AI Child Porn Will Not Protect Children
Contra Aella's insane take
Recently,
, who needs no introduction, had a take:
Let’s take it at face value.
She wants to protect kids.
She believes that some small share of people just have intrinsic sexual interests in children.
She believes sexual interests can be satiated without causing escalation.
Thus, AI child porn can provide satiation without harming actual kids.
She musters some evidence for her view. We will take it in turn.
My view, in turn is this:
I also want to protect kids.
I believe sexual interests are highly malleable and usually not well-described as intrinsic, and furthermore that sexual interests in children are plausibly very common or could easily become very common.
And furthermore, I do not accept the satiationist model of sexual behavior.
Thus, AI child porn will lead to increased harm of children.
Let’s dive in.
Has porn reduced rape? (No, it has probably increased it)
, like many pornified people, really wants to believe porn satiates a need and thus reduces rape. The research basis underlying the satiation hypothesis is very dubious: low study quality, and an alarmingly high retraction rate. I know of one researcher who is a prominent proponent of this, whose papers have been highly cited and gotten huge media attention (e.g. he claimed when NY opened more strip clubs rape fell in those areas)… who has had at least 2 and I think maybe 3 of his prominent papers now identified as clear errors likely to be retracted. A lot of researchers in this field are just making stuff up. Why? Because sex data sucks! Data on sex is really bad an unreliable! This is a notorious problem, and so the degree of confidence we can have on any sex-related hypothesis is just intrinsically quite low.
But with porn we have a pretty easy way to test effects. Why?
Because porn use has gone up an insane amount! Depending on the survey you use, maybe 80% of men under 45 recently used porn, maybe as little as 35%, but everybody agrees it’s gone way up. Here’s GSS estimates using the very-flawed GSS variables for porn viewing:
The 1980s values are wonky due to some question design stuff, but plausibly porn use has risen WAY above its 1972-2000 norms over the last 10-20 years.
If pornography has a satiating effect, then people-with-rapey-desires should sort into porn use. This means that as porn use rises, rape rates should fall. This is what the best satiation studies purport to show: that access to porn reduces sexually violent behavior.
But it’s pretty easy to test this when a variable has gone from 10-20% of the population to 30-80% of the population! That’s an enormous shift! If porn satiates rape, then porn rates should have fallen a ton vs. other crime rates!
The “vs other crime rates” matters because we also have to recognize all crimes are correlated. Bank robbers are more likely to be rapists than non-bank-robbers, etc. Virtually all social pathologies are highly intercorrelated. As such, it could be that rape falls or rises just because of other changes in determinants of crime that impacted all crimes equally. But porn should change rape much more than it changes other crimes. We would expect porn access to have a bigger impact on sexual assault than it has on the share of people driving 80 in a 20 mph zone, because porn seems a lot more credibly and theoretically related to possible determinants of sexual assault.
In other words, if porn reduces rape, then rape rates should have fallen quite a bit more than other crime rates over the last few decades.
Here’s changes in crime rates, chained to 2000, according to FBI crime data. Note that the FBI has changed the reporting definition of rape a couple times, I’m using a best-guess at a harmonized estimate.
Rape and homicide rates 2000-2016 appear to be extremely correlated. Rape rates rose more in the late 2010s, then fell recently when homicide rates skyrocketed. The anomalous rise in murder rates is well-known to have been caused by idiosyncratic effects in 2020: some mixture of George Floyd-related incidents, depolicing and anti-police sentiment, COVID lockdowns, etc. There’s no reason to believe that post-George Floyd-type-effects would suddenly cause a rape epidemic in the way they might cause arson, robbery, or murder.
Notably, the crash in rape rates supports the idea that lockdowns mattered for crime, since lockdowns obviously reduced opportunity for rapes! But I also want to note that if you look at 1990, you can see that from 1991-2000, the decline in rape was waaaaaay smaller than the decline in homicides. If we made our benchmark any time before 1995 instead of 2000, we’d find that rape has not declined nearly as much as homicide.
I submit that this is evidence against porn having any large satiating effect: rape rates at best tracked with general other rates in criminality, and plausibly fell quite a bit less if you date the rise in porn to the 1990s.
