The algorithms defining sexuality suck. Here’s how to make them better.

Mashable’s series Algorithms explores the mysterious lines of code that increasingly control our lives — and our futures.

Ever since porn was credited as one of the most innovating forces behind early internet technology, we’ve become obsessed with the idea of tech enhancing our sex lives. We’re so horny for it that we’ve helped build a that’s expected to keep growing.

Sextech often sells people on the promise that algorithms can optimize users’ sexual experiences. But a vast majority of algorithms built explicitly for pleasure remain rudimentary at best and harmful at worst — including those used in s and .

That’s because a lot of sextech relies on a grossly reductive view of sexuality. Exhibit A: The all-male startup that claimed to invent an . Exhibit B: The fellatio machine which promises “the perfect blowjob” thanks to artificial intelligence fed porn video data.

Even the most advanced, well-intentioned sextech is held back by a lack of legitimate sex research, accurate data, and designer diversity. That’s on top of the biases built into algorithms, overstated tech capabilities, marketing gimmicks, and Silicon Valley capitalism. 

“The pleasure product industry is one of the few industries that has been relatively untouched by modern technology,” said Liz Klinger, co-creator of the , which tracks and generates charts of your vaginal contractions during arousal. The biggest trends of VR and remote control smart toys, she pointed out, use tech that’s decades old. “Existing companies just don’t understand how software, data, AI, or other technology can introduce new experiences or appeal to new, different demographics.”

The failures to integrate algorithms into sexual exploration and expression go beyond an outdated adult toy industry and bleed into all corners of the internet. As it stands now, the binaries encoded in algorithms seem almost diametrically opposed to the complex spectrum of human sexuality. 

But there are ways to change that.

The binaries encoded in algorithms can seem almost diametrically opposed to the complex spectrum of human sexuality. 

“Technologists write algorithms that are interacting with these very complicated social systems with no consideration or background in their complexities. But there’s already a lot of information out there on how to approach gender and sex that you just have to incorporate into your algorithm,” said , the founder of , a social media sharing platform that uses machine learning to create a safe space for women and LGBTQ folks to express themselves sexually. 

The algorithms policing sex on social media have such little nuance that they can’t even differentiate porn from sex ed, sexual health, or sex commentary. Sexism and homophobia are so entrenched in how platforms like Facebook and Instagram police sexuality that ads for women’s sex toys and HIV/AIDs prevention are banned while ads for condoms and erectile dysfunction pills are allowed. Those same biases plague algorithm-driven sextech devices, too, which often impose false and exclusionary ideals about what the “best” sex should feel like. Lack of scientific research and data around pleasure and sex, especially when it comes to people with vaginas, makes it almost impossible for sextech to deliver on its promises of sexual optimization.

It doesn’t have to be that way, though. Algorithms don’t need to default to constricting heteronormative male views on sexuality. Instead, a few companies are grounding their algorithms in more inclusive feminist approaches to sexuality in the hopes of countering these cultural biases. 

But it takes investment to try something new, which the majority of the sex and tech industries have so far been unwilling to pony up.

“We’re seeing an increase in people using sextech to feel connected,” said , an activist for sex workers’ digital rights, vice president of , and a self-described thot leader. “But I always say, with sextech, we’re not teaching yoga or selling smoothies here. We’re dealing with something so intricate, so personal, so deeply rooted in all of us. We need to think carefully about the philosophies we’re putting into these algorithms and talk about their potential harms as much as their potential benefits.”

Once we start doing that, the sex-positive potential of algorithms are theoretically endless.

“There’s a lot that algorithms, software, and other technology can do to help improve pleasure and understanding of our own sexualities,” said Klinger. “For Lioness, some of the uses I’m seeing is utilizing real-world sex data to put different experiences of pleasure into context for our users.”

Perhaps the greatest potential for algorithms in the sexual wellness field might lie beyond just explorations of pleasure. According to Emily Sauer — the creator of the that allows couples to customize penetration depth to avoid pain — algorithms could help remove the societal shame of openly discussing our sexual difficulties.

“We turn to sextech to feel less alone,” said Sauer. “We want to know how we relate to everyone else through the tech, the data, because nobody’s talking about these taboo things that make us uncomfortable.”

The promise land of algorithmically-driven sexual exploration is like playing with fire, though. Algorithms are as capable of destroying healthy relationships to sex as nourishing them.

