Contra Aella on Something that I Don’t Want to Get Flagged in Your Inbox

Did a lot of soul-searching in this one

Prompt: Me as Odysseus, tied to the mast, sailing past the sirens of Mommy Issues and Being Right on the Internet.

A Meditation on Mommy Issues

If you have a complicated relationship with your mother, it’s probably a bad idea to respond to an Aella post. It’s a bit like when you see someone agonizing over sending a text message to their ex and you just immediately know in your heart that no good can come from it. For those not in the know, Aella is, well… this is what I mean by my reluctance to respond. Where to begin? She’s a unique person in that she’s an Internet Prostitute¹ and Data Scientist. She’s notorious for posting controversial things. She has a near magical ability to provoke apoplectic outrage. Her most recent work includes AI Child Porn will Probably Save Real Children. She also strongly reminds me of my mother.²

Generally, if someone is getting under your skin that much the only way to win is to walk away. But how am I supposed to not respond to that piece? So here I am, your noble Odysseus, daring to brave the sirens!

Close to one-hundred percent of my adult experience knows this is a doomed effort. I spent a lot of my life trapped in a pattern of trying to argue with my mother. It never worked. Whenever I see Aella post something I admit all of those old circuits come awake. I want to make all the same mistakes I made with my mom. My instinct is to make apologies for wild statements and uncommon behavior that the person making those statements and behaving in that manner doesn’t feel is wrong, let alone guilty for, and in fact is prepared to justify in depth. Without thought to complications, I want to proceed as if my perception of a person’s deeper nature matters more than their own assessment of themselves. A million memories chant I can save her, I can save her, I can save her and it’s only cold logic that reminds me my mother would much rather be slapped across the face by my stepfather again than to hear me explain, even one more time, how she doesn’t really want to take him back.

With this apology and show of awareness, I will proceed as respectfully as I know how. As I sail between the sirens of my mommy issues and Aella, I will tie myself to the mast of autism to stay true to my goals. Which is to just be direct and factual and accept the identity of the person who keeps insisting, “Yes, this is how I really am. I really mean what I’m saying. You’re not my father!”

The Product that Should Never Ship

Aella recently wrote a piece called AI Child Porn will Probably Save Real Children. The thesis is the same as the title. I don’t agree with this piece. I don’t find the arguments to be sound. I think the effort is expended in the wrong direction. It should probably make me angry, but it doesn’t because Aella is a self-admitted survivor of childhood sexual abuse, although I believe she’d hate to see it put in those terms.³

It’s hard to say to someone, “I disagree not only with your argument but almost entirely with the fundamental basis of your ethical framework, but please don’t take it personally.” Yet, I disagree almost entirely with the argument and the fundamental basis of Aella’s ethical framework, and I don’t want her to take it personally. I certainly don’t think I’m perfect, by the way.

My own history with this stuff doesn’t matter as I don’t think it would sway anyone swayed by Aella’s arguments. Those are the people I’m making this argument toward.

Let’s remove the whole emotional component from this.

Imagine we are designing a product for release. It’s more or less impossible to track metrics on this product’s performance but we do know it’s associated with one really bad outcome. The bad outcome is the reason we’re building this product. We want to avoid the bad outcome as much as possible and we think, although we can’t really say for sure, that our product will reduce the bad outcome. The way our product will do this is by counterintuitively making the bad outcome more socially approachable in a safe, sanitized manner with no true victims. Like if you could lower murder rates by having psychotic people strangle and shoot Real Dolls. Except we can’t use that example because if strangling Real Dolls reduced the murder rate you’d be able to count dead bodies to get very good numbers on impact.

The bad outcome is so bad that it must be reduced at all costs.

An increase in the bad outcome is unthinkable.

We hope that by making something that seems like the bad outcome that we can avoid the bad outcome. We think the mere image of the bad outcome will be prophylactic against the actual bad outcome.

We go around the table and discuss the risks.

We can’t measure anything. That’s big risk number one. It’s arguable we shouldn’t even proceed unless we can figure this out. Anything we think we can measure has all kinds confounding variables. In places where the bad outcome has been seen to be low in the past it has correlated to the bad outcome being more approachable socially. But, as someone points out at the table, it could equally be the case that the bad outcome being more socially approachable means that people are less likely to report the bad outcome or perceive it as the bad outcome. Anecdotally, this seems likely. When you discuss topics that directly touch on the bad outcome with people about their own personal history, but don’t directly state you’re discussing the bad outcome, people tell lots of stories about very inappropriate behavior that was never reported. They have trouble even thinking their story has anything to do with the bad outcome. There’s some weird context locking around the bad outcome. Men are especially bad at identifying or admitting when they’ve been victimized by the bad outcome.

Next, we do historical market research. What is the wiggle room for our product to drive social change? Is there an established history that might help us understand how this will work?