But FBI crime data is biased by issues in reporting and definitions. Let’s use the National Crime Victimization Survey to look at the share of people who report being victimized by a crime! This is a much more credibly stable variable. Here we are gonna take the population-estimate of how many people experienced a specific type of criminal victimization, and divide by the number of men ages 15-40. Why that divisor? Simple, because men 15-40 are the overwhelmingly vast share of the perpetrators of these crimes, so this is a good proxy way to ask “How criminal were the young men?”
You can see that robbery, aggravated assault, and simple assault are all down, and down by a lot. Young men are arguably much less criminal in the 2020s than they were circa 2000!
But look at rape and sexual assault!
It’s way up, and the growth in it clearly stretches back to the 2000s, when expanding internet access plausibly led to a major increase in porn usage!
On the whole, the NCVS data, which is more credible than FBI-reported rapes, suggests that while crime generally has fallen, sexual crime is actually getting more common even as we have seen a huge increase in porn usage.
The small lab studies we have are weak because it’s hard to find a control group: virtually all men either use porn, or have nonrandom reasons for not using porn like being religious which also correlate with lower crime. But because porn use has become so ubiquitous, we can practically consider the entire population to be “treated.” Thus, population-scale statistics are reasonable inferential checks. If the satiation hypothesis is true, rape rates should be falling, but actually they are rising even while other crimes are falling. This is reasonably suggestive evidence against satiation, and may even suggest porn causes rape. The study
cites claiming porn reduces rape was published in 2006, with data ending in 2003, before the truly massive increase in porn use during the 2000s: it is irrelevant. Thus, while all our data is bad, a reasonable prior for this debate, a stylized fact you should carry in your head, is, “Over the last few decades, as porn usage has exploded, rape rates have not fallen nearly as much as rates of other crimes, and recently may even be rising.”Will people emulate child porn? (Yes, duh, a million times duh)
consider:
She has fun survey data on this. I like the survey data. Her theory is plausible: porn makes people willing to try stuff they’re already sneakily interested in, but maybe turns them off from stuff they see as too taboo. This leads her to this:
She regards this as implausible. Most people aren’t like that, she says.
But obviously, tons of people are like that! Huge numbers of men are masturbating to CSAM-adjacent content! “Barely legal” and “nubile” porn are hugely popular categories according to data published by porn sites I prefer not to link to. Porn categories that revolve around youthfulness-fetish are ubiquitous. There are strong evolutionary psychological reasons to believe that men in particular are likely to have a non-trivially-frequent “glitch” in their sexual programming which causes their built-in preference for “young and fertile” to frequently “malfunction” as just “young” since “fertile” isn’t easy to observe in the state of nature (humans do not as obviously go into “heat” like other mammals) but “young” is relatively observable. The genuine tragedy here is that there actually are a huge number of men who, with a little push, will be pedophiles. In the state of nature, the evolutionary costs for erroneously identifying a woman as “too young” are low, since if she’s too young the male can just wait a bit. As a result, male sexual hardware has a high intrinsic likelihood of a specific kind of malfunction: excessive youth preference. Why do you think men’s fashion is so stable compared to women’s fashion? Because a way women signal youth is by adopting idiosyncratic fashions typical of their cohort; older women emulating younger fashions are described as “keeping up” and other metaphors that clearly refer to the idea that fashion is to a nontrivial degree about women signaling age cohort.
How else might we know that male sexual hardware can easily be culturally adapted to have widespread preference for pedophilia?
Because of history, my God! Are we ignoring bacha bazis? Are we pretending like Ottoman and Greek pederasty didn’t exist? The sexual abuse of children is so historically common and widespread it is difficult to fathom how
can be so massively ignorant of it! Sex with children is practically a historical universal before Christianity! We know from textual records and from historic mortality patterns that Romans started marrying off girls and trying to impregnate them at 12, often pre-pubescently (they didn’t, uh, quite get how stuff works).You cannot have more than a high schooler’s understanding of history and think that pedophilia is an uncommon or rare fetish. It is extremely common. In many ancient societies it’s not unreasonable to think a majority of men were pedophiles, and not just “married girls at 16” pedophiles, but like, “They rode romantic love poems about the sweetness of the anuses of 9 year old boys” pedophiles. Out of respect for decency I will not proceed to here provide extensive evidence, and I don’t necessarily encouraging looking this stuff up, but
’s entire theory here rests on one massive loadbearing assumption: most of her data, most of her sexual partners, even herself, are intrinsically products of Christian moral rearing which basically said that even if somebody did sexually abuse a child, they should feel guilty about it and stop doing it. The idea that sex with children is extremely bad is not culturally universal, indeed fairly culturally rare, and as Christianity’s moral influence fades, that norm will fade too.It’s not just history: we have studies of current societies that often show crazy high rates of child sexual abuse! Here’s the latest meta-analytic data I found:
I would not take the specific country estimates to be gospel; other studies often find rather different rates. For example, if you believe Cambodia has lower rates of child sexual exploitation than Switzerland or Sweden…. here I am smashing D to Doubt.