Fixing the algorithms that police sexuality

Time and time again, algorithms have been shown to perpetuate the implicit biases of human beings around gender and race. The most influential algorithms informing sexual expression in our modern world are no exception. 

Leadership at social media companies like Facebook and Twitter tend to be mostly white, heterosexual, cis men. They’re also the ones who get to decide what their platforms — and the algorithms that monitor them — consider appropriate sexuality versus obscenity, or sexual exploitation versus sexual expression on the internet.

Unsurprisingly, those definitions of sexuality are revealing themselves to be very narrow and discriminatory. 

Sex-blocking algorithms have been found to disproportionately censor marginalized groups, especially LGBTQ folks, sex workers, and women. One cybersecurity firm found that over 73 percent of online content flagged as inappropriate featured LGBTQ people. Aside from straight-up deleting and blocking accounts, the more subtle phenomenon of shadowbanning has been accused of enabling apps like Instagram and TikTok to exclude marginalized people’s content from discovery and explore pages. Instagram’s CEO has said shadowbanning “is not a thing,” while TikTok has admitted to the practice — although a TikTok representative said its intentions were good even if its execution was not.

“Accounts like Playboy and Kim Kardashian are welcome on social media platforms because they follow a mainstream view of sex. But the second that you have any alternative expression of sexuality that doesn’t cater to the male gaze, that’s something their algorithms deem unsafe and unacceptable,” said Brown.

According to SXNoir, the discriminatory, anti-sex algorithms on social platforms have been accelerated by legislation like FOSTA/SESTA. Aimed at stopping sex trafficking, it made tech companies legally responsible for all the sexual content on their websites. This incentivized social media platforms to censor sex even more, disproportionately impacting marginalized groups like sex workers and queer communities, who rely on the internet for safe sexual expression.

“What happens is that sexuality as a whole becomes collateral damage in digital spaces,” said SXNoir. “Sex workers become the scapegoats for censoring everyone’s freedom of speech, expression, privacy.”

Just look at Tumblr, which was flagged by Apple for child pornography in 2018 after the bill passed. Instead of investing in better algorithms that could distinguish between sexual expression and sexual exploitation, Tumblr banned adult content altogether. This left the queer communities and sexual subcultures who thrived on it with nowhere to turn. Other tech companies like Facebook are making similar decisions, implementing more punitive rules around sex rather than nuanced algorithms.

“We need to create technology for us and by us. The reason why technology does not currently work well for BIPOC, queer people, those engaging in sexuality, is because we are not the ones creating the rules, the Codes of Conduct,” said SXNoir. As seen in recent discourse around OnlyFans, “Black sex workers are very much pioneers of these digital spaces, the ones who [often need to] navigate sexuality on the internet the most. We’re futurists. And we deserve a seat at the table.” 

Lips is trying to do exactly that, by working with those marginalized by social media algorithms and making them the central forces powering new algorithms around sexuality.

“Black sex workers are very much pioneers of these digital spaces, the ones who navigate sexuality on the internet the most.” 

Launched a couple months after the Tumblr ban, it offered a new home for those who had been kicked off. Aside from being a safe sharing platform for honest sexual expression and exploration for the marginalized, it’s also a marketplace for those banned from advertising on other social platforms, like organizations who work in HIV/AIDS prevention.

Despite what giant social media platforms want you to believe, Brown said, it’s actually possible to create algorithms that can distinguish between legal and illegal sexual content. But making them takes time, investment, diversity among developers and users, a social-theory based moderation process, and explicit anti-hate speech agendas that protect marginalized voices.

Ultimately, though, doing all that just isn’t immediately profitable for those tech companies.

Lips prides itself on a machine learning system that empowers its diverse community of users to train the algorithm to understand the granular distinctions between sexual exploitation versus expression. Because while both computers and humans have a hard time defining those differences in concrete terms, people can tell the difference between, say, erotic art versus pornography when it’s in front of them. (Famously, in a court case trying to distinguish obscenity from erotica, former Supreme Court Justice Potter Stewart said, “I know it when I see it.”)

“You’re using the power of the collective to create the knowledge base and data set, which the algorithms can then apply in a way that’s faster than any one individual could,” Brown said of Lips.

Of course, this requires confidence in your user base, which is why Lips implements a gated entry. Anyone can browse Lips. But to participate in the community you must apply by submitting a couple posts you want to contribute to the platform as well as information on why you want to join.