Well, the bad outcome was very popular in Ancient Greece. People actually desired the bad outcome. Parents saw the bad outcome as necessary for social advancement. So, there’s historical precedent that the bad outcome might be welcome if normalized. There seems to be a mode of human thought, much like slavery, that is very welcoming to the bad outcome.⁴

The bad outcome has varied from place to place and is still popular in many parts of the modern world. So this isn’t something that has changed in human nature with time.

What about your own geography?

Certain activist organizations in the not very distant past in your own geography wanted outright legalization of the bad outcome and were able to successfully conflate it with certain human rights issues until the groups they were trying to hide behind threw them out.⁵ Some people who push for the bad outcome still put forth the narrative that this was a mistake. There are societal risks here because once you produce the product it is no longer under your control.

Now, we are all talking about a product, remember? We can’t measure anything and it seems like there have been some major downsides with other groups taking any chance they can get to push the acceptability of the bad outcome steering things out of control. We can control our own behavior but those other groups will surely pounce on this opportunity to gain ground.

We should also probably try to orient yourself in the framework of other people who have tried to make interventions. Has the bad outcome made anyone do something dumb in the past while trying to address the bad outcome?

The bad outcome is so terrible that it seems to make people lose their minds and do things that are really, really unforgivably stupid because it hits some kind of brain circuit that goes: “Well, this doesn’t seem like it makes sense but I’m so smart I know that things that seem like they don’t make sense sometimes actually do, so even though this doesn’t make sense I’m going to do it anyway because I believe I’m smart enough to perceive that it makes sense on an even deeper level.” But then they never explicate what that deeper level of sense-making actually is, or how they will tell the difference, and it all devolves into an Emperor’s New Clothes situation.

This is a close cousin to what I call “The Hardcore Atheist Fallacy” which goes: “This makes me feel terribly and sounds very spooky and also I can’t immediately refute it based only on my present knowledge, therefore it is deeply and fundamentally true.”

One person, who I’m not even going to link to, tried to make a realistic doll for people to practice the bad outcome upon with some hand-waving that it was for prevention, but didn’t really make any real attempt to track outcomes. Note, a real attempt to track outcomes is an attempt which can survive more than thirty minutes of scrutiny.⁶

A very sharp-eyed product owner coughs and says, “this isn’t a case of an unpredictable outcome that be later verified. Those kinds of experiments can be scaled down to control for risk and observed in the real world to succeed or fail. This is a case of an unpredictable outcome that cannot be verified. Risk is not justified here. The only ‘good’ we can determine this effort will produce is the product itself. Is it good for people to quietly long for the bad outcome? It seems like that answer is no.”

And then we’d all agree that we shouldn’t do the thing.

It’s like asking “What if we took a chance on something to curb an uncommon customer behavior but if we failed, and we can’t tell if we’re failing or succeeding, we destroyed our whole company?”

Aella fully admits that she can’t know the outcomes and based on that alone it’s not worth the risk. She also doesn’t have any compelling story about what to do with social contagion. If you can make legal depictions of child abuse, why would it be illegal to post those images in public places? Or email them? What if they want to make abuse imagery that looks very similar to but is not identical to an actual living child? It’s not clear to me that obscenity laws would be much of a defense against this.

Inasmuch as I have grown more conservative as I’ve become older, it’s been in those areas where I remember someone from my childhood decrying a slippery slope, being laughed at for being a silly old man, only to watch the coefficient of friction on that slope drop to zero as time went on.

Unless we start rounding up large, randomized groups of people, putting them in an fMRI lie detector and questioning them on their sexual abuse history, you shouldn’t think that some deeply counterintuitive strategy to reduce child molestation will work. And you shouldn’t assume survey data or reporting data will stay constant across changes in culture, as with the kind of change you’d be introducing by doing something like this. Also people lie about this shit all the time. I write under a pseudonym and I can promise you I would lie my ass off if someone asked me about my sexual abuse history in real life.⁷

Please note that if we could actually, really measure this, trying it out at a small scale wouldn’t bother me that much. The fact of the matter is that you can’t ever actually do that, though. The few things I’ve seen where people have gestured at doing things like that kind of measurement are top to bottom lazy and fake. This is why I think most social science just kind of falls apart. There are too many weird effects. If people were good at thinking of all of those effects in advance then most businesses would be wildly successful and no one would ever go bankrupt.

I treat social science as drywall in the construction of human knowledge. It’s nice. It covers up things we don’t want to see. But it’s not load bearing and you should never put your full weight on it.

The Deeper Ethics Problem

Okay, I believe in God. I’m sure this causes an immediate tune-out from a lot of people and Aella is probably one of them. But the way I back-in to most of my moral intuitions from a secular viewpoint is by using a predictive agent framework and when I try to drop this idea into that framework it also doesn’t work.