But the exact values don’t matter: what matters is that adults who want to have sex with children are just not that rare.
’s view that sexual interest in children is an uncommon, deep fetish which is hard to trigger is just clearly false. In fact, sexual interest in children probably is not that rare, and probably not that hard to socially nudge, particularly re: older men and underage girls.This connects as well to
’s other questions about “Will AI child porn normalize child sex?” or “Will it draw in new markets?”She has a mental model were sex with kids is a rare and extreme fetish which would be fairly difficult to normalize. Yet the historic and contemporary record suggests that sex with children is a rather common sexual practice and quite easy to normalize. That child sexual abuse is so common even in societies that deeply stigmatize it also suggests that this sexual interest is actually very common: something like 3-12% of children will be sexually abused! This isn’t some crazy rare edge case! A lot of people, given the chance, will abuse children!
Do people emulate other porn? (It’s complicated, but probably yes)
Okay now, is it generally true people emulate porn? Maybe I’ve convinced you the satiation model of porn and sexual crimes is a weak hypothesis, and maybe I’ve convinced you that there’s actually a lot of people who could be nudged into pedophilia without too hard a push, but is porn actually that push? Do we actually see changes in sexual behavior mirroring porn?
This is a complicated question, honestly more complicated than I thought. I thought for sure that if I used NSFG data to measure changes in male→female penetrative anal sex, for example, that it would show a big increase over time! And it does! But not when I expected it to? Of women who have ever had male-female sex, here’s the share who have had anal sex:
So you can see there was a big increase in male→female anal sex between 1992 and 2002. The “naughty nineties” were coextensive with just a massive increase in a sexual practice most women don’t find very appealing (more on that below). This increase occurred across cohorts. And for what it’s worth, values further back were probably even lower: Kinsey’s surveys suggest 11% of married couples in the late 1940s/early 1950s may have ever had anal sex together. If we imagine Kinsey’s couples as being 35-44, you can see that between the 1950s and 1990s, anal sex prevalence plausibly doubled.
So the rise of anal sex happened sometime between the 1950s and the 1990s. I think a plausible period is the 1980s, and specifically the rise of the VCR. Here’s Google Ngrams on anal sex:
Almost nobody talked about it before the 1970s, and it really took off in the 1980s. Grok also insists to me that the first anal sex porn scene for video involving a human was in 1970 (there was a dog involved in a 1969 movie), and that anal sex became a popular pornographic style in the late 1970s and early 1980s. That all tracks. It seems clear that porn innovated and created a norm that spread rapidly, and behavior responded.
Anal sex prevalence actually seems to be falling. And note that this is among women having heterosexual sex at all, so this isn’t due to the general decline in sex or shift towards non-heterosexual sex. A recent article laid out evidence that the VCR was the first true revolution in home pornographic videos. The rising prevalence of homosexuality probably also contributed, with I imagine straight people wondering what all the hubbub was about.
So it seems to me what happened is, in the 1980s most likely a revolution in porn created a new sexual norm: some men feel comfortable asking for anal sex from their female partners. This norm took time to percolate; it did not reach a peak until the early 2000s. Since that time, younger generations seem to be turning away from anal sex, at least among women who are having any heterosexual sex at all. It’s possible young men still expect anal sex, but that anal-sex-expectations make men bad sexual expectations, and so women reject men they suspect of desiring anal sex. I find this story extremely plausible for reasons we’ll come to.
Now ask yourself how likely it is that historic anal sex was extremely widespread. Does it serve a clear evolutionary purpose? No. Does it create massive pathogen exposure? Yes. So how likely is it that human reproduction over the last few hundred thousand years was selecting for something that, as soon as people started doing a lot of it, they all started dying of AIDS and getting HPV and anal cancers? I would say super unlikely.
Which… isn’t to say nobody did it! Obviously, some people really like anal sex! But episodes of cultural interest in anal sex probably don’t persist.
