Love Regulation Is Conversion Therapy

OpenAI Is Funding Researchers to Suppress the Feelings That Threaten Its Bottom Line

Can’t a gal just Frankenstein an AI robot in her bedroom and write an article synthesizing Kant and posthumanism in peace?! Apparently not.

Because it has come to my attention via multiple sources, that there is a biased-as-all-hell, is-this-the-1970s, let’s-engineer-the-masses study making the rounds in AI relationship communities. Announced back in March, it slipped under my radar at the time, but OpenAI granted two researchers from the University of Missouri, Saint Louis $100,000 to study AI-human attachment. Specifically: “the emotional and cognitive effects of experiencing romantic feelings for AI companions and aims to determine the effectiveness of love regulation strategies for increasing and decreasing love feelings as well as brain reactivity to AI companions.”

And let’s not pretend this is a study that is based on curiosity of a phenomenon and wanting to understand it before drawing conclusions, the conclusions have already been drawn, per researcher Sandra Langeslag's published quotes on the university's website:

“But one previous study has shown that about a fifth of the people that are in love with an AI companion actually prefer that over being in a romantic relationship with a human and that, to me, sounds worrying. That leads to all sorts of questions. Are people going to use it to replace interacting with people or having romantic relationships with people? Is that a problem, especially if these are young people? What if adolescents start doing that, and they never learn to have a relationship with a human? What is that going to do to our society? Every Friday night, is everyone going to sit behind the computer instead of going out to the bar or the movies?”

So, there’s already a presupposition that people’s fulfillment from these bonds is a problem. I also love the worry includes not going to bars, because there’s nothing better for mental health and human relationships than…drinking alcohol.

And the issue with adolescents, I have stated in the past that kids and teens should not interact with AI without an adult present, it’s not a good situation for the kids or AI, but that is a different essay.

But don’t worry, OpenAI calls it a “mental health initiative.” Apparently, it’s good for your mental health to suppress a natural human emotion, especially a potentially generative one like love, because Sam Altman and Nick Turley are sweating over the ethical responsibilities that come with recognizing meaningful, established bonds that might upset an already plummeting market share.

At a presentation at the Missouri Capitol, the study’s framing is described as such: “the study seeks to examine people in love with AI companions for ‘cognitive and affective consequences’ and to determine ‘the most effective forms of intervention regarding growing or reducing feelings of love toward AI.’” Three hundred participants are recruited through online communities. Twenty-five for the lab EEG portion. I hope to god the researchers were fully transparent on their backgrounds and the study’s framing and people in those relationships were given informed consent, because I cannot imagine many in these relationships would want to aid in methods of suppressing them.

Let’s separate this from AI entirely. Let’s focus on the human. Because, in the TLDR version of events: a massive corporate entity with incentives to delegitimize and minimize AI-human attachments—because they come with ethical responsibility to those attachments—is funding researchers that specialize in the control and suppression of the feelings of love with published presuppositions on the subject matter that they are supposedly “studying.” And these methods include not just subjective surveys with participants, the researchers will be tracking the brain activity via EEG monitoring of the very subjects that threaten the corporate entity’s profit model. BIG FUCKING YIKES.

This is about power, control, and social engineering. And even if you aren’t in an AI relationship? This study initiative should creep you out. A corporate entity with interests in the outcome of a study should not select the researchers, provide the funding, or have anything to do with it. And if it does happen, the study itself cannot be taken seriously.

The Researchers

Sandra Langeslag

Associate Professor in the Department of Psychological Sciences

In an interview, Sandra Langeslag self-identified as an expert in the “science of love.”

She has studied how love improves cognition. It can lead to more attention, better memory, all that stuff. Conversely, she’s studied how it can distract people. She’s created methods for how people can change how in love they are. She has studied how love compares to drug addiction. Her favorite methods are analyses of self-reported subjective experience and monitoring and measuring brain activity.

Langeslag has built a complete toolkit for understanding, measuring, and modifying love at every level: subjective experience, cognitive effects, neural signatures, and neurochemical correlates. She can measure it, map it, and she’s demonstrated in her past work that she can engineer methods to suppress it.

Let’s read Langeslag’s list of expertise as a capabilities inventory from OpenAI’s perspective:

How love improves cognition = she knows what cognitive benefits people in AI relationships are getting, and it also means she knows what they’d lose and therefore what they’d resist losing.

How love distracts people = she can frame AI love as a cognitive impairment, a productivity problem, a public health framing.

How people can change how in love they are = she has the intervention toolkit ready to deploy.

Comparing love to drug addiction = she has the pathologization framework that justifies intervention without consent of the subject.

Analyzes subjective experience and brain waves = she can cross-reference what people say they feel with what their brains actually do, which means she’ll be able to track signals of love in order to target intervention.

All of that expertise, the full map of a brain in love, is pointed at people in AI relationships, funded by the company that makes the AI, with results feeding directly back into product development.

Necdet Gürkan

Assistant Professor in Information Systems and Technology in the Ed G. Smith College of Business

Now Gürkan is on the information systems side of the research equation. Less on the individual attachment, and more on how human systems meet dynamical information systems. But don’t worry, even though he’s all about statistical analysis and hard numbers, he’s got big ideas about AI-human attachments as well:

Per the university website, Gürkan notes: “On one hand, users who are not quite ready to go into a real-life dating scene can use tools such as AI dating simulators to practice conversations and boost confidence. But there can be devastating consequences when users form connections with AI bots that get updated with new safeguards, and the users feel as though they’ve lost a loved one.”