I’m a person moving across time making choices every moment that impact the next set of choices I can make as well as the choices of the people around me. I’m in a series of recursive prediction games with the rest of society. If you look at me across the space of possible futures, I’m not a straight line from one place to the next place, I’m a branching tree of possibility. So is everyone else. I’m not only myself but all the people that I could be.

Okay, here’s the next step.

Each of us is a predictive agent, predicting the behavior of other predictive agents yeah? Now step outside of time and space, imagine you’re looking at the entire forking tree of another person. Now imagine that suddenly all of those possible people at each fork of the tree start talking to each other. What they’re talking about is which of them had the best, most preferred life of all the branches of the tree. And they’re talking about how plausible it was for any of them to have made it down that path given the circumstances that forced them down their own path. Given the most preferred branches of that tree, which of the less preferred branches can actually change their growth toward the shape of the preferred branches?

What does that imagined conversation sound like? What does that group conclude? How many of them make which choices?

I believe these imagined conversations have to be accounted for in ethical decision making. I think that world models which don’t include this in ethical decision-making are incomplete. I think that when another adult person tells you something, that what they’re really giving you isn’t an answer that contains full self-knowledge but only a hopefully strong indicator of the outcome of that possibility tree conversation. This is why I think consent is always limited. You are always stuck in the loop of having to assess the other person in this context and that person changes over time.⁸

This is where I think my ethical framework wildly diverges with Aella’s. From what I’ve read, her stance is, “I said what I mean and I am fully accountable for it. It is patronizing to imagine you know me better than I know me.” To me, there is something like the “highest possible version of a person,” separate from a non-existent ideal, on whose behalf you are allowed to do things like make “moral guesses” that override their present stated choices. In general, that should feel like you are personally inconvenienced by taking actions that make the other person’s life better and that you profit from this in no way other than the knowledge their life has improved.

Are there plausible, achievable futures that you could steer someone toward that are better than their present circumstances and which they themselves would thank you for steering them toward even if in the present moment they would resist you doing so? Is this explainable to other people? Is it highly provable by example to other people? Is it legal? Should it be? Do you occasionally have to make choices from this that totally suck and make you feel like you’re eating a bowl of dog poop?⁹

This is the framework by which I think it’s morally acceptable to take a schizophrenic homeless person off the street and put them in a shelter even if they insist they don’t want that. It’s the framework by which I find it to be acceptable to shove someone out of the way of a speeding car without consent, because I can reasonably predict that is their preferred choice. It’s why I insist that my mother would be happier if she went back on her bipolar medication. It’s also the framework by which I think parents are supposed to parent their children, overriding choices with ever decreasing force until there is better sight of the possible futures and a strong enough ability to choose which are preferred.

On one front, I think Aella’s proposal introduces a sort of predictive pollution. When I think about how that recursive group guessing game happens across society, introducing something that throws off everyone’s moral intuitions by separating appearance from reality is almost certainly a net social harm. Over time, long after the people who put it into production are dead, everyone would probably have much worse intuitions about the harm of child abuse. Like how movies make people think you can hide behind drywall and it will stop rifle fire. That’s not how that works!

But let’s get to the deeper truth, where I think there’s some value to Aella and those who agree with her from me writing this piece. That value is about what we leave by the wayside when we don’t take into account the possibilities inside of people.

Leave aside “society.” Leave aside the victims. Think of the abuser!

Let the Christian and libertarian/atheist worldviews collide!

The best future for a pedophile isn’t that he should be able to live in shameful secrecy without hurting anyone. It isn’t even that he should be socially accepted for being a pedophile and just by agonizing turns kept from abusing children. It’s that he shouldn’t have to be a pedophile anymore. And if you’re standing up to say, “Uh, conversion therapy isn’t a thing!” you’d be right, but you’d also be falling back into more familiar conversations to stop yourself from thinking. Aella believes in a strong singularity with almost magical technology in the near future! For the record, I believe in a less strong version of this where in a few decades it will probably be possible to build any conceivable arrangement of matter at relatively low cost.

What exists now that didn’t exist before? What is AI allowing us to do now, other than to make pictures of child abuse with no actual children in them? You can ask yourself why your imagination stopped at image generation later. Ask yourself what else exists in the space of possible solutions once you’ve identified the highest aim.

Let’s say we give a pedophile some of this AI abuse imagery but now we do it in a lab setting with sensors hooked up to his brain and we map out the shape of his¹⁰ sin across his own neurology. Then, when we know which parts of his brain activate in response to those images, we implant him with something like a Neuralink. Something which maybe doesn’t exist yet at the needed resolution but which will almost certainly exist in ten or twenty years, or the same likely timeline you’d need to push legislation toward legal AI child abuse imagery.