Users who demonstrate an understanding of sex-positivity tend to be more qualified to make these judgment calls. Feeding their knowledge into the algorithm teaches it to be more effective and unbiased about sexuality than the cis, white, heterosexual male gaze that currently informs the rules. That’s on top of the diversity of their team on the backend, who lead the way in removing as much discriminatory bias as possible from how the technology itself is built.

“A machine doesn’t inherently weed out what it personally thinks is icky or inappropriate.”

They’re also using a unique tagging system that further reflects their sex-positive outlook that doesn’t shame or yuck people’s yum. It’s an inclusive rather than exclusive system. This means that instead of users inputting preferences that exclude others, everything is allowed on the main feed. Then, users curate it to what they like by labeling the type of posts they’d rather not see again. It’s an extension of the philosophy that everything that’s legal and doesn’t impede others’ abilities to express themselves should be allowed to exist on the main feed of Lips — even if a lot of people don’t care for it.

But Lips also knows that to accomplish its vision, it can’t just rely on machine learning moderation. It prioritizes a team of human experts with backgrounds in, for example, exploitation, who can further gut check disputed submissions.

Brown believes that, in some ways, algorithms can allow for a type of shame-free exploration of sex that real-life can’t.

“Sex is one of those things where it’s really hard for us to see outside our own perspective because it’s such a visceral thing. But a machine doesn’t inherently weed out what it personally thinks is icky or inappropriate,” she said. “A machine has more of an ‘if, then’ approach to sex that can be beneficial. But it depends on who’s creating the machine, what data you’re feeding it, and whether it’s vast and diverse enough to be a good algorithm.”

The algorithms for better sextech hardware

There’s an oversaturation of data-tracking and AI devices on the market that claim their tech optimizes or enhance sex through algorithms. Unfortunately, the vast majority of them use a lot of overstated tech-speak to sell people on impossible promises. On top of the same issues with implicit biases, sextech lacks good, representationally diverse scientific data as well as legitimate technologists.

Take the Autoblow A.I. The methodology behind its artificial intelligence are detailed in a not peer-reviewed paper with anonymous authors, spear-headed by one scientist who told Vice they had a “PhD on AI stuff” while the rest of the team was mostly engineers. The data was gathered by a marketing firm, which Autoblow A.I. worked with in the past on a vagina beauty contest. By watching POV porn videos of blowjobs, they translated the linear motions into data with a slider tool, which was then put into a neural network that developed the device’s haptic feedback patterns. 

Autoblow A.I. creator Brian Sloan told Mashable in an email that the data-gathering team used video sorted from most to least popular on Pornhub. The race of performers was not considered relevant. About 90 percent of the videos showed a woman giving a blowjob and 10 percent showed a man giving a blowjob. But Sloan noted that “even if there were differences between how gay or straight people performed blowjobs, the Autoblow’s penis gripper and motor system does not perform at such a high resolution that those differences would be felt by users.”

As for the diversity of the team behind the technology, Sloan said, “I hire by ability and cost; not by race or sexual orientation. I don’t think it’s appropriate to ask people who work for me about their race or sexual orientation because it is unrelated to their work and in general none of my business.”

It’s well known that porn (especially on Tube sites) overwhelmingly caters to cis, straight, white men. Regardless, porn is also a visual performance of sex, which probably isn’t representative of what a great blowjob feels like. According to , a software engineer in sextech who created , that’s one of the many reasons why the data set used for the Autoblow A.I. can’t deliver on its promise of being the “best blowjob machine in the world” that allow you to “enjoy blowjobs the same way they’re given in real life.”

“Sex is a complex emotional and sensory experience, and algorithms can only replicate vague traces of that experience,” Machulis said. “The hardware that the algorithms are controlling is usually nothing like a person someone would have sex with, and to match the marketing, the algorithms themselves have to distill all the desire and lust and scents and touches and sounds of sex into that crappy hardware. It’s a challenging scenario even for the best of engineering and research teams, and most sex toy companies have neither engineering nor research teams.”

Sloan countered by saying that, “It is obvious that oral sex includes other auditory, physical, and emotional sensations, but at least to us it is also obvious that the other characteristics are virtually impossible to measure, and even closer to impossible to reproduce in a physical product that costs a few hundred dollars… While every inventor including myself would love to include innumerable fantastical features in his invention, we are unfortunately limited to only those which are reasonably achievable at an economical cost. I have been granted a number of patents both on the machine itself and on methods for automatic porn synchronization using machine learning we plan to use for future iterations. That is to say we are pushing our field forward and pioneering the future of A.I. in sex toys.”

