critics are still overwhelmingly white and male. White male writers and directors account for 91.9% and 86.2% if their fields, respectively. This is closer to the 73% male, but still a crappy, crappy excuse for the default assignment of AI as male. Representation matters and this is sorry representation.
P.S. Don’t get uppity, real world
The Global Gender Gap Report issued on 17 DEC 2018 by the World Economic Forum showed (in collaboration with LinkedIn) that women only occupy 22% of jobs in AI professions. (See page viii, 28–35 of that report.)
You probably had a general sense of this disparity from simply being an audience member. But it’s “nice” to have some data to back it up. Be forewarned: It gets worse when we look at correlations. (No, really.) But before we do that, we should look at the rest of the distributions, starting with embodiment in the next post.
Men are machines. Women are bodies. Male is extreme. Women are nuance. General AI has gender. Other AI does not. Male is free-will. Machine is subservience. Male is default. Women when it’s necessary.
At least in screen sci-fi.
Let me explain.
In November of 2018, a tweet thread between Chris Geison and Kathy Baxter called my attention to questions about the gender of AI in sci-fi. Baxter noted that most AI is male, and how female AI is often quite subservient or sexualized. In this thread, Gieson added Cathy Pearl’s observation that embodied AI is often female and male is more often disembodied and regarded as a peer.
I already had a “database” (read: Google Sheet) of AI in screen sci-fi from Untold AI, my 2018 study of the stories screen sci-fi doesn’t tell, but should. So, I thought I could provide some formal analysis to this Gendered AI discussion. To that end I’ve added around 325 AI characters to the Google Sheet, and run some analyses. This series of posts will break it all down for you.
Now, it can get a little dry to talk about percentages and comparisons and distributions, so I’m going to do my best to keep tying things back to the shows and the characters and the upshot of all this analysis. But the way we get to that upshot is through the numbers, so stick with me. For this first post, I’m going to share what I captured, and what counts as an AI character for purposes of this study.
327 AI characters from science fiction (see the full list in the live sheet)
Movies and television shows from 1927 (Metropolis)–2018 (Upgrade)
Call to action: Of course I missed some movies and TV shows. Add them in the comments, including a link to their IMDB page.
The survey that drives this site has always focused on screen sci-fi for its ability to depict interfaces that can be reviewed. Literature is much more free to experiment with ideas than screen sci-fi, and so will have lots of additional examples, but won’t appear in the survey.
Each character is tagged multiple ways. More detail on particular attributes below.
Movie or Show Title and Episode if appropriate
Gender Presentation (which is a roll-up of four separately tracked variables)
Appearance or evidence of primary sex characteristics
Appearance of secondary sex characteristics
Pronouns used by other characters
Subservience to humans
Germane-ness of gender (more on this in its own section)
If not free-willed, the gender of the master
Category of AI (Narrow, General, or Super)
Whether their gender presentation changes over time
Genesis, or how the AI came to be. This is mostly used to distinguish AI that are copies of humans (whose gender would thereby be inherited).
Call to action: If you think there’s some critical attribute that I’m missing, pipe up in the comments. I can’t promise I can get to it before the next post, but I can consider it as a future enhancement.
Yes, but which Skynet?
With the exception of the flag marking changed genders, when characters change other attributes over the course of their stories, they are tagged for their final state. For example, the Maschinenmensch from Metropolis begins an anthropomorphic robot, but after Rotwang transfers Maria’s likeness to it, it becomes indistinguishable from human, and so is tagged as such.
If you’re looking at the Sheets data, you’ll see that text values have corresponding numerical columns to allow for easy sorting and graphing data, but I tried to gray them so they don’t distract from a reading of the raw data.
Full disclosure: Possible problems with this data
Sci-fi is a vast supergenre. There are certainly examples missing from the survey, so it should not be regarded as exhaustive. (I tried to get as many as I could.)
I generally target well-known examples rather than limited-release or student projects.
The sci-fi interfaces blog usually eschews comedy that breaks the 4th wall routinely, (e.g. Spaceballs), as this makes for very complicated analysis, and so the survey will be missing these examples as well.
I only speak English fluently, and so have only reviewed shows in English, with English dubbing, or with English subtitling.
I am not a data scientist. I’m a smart guy who tried his best, but may have made some errors in the formulas.
I am not an expert in gender issues. I may make unintentional errors in discussing or categorizing genders, use insensitive language, or have naive errors in my thinking. I have engaged a professional sensitivity review, but of course they might not catch everything, either.
I am a progressive, liberal, (imperfect, see above) feminist. Though I tried not to, my bias may have colored how I coded the examples and of course the interpretation of this data.
I have to go on a LOT of common-case presumptions. For example, men can have breasts for many reasons, but I used the presence of breasts as one marker of female-ness. I suspect this is a disservice to the real complexity of gender and sex in the world, but presuming the audience sees gender as primarily binary, it marks how these characters are likely perceived rather than what they are.
I’m not too worried about these caveats, though, since what we’re aiming for here isn’t precision engineering specs, but rather to get a numbers-based sense of the big patterns in screen sci-fi, and for that, a little bit of noise in the numbers is OK.
Lastly, not every character that you think might qualify does, so I should explain my rationale for what got in and what got left out.
What counts as an AI character?
I’ve tried to be strict about what counts as AI in that the intelligence of the character must be housed in non-biological circuitry. This leaves out some characters that on a cursory consideration would seem like a natural fit. For an example, compare The Stepford Wives (1975) and The Stepford Wives (2004). The wives in the original were robots through and through—mechanical, lookalike replacements of the original humans. But the wives in the remake were cyborgs, with robotic bodies housing their original, human brains. This means that in the original, the wives count as AI and appear in the survey. But because of this cyborg technicality, none of the “robotic” characters from the remake make it in. Not even the little cyborg dog.
Meanwhile, Rachel and Deckard, replicants from the Blade Runner universe, had a baby (according to Blade Runner 2049) so we can generalize and say replicants are capable of wholly biological reproductive acts. Given this you might think they’re out of the survey, but, since they are fabricated, they get into the survey.
Also, T-800s Terminators (the Arnold kind) get in, because even with their wetware bodies, the intelligence they carry is non-biological.
I know, it’s complex and sometimes counter-intuitive. Such is data.
OK, so looking at those attributes for those characters, the first thing we should look at is the distributions. This included all sorts of questions like: How many AI present as men? How many as women? How many are nonbinary? What kinds of bodies do they have? Who is master of whom?
It’s thrilling, thrilling data analysis action, so stay tuned.
Now we come to the end of Idiocracy, if not yet the idiocracy.
This film never got broad release. There are stories about its being supressed by the studio because of the way the film treated brands.
But whatever the reason, I’m happy to do my part in helping it get more awareness. Because despite its expositive principle being wrong (and maybe slightly eugenic), the film illustrates frustrations I also have with some of the world’s stupider ills, and does so in funny ways. Also, as I noted in the last writeup, it even illustrates speculative and far-reaching issues with superintelligence. So, it’s smarter than it looks.
I’d recommend lots and lots more people see this, generally, if only to reinforce the demonization of idiocy and make more people want to be not that. So first let me say: If you haven’t yet, see the film. Help others see it. Make People Valorize Enlightenment Again.