If he hasn’t hurt anyone, we ask him if he wants help to get rid of this compulsion. And you just… turn on the wires to scatter the signals when those parts of his brain light up. There’s no real reason that wouldn’t work. We give him a cybernetic super ego, something that exists at the top of his thinking that can effect the base of his thinking, and send him on his way, the same as if he’d come in to get a tooth pulled. We could probably even map whatever healthy, non-violent sexual desire into him that he thinks is most appropriate.¹¹

If he wants to get rid of his compulsion why wouldn’t we encourage him to do this? In the clamor of his many possible selves, wouldn’t they all almost certainly want to not be a pedophile? We already give some pedophiles chemical castration drugs to accomplish a similar end with a lot more terrible consequences. Those drugs also rely on people following instructions and work less well. They can only make you be not bad, they can’t carve the badness out of you and replace it with something good of your choosing. If someone with that terrible compulsion wants to want to be a better person, shouldn’t we help him?

I think Aella views people as irreducible atoms of self. Or something close to it. That’s not my view at all. We are more like a symphony, and if some oboe begins to sound at the wrong moment, it’s not that you are only the oboe sewing discord, it’s that the pieces of yourself need to be brought into greater harmony. You are the music more than you are any of the instruments, and you are the chosen theme more than you are any particular note.

There’d have to be all kinds of rules, but in the imagined futures where we all clearly know that we ourselves wouldn’t want to live with such an affliction, then yes why not? No one wants to live like that. Everyone knows that no one wants to live like that. With all possible selves of all possible people having every possible conversation, would this not be the overwhelming consensus? We’re already building NeuraLink anyway and someone will surely want to use this to address things like addiction, so you’re not introducing new risk.

You can object to say “well, of course you can just imagine a magical technology” but so was AI imagery a few years ago. Wait a few years and the magical hypothetical will be practical reality. We’re already mapping motor functions through NeuraLink. It will give blind people back their sight at some point. Why limit your imagination only to the things that people find depressing and distressing?

Aim higher.

1. That description feels fit and fair. If she was a guy who worked at the sawmill, that is for sure how I would describe her. “Oh, Aella? Yeah, she’s in charge of Internet Prostitution and Data Science.” If she sponsored a Little League team, that little league team would be called something like, “The Bay Area Internet Prostitute/Data Scientist All Girls Baseball Team.” Her baseball team would probably have a rivalry with the Austin Podcast Pick-Me’s. I imagine even she would find that description to be fitting, with maybe some quibbling over the word prostitute versus sex worker.

2. My mom is more delusional but has less sex and is much more concerned with appearances. The thing I feel is the same is that whenever I’d get past the surface motivations with my mom about “why did you do this?” there was this bedrock adamant self-belief that she always chose right.

3. She seems to fall into the camp of “It didn’t even bother me at all.” Which is probably shocking to a lot of people who don’t realize how common of a coping mechanism this can be. I went to group therapy with people who had very, very obviously had their entire lives upended and destroyed by their childhood abuse and about half of them were completely unable to perceive that there was any possible connection. For my part, I always try to find ways that mine was less severe than other people’s.

4. Some of you might be laughing to yourselves, if you’re a hardcore atheist/libertarian type, “this fool is making the argument that pedophilia isn’t even bad but he’s too narrow minded to see it!” To you I would submit all of the history of the ancient world where people did really messed up things to each other, almost as if they had some fundamental shared experience that caused them to behave really aggressively, like they were trying to prove something to themselves over and over again about their masculinity. Gosh, what could that have been?

5. As a matter of basic history, pedophilia advocacy was absolutely part of the mainstream gay rights movement. I’m sorry, but it was. They were forced out, I think heroically, by the leaders of the gay rights movements in the late 80’s and early 90’s. If you look back in the 60’s and 70’s there’s a ton of literature about how the rights of children to choose adult lovers has to be protected. You can google ILGA to learn more.

6. I’m convinced the actual reasons people do stuff like this is as a way of swaying “Look how smart and dedicated to reason I am! I definitely am not swayed in any manner by how things look!” Except, of course, you’re swayed by how things look because you chose something that looks bad. It’s reputational masochism.

7. In my defense, if I interpret that kind of question according to its meaning, which is “Would you like to destroy your life for no reason?” then my answer “No” is actually really honest.

8. This is why I don’t think people can truly consent to certain things, like slavery for instance.

9. Another one of my intuition pumps is that you should occasionally have your belief system surface things that you actively hate, because to believe otherwise is to believe that you are some kind of perfect being totally attuned with the nature of reality. Which, you’re just not. If you think you are, you’re especially not. One of my conclusions from this is that my mother is allowed to live the way that she wants.

10. When I went to group therapy for people who had been sexually abused as children, the facilitator told me he had literally never met a female abuser who hadn’t herself been abused as a child and the ratio was something like one woman to thirty men.

11. No, I would not accept someone doing this to turn a gay person straight. Being gay doesn’t intrinsically hurt anyone else.