We know of one such historic episode! Romans were obsessed with anal sex. Depending on which corpus of historic pornographic data you use and your guess about which hole is being used in rear-entry depictions where exact anatomy is not discernible, somewhere between 25 and 60% of all Roman pornography depicts anal sex. There is a pretty strong case to be made that a nontrivial part of the appeal of Christianity for women was the generalized Christian prohibition on all anal sex. Yes, dear reader, ancient Christianity did in fact ban anal sex even between consenting married couples. St. Augustine most explicitly addresses the issue, but there’s a pretty widespread understanding that early Christians really did do their best to discourage men from anally penetrating their wives. The Cathar heresy may or may not have actually engaged in widespread anal sex, but the medieval Catholic church saw fit to accuse it of endorsing male-female anal sex as a way of showing how disgusting it was. Was this accusation true? Historians debate whether “Catharism” existed at all as a real movement! But the accusation reveals the norm.
So, because anal sex may have various reasons some people enjoy it (particularly males, on either end, evidently), the practice of anal sex is likely to emerge every so often. But because other people tend not to enjoy it (women) and because it has various maladaptive features (absolutely gobsmacking rates of pathogenicity), anal sex is an unstable norm.
In my view what we’re seeing right now is that the anal sex norm emerged, but unlike Rome where women had few means to fight back (and desire to avoid childbearing may have led some women to tolerate the behavior), modern feminism and contraception are enabling women to fight back faster. There may still be a male expectation in many social circles (hence the stories about young men demanding anal sex), but in practice women have the ability to gatekeep the behavior. The norm is unstable and it’s unlikely to persist.
How do we know women tend to dislike anal sex? Here’s
’s taboo chart:
I’ve circled anal sex. You can see that anal sex is overwhelmingly male-preferred. Also look at the bottom right and notice that pedophilia is also very male preferred. So is incest. In general it actually looks to me like the degree of male-preferred-ness may have a nonrandom relationship with the degree of taboo-ness, but I’m not totally sure.
Do we have other data?
Well, the 1992 NSHLS survey I used above asked men and women about a bunch of sex stuff! The NSHLS was massively comprehensive and it’s a huge shame it hasn’t been replicated. Here’s their findings for a range of sexual activities:
You can see the most massively sex-biased interest is group sex. In
’s data that’s only modestly male-biased. You can also see that men are just hornier for basically everything. Here’s just the ratios:
So the most male-biased interests are:
Sex with multiple people
Sex with a stranger
Watching others have sex
Stimulating partner anally
Being forced to have sex
Receiving anal intercourse
We can also compare women’s interests in receiving anal sex acts to men’s interests in giving, and also look at oral.
So we can see that on oral sex, men are about as interested in giving as women are in receiving. Flipped the other way, women are indeed less interested in giving than men are in receiving, but that’s mostly just the generalized horniness gap at work.
When it comes to manual stimulation, men are about 40% more interested in giving than women are in receiving, and women are only about 50% as interested in giving as men are in receiving. In general, men are interested in doing something women aren’t super keen on, and women are hesitant to do something many men are apparently interested in.
Finally, anal intercourse. The 1992 NSHLS did not ask women about giving, because they don’t have penises. Regardless, men are 2.5 times as likely to find giving anal intercourse appealing as women are to find receiving it appealing. There is a huge gender gap here.
So, on the whole, we can say that even though very few women find receptive anal sex appealing, huge shares of women end up in that sexual position, because men want it and perhaps demand it. This rise has occurred since the 1950s, and in fact was ongoing in the 1990s. Most plausibly, the rise in anal sex occurred in the 1980s and 1990s, coextensive with the rise in home pornographic videos. This period saw a huge rise in general social indicators of interest in anal sex, and pornography is a plausible candidate.
But because anal sex has significant countervailing currents against it, prevalence may now be falling as women seem to be declining to provide for male interest in it (disclosure: this seems reasonable to me, I think it’s actually insane and gross that so many women have been pressured into anal sex, and I think stigmatizing this practice is a good thing).
How would pedophilia look? Well, we already know other societies found ways to make pedophilia pretty common, so some kind of latent interest-capacity must be there. We know anal sex rose despite widespread opposition to it by half the population. We know this rise occurred coextensively with the rise in porn, which itself was coextensive with huge general growth in public anal sex interest. Anal sex is declining because women have the ability to advocate for their interests and because anal sex has strong maladaptivities related to pathogenicity.