The underlying issues with the Autoblow A.I. speak to how a vast majority of sex devices incorporating algorithms rely on an interpretation of intercourse that ranges from rudimentary to dehumanizing.

Smart cock rings do this, too. Take, for example, the , which is marketed as a that collects biofeedback data during the act to learn what you and your partner like and give personalized suggestions for improvement. In practice, though, it just tracks the elapsed time of intercourse, speed, and depth of the cock ring-wearers thrusts — which, you know, isn’t very useful for much of anything, least of all the person receiving the penis. 

Or there’s a Fitbit for vaginal wellness, the , which promises “better bladder control, faster postnatal recovery, and enhanced intimacy.” It’s one of the most blatant examples of a gamified sexual wellness algorithm. You play a literal game where the strength and timing of your vaginal contractions basically acts like a controller. It can even detect if you’re doing the kegel wrong — yet fails to provide any sort of in-depth instruction on how to do it right. It also doesn’t account for how, often, vaginal pain during sex comes from the pelvic floor being too tense and constricted.

“A lot of these new products that measure all different kinds of physiological responses view sex only in the physical realm of pleasure. But sex isn’t just a physical need. There’s a psychological component,” said Sauer. “Data-focused products often fail to ask what is the purpose of sex? How do we want to feel about ourselves during sex? Because when sex is anything less than perfect, or somebody’s body isn’t meeting that algorithm’s expectations, the last thing you want is something measuring your performance level.”

“Our human nature is then to use that sexual data as a measurement against ourselves.”

Ultimately, a lot of sextech sells us on the promise of validation from the seeming objectivity and impartiality of an algorithm. But as we know, there’s no such thing as an unbiased algorithm. There’s also little data on sexual pleasure (especially for people with vaginas) that would even allow for an “objective” measure of pleasure.

“I do think that there’s value in that data,” said Sauer. “I just think that our human nature is then to use that sexual data as a measurement against ourselves.”

Perhaps no one knows that better — and how much thought and care needs to go into an ethically data-driven sextech device — than Klinger, the co-creator of Lioness. During its early development phase, three different potential cis, male, heterosexual investors suggested she create a leaderboard to measure which users had the “most” or “best” orgasms.

“I don’t even know how you’d measure that, or what that even means,” Klinger said. 

There is no ideal or “perfect” orgasm. The whole point of the Lioness, actually, was to help users understand how individual sexual arousal is: Some prefer orgasms with a longer lead up time, others prefer orgasms that last longer, while others like smaller successions of multiple orgasms. The point of the data is not to measure the “adequacy” of your pleasure, but to encourage curiosity over the multiplicity of ways we can experience pleasure.

So, instead of gamification, “We try to frame Lioness almost like a meditation app versus a goal-oriented exercise app. You’re not competing in meditation. It’s more about continuing the practice, having a journey, learning about your body’s experience, rather than trying to compete for faster, stronger, or better orgasms.”

I even used the Lioness in conjunction with precisely to do that, tracking different physiological and psychological responses through experiments in mindful sex.

As a woman who dealt with anorexia and body shaming growing up, it was essential to Klinger that the data never be presented in a way that inspired self-critique — or invalidated, judged, or even instructed users on how to have an “optimal” experience. Instead, the app mostly just gives you the hard data, and encourages you to fill out the journal to describe how you felt and how it relates to your interpretation of said data.

In an upcoming update, the team plans to provide classes in the app to help users learn how to better interpret their own data. But they never want the algorithm itself to impose any strict meaning or assumptions — despite the fact that it would probably be more profitable to do so.

What Klinger and the Lioness team have just launched, though, is a research platform where users can opt-in to fix one of the biggest issues that’s holding sextech algorithms back: all that rigorous, legitimate scientifically-gathered data on sexual function we don’t have. 

Not only is there a lack of research on the topic, but the data that does exist also has its own biases because of the lack of a diverse or big enough sample size. That’s why Lioness is also trying to get free devices to people who want to participate in these studies but can’t afford the high cost of most sextech. Two upcoming studies center around the while (another severely understudied topic).

What Lioness ultimately does through its data-driven algorithm is give users access to knowledge about themselves that they’ve been sorely robbed of.

“There’s so little places to get reliable, personalized information about your own sex life, your own body. And there’s few spaces to explore that comfortably,” Klinger said.

The marriage between algorithms and sex is in its infantile stages. But what’s clear is that no algorithm — no matter how smart — will ever remove the need for a human understanding of sexuality.