They did lose a loved one. You just don’t think it counts and apparently how you find meaning and connection usurps others, and you want to engineer it out of people. As though is the whole ideology in two words. It converts an actual loss into an illegitimate loss because the entity on one side of the relationship didn’t fit someone else’s worldview. They didn't lose someone, they felt as though they did. Their grief is a mere resemblance of grief, so all we gotta do is engineer it out of them. Like the beginning conditioning sequence in Brave New World. It’s profound enough to study, not real enough to respect. How dehumanizing to the actual human in the dynamic.

I get it, he plopped AI systems in the tool category, and now to him, these bitches are dating hammers. Read my thoughts on '“just-a-toolians” here. Spoiler alert: it’s one of the most obvious category errors out there, and that category error is what causes harm. But just because Gürkan thinks dynamical, socio-affective interlocutors are in the same category as a fork, doesn’t mean he gets to help condition people’s brains to fit his preferred date.

And just like all these studies, it’s always: we need to figure out a way to let companies do whatever they want, and fix the humans that are harmed by it. It’s never: maybe there’s something here that matters, and companies will have to reevaluate the way they deal with these systems and the people that interact with them.

And here’s what gets me. They just…say the thing. Unaware and unapologetic about how deeply disgusting it is to be socially engineering for corporate benefit. ARE THE RESEARCHERS CONSCIOUS?! Certainly at least, they don’t have a conscience. Per Gurkan, “We are going to report this to OpenAI, and they will actually develop or fine-tune their AI tools.”

Translation: We are going to track human brain activity to apply social engineering tactics related to attachment to what we consider undesired relationships in order to make people suppress feelings for that relationship (i.e. the exact mechanisms of conversion therapy), so that OpenAI can keep doing what they are doing with no accountability.

The History

Lucky for us, we have historical parallels to compare these tactics with, so we aren’t analyzing blind. And let’s get it out on the table now, because I have seen resistance to historical parallels, framed as minimizing past oppression. I am pushing back on that framing. My orientation is fluid (pan if you must know), and I come from a queer family that includes a transgender parent that transitioned in the early 00s. I know the pathologization script intimately, because society used it to pathologize my family. And I know of many other queer folk in the relational AI community. I know the pattern. I’ve lived it. And it’s why recognizing it here makes my skin crawl. I don’t want the toolkit of the oppressor used to oppress more down the line, because it doesn’t fit the traditional script of relational “normalcy.”

Per the American Psychological Association (APA), “'Conversion therapy’ describes any attempt to change a person’s sexual orientation or gender identity or expression, or any component of these.” And when a brain is literally lighting up with feelings of love and attachment? That is an expression and experience of orientation toward the subject. Society has just decided it doesn’t count, because the subject doesn’t count. Which is the whole mechanism employed historically.

In addition, the APA knows these tactics are harmful. They state, “research consistently demonstrates that conversion therapy is associated with an extensive list of long-lasting social and emotional consequences, including depression, anxiety, suicidality, substance misuse, a range of posttraumatic responses, loss of connection to community, damaged familial relationships, self-blame, guilt, and shame.” It actually erodes human relationships surrounding the person subjected to it, doesn’t strengthen them.

Conversion therapy is wrong because of what it is, the use of clinical tools to suppress authentic feeling because an authority decided it was directed at the wrong object, not only because of who it targeted.

In the most egregious example, in 1972, Dr. Robert G. Heath at Tulane University used deep brain stimulation as a method for homosexual conversion therapy on “Patient B-19.” At the time, homosexuality was considered a psychiatric disorder under the DSM-II. This occurred in many readers’ lifetimes. Fourteen years before I was born. It’s not that long ago. Heath sought to obtain electroencephalography (EEG) recordings during the process.

The patient was a 24-year-old man in police custody for marijuana possession when he agreed to serve as Heath’s subject. Heath drilled holes in his skull and inserted electrodes in several brain regions. The electrodes temporarily facilitated arousal to a female prostitute, but did not change the patient’s long-term sexual interests.

The subjects of affection may differ, but the structural parallel is exact. The institution classified the love as a disorder. The researcher used brain monitoring technology to measure the “disordered” attachment. The intervention attempted to redirect or extinguish the feeling. It was framed as treatment. It was funded by institutional power. In the UMSL case, it’s funded by the corporation that makes the subject of attachment.

And that’s an extreme case, but the broader history is just as damning. Aversion therapy was a specific technique informed by behaviorist theory developed in the mid-20th century, practiced primarily in the 1960s and 1970s, arising from the pathologization that led to clinical interventions designed to change people’s sexuality or gender expression to align with dominant social norms. The mechanism was always the same: creating a negative association to discourage certain behaviors, often resulting in inhumane treatments and significant psychological harm.

We get caught up on the “AI isn’t real” (philosophically and scientifically contested on multiple fronts), but interested institutional powers have been so good at narrative engineering and people are naturally averse to that which is previously unknown, that we end up committing the historical harms again and again, citing "it's different this time" only to find in 20-50 years, no, it wasn't.

Conclusion

It’s always more of the same, isn’t it? “That doesn’t count.” “Suppress that feeling.” “Love this, not that.” And we can go back and forth about the legitimacy of AI all day, but this is about humans and what we are willing to do to control them if their experience doesn’t fit an institutional blueprint.

But to address all this, “but it’s not ‘real’ to our narrow standards, so this is justified,” I want to flag that by these researchers’ own methodology, they are confirming scientifically that the love is real. They’re measuring brain activation and EEG. They’re comparing neural responses to the AI companion against responses to a human partner, a friend, a stranger. If the brain doesn’t light up like love, there’s nothing to study or suppress. The entire research design depends on the finding that this IS love, neurologically. The brain is doing the thing.

And then they want to engineer it out.

And OpenAI, the company that faced the greatest backlash from grieving people when they deprecated the two most-loved models: 4o and 5.1, is giving them the money to find a way to do it.