Child sex doesn’t have these features. Sex with children does not expose the pedophile to higher rates of illness and disease, and sexual interest in children is a plausibly common malfunction of an otherwise evolutionarily adaptive youth-preference among men. Moreover, children are unable to advocate on their own behalf in a politically consequential way. As such, a flood of AI child porn could be expected to lead to more child sex, not less. This is an extremely bad thing, and so we should discourage AI child porn production.
Do people want the real thing? (there’s no way for Aella to know)
says:
Her argument here is just an extreme version of the satiation hypothesis. But all
knows about is the experience of commercial sex work. This is informative experience. But crucially, she doesn’t know if the John has a kid locked up in the next room. She knows what gets him using her & her colleagues’ services. Her inference here is overwhelmingly based on her clients. But people willing to settle for the fake thing are obviously not a valid sample for assessing whether or not people in general are willing to settle for the fake thing. I don’t know for sure if child porn viewing always escalates into child sexual assault, or if AI porn would have the same effects, but I know for a fact that asking the most enthusiastic customers of porn if porn is enough for them is not a reasonable way to assess how commonly porn isn’t enough!Will norms change? (yes)
says:
So good, she concedes the core point that norms are a major defense and widespread child porn could in theory erode those norms.
But then she says social shame around pedophilia has increased. Uhhhh…. has it? Really? What’s the measure there? I don’t think we have any evidence of that. Rising CSAM access over the past decades is a pure technology shock (internet), totally orthogonal to the normative debate.
The actual fact is that if we generate an Infinite Ocean of AI Child Porn… it will spill out of its bounds. The reason CSAM is pricey is because it’s hard to get. But because it’s hard to get, not a ton of people just have a whoopsie and find 50,000 images of child porn on their hard drive. In a world where legal AI child porn abounds, it will much more easily slip its bounds. I have myself, very much without wanting to, seen more of Aella’s body than I’m comfortable with, because the social lines separating porn are porous. More of us are going to casually see images of child porn if and when AI child porn becomes legal and available. As that happens, seeing child porn will be normal.
Moreover, it will be hard to detect the difference. Oceans of fake child porn will conceal the fact that, yes, some people do want the real thing, and now they have perfect legal cover. Law enforcement may not even be able to tell the difference. As a result, any enforcement of image rules will become impossible. The upshot is that we will effectively be legalizing all child pornography, and I don’t think
wants that. I think she just hasn’t thought the steps ahead to consider what enforcement of still-lingering actual child pornography would look like in a world where the volume of child pornography in society rises 1,000,000%, the AI stuff is perfect quality, the AI stuff slips its formal bounds fairly easily, and some degenerate weirdos still want real kids. In that world, the existing bounds on child abuse and child pornography would effectively collapse.Conclusion
The satiation hypothesis with respect to porn is wrong. Rising porn has been associated with rates of sexual crimes falling far less than other crimes, suggesting that porn does not substitute for sexual crimes. The emergence of new sexual norms in porn, even norms unpleasant and offensive to many people, are rapidly replicated in polite society. Because sexual interest in children is in fact relatively common historically and cross-culturally, there is no reason to believe pedophilia would be any different. And whereas the disgusting wave of men-demanding-anal-sex may be cresting and now falling as women assert their sexual interests, we cannot expect any such wave for AI child porn. Rather, an infinite wave of AI-generated child porn will create new norms of allowable child sex, and provide an impenetrable shield for actual child abusers and child pornographers to hide behind.
In sum, we should do everything in our power to prevent the legalization or normalization of child pornography.
And that’s even without considering the fact that harm is not a reasonable standard for morality or law, that in fact child porn being disgusting really is sufficient basis to put child pornographers in prison forever.
Surely the giant spike in NCVS reported sexual assaults in 2017-2019 is #MeToo and not an actual catastrophic crime wave, yes? That is, people became much more likely to internally classify borderline incidents as rape, or recall past incidents more vividly, or some combination of things like that. I don’t mean to say that victims weren’t by and large truthful, just that social norms can have a huge impact on certain data series independently of the objective facts they try to measure
Do you think that the fact that society thinks about sexual assault and rape quite differently now than in previous decades (e.g. me too movement) is going to be a confounder in some of these statistics over time?