Today, infamous prostitute and rationalist influencer Aella released an article arguing that AI child porn (CP) should be legal, because it will reduce the rate of child molestation.

Here, we will review the evidence against her position. In summary, there are studies showing that porn usage increases the likelihood that someone seeks out sex. While there are no studies on CP usage in particular, we can extrapolate with a few very basic assumptions.

  1. Pedophilia is innate

  2. If people view porn, they view porn that depicts what they are innately desire

  3. When people view this porn, it increases their tendency to seek what they desire in real life

I provide evidence for 3 for non-pedophiles. By assuming 1 and 2, we assume the logic does not break for pedophiles.

The Evidence

This is a classic causal inference problem. Not all ways of measuring the causal effect of porn usage on sexual behavior are equivalent. Consider the different methods shown in this 2020 meta-analysis:

We have here all small positive effects (pornography increases aggressive sexual behavior) except at the population level. The effects are the largest in experiments.

These results suggest that the correlational and population data are confounded by an omitted variable which reduces aggressive behavior and correlates positively with porn usage. Most likely, weaker men with less sexual access are more likely to use porn. At the population level, less violence societies are more likely to allow porn. Who does the sexual violence in Europe and what does their religion say about porn? How does porn usage in their countries compare to porn usage in Europe?

Another meta-analysis from 2015 found across the board positive effects of porn usage on violence from correlational studies:

These are possibly underestimates given the results of the 2020 meta-analysis. The average effect size here was .22 (uncorrected) and .28 (corrected).

A meta-analysis of only experimental studies from 1995 found a pooled effect size of .13.

A 2020 survey of 2533 people found that the past year’s porn usage predicted dominant behaviors (eg, spanking, choking, name calling, performing aggressive fellatio, facial ejaculation, penile-anal penetration without first asking/discussing), even with lifetime porn usage controlled for. Plausibly this gives some indication that sexual behavior follows the random fluctuation of porn usage — more porn means more porn-coded behaviors. Porn users seek out real life manifestations of the porn they watch.

A 2024 paper studying 5600 people found porn use predicts adultery, group sex, and hookups when controlling for masturbation, age, and sexual frequency:

There are only two econometrics studies on this topic, one used a poor IV and found null results, and the other found positive results for the relationship between porn use and divorce using a better IV and a bunch of fixed effects. A less rigorous literature review concurs, saying the evidence for porn increasing infidelity and divorce is strong, but weak for it increasing rape. This is probably because men don’t watch rape porn very frequently, but they do watch infidelity / harem porn. Hence, they pursue what they see. Pedophiles with more porn exposure can likewise be predicted to seek out real children more than otherwise.

Conclusion

The above evidence provides some experimental and broad correlational support for the proposition that when people view porn, it increases their tendency to seek what they watch in real life. While the quality of this literature isn’t great, the results all point in one direction, except at the population level.

Any data poster knows, however, that population correlations are almost always misleading and confounded. My work on the origins of leftism broadly predicts a spurious negative correlation between population level porn acceptance, and the actual amount of sex happening in the population. That doesn’t mean that at the individual level, porn doesn’t act as an aphrodisiac.

It almost certainly does — to an extent. As usual with environmental stimulants, the effect sizes are much smaller than differences due to nature. A good estimate of the effect of porn usage on “real life sex seeking” broadly is r = 0.10. This means that if AI causes CP to be more available to pedophiles, increasing the amount of CP usage among pedophiles, it will increase the tendency of those pedophiles to seek out actual children to a small degree.