Half A Million Kinksters Can’t Be Fallacious—Asterisk

At a convention not too long ago, somebody pulled me apart and informed me they’d be down to supply me funding for my analysis. I used to be like, superior, as a result of for years folks have been yelling at me for doing surveys on the web.
For those who’re new to my work, I’m a self-taught researcher, pushed to be taught by lucky entry to excessive pattern sizes and sheer curiosity. As of this writing, over 700,000 folks have responded to my Big Kink Survey on fetishes and intercourse — probably one of many largest surveys on the subject ever achieved. It grew out of my need to reply two questions: How bizarre is my very own sexuality, precisely? and Why are there far more girls all for submission than males all for dominance?
I initially began doing surveys by way of porn (I additionally do porn), which is a tremendous straightforward option to get a whole lot of sexy social media followers prepared to do something you ask them to. I’d ask them to assist reply random questions I had, and shortly realized the info was horrible as a result of my questions have been horrible.
This began years of quickly iterated survey studying. I’d pump out tons of questions, take a look at the ensuing information on Google Sheets, after which rock forwards and backwards with agony that I hadn’t totally thought by how folks would possibly interpret query 5 and now all my correlations have been ineffective. I’d usually observe with questions in polls on X (previously generally known as Twitter), which received a lot of folks within the feedback reliably misinterpreting my phrasing in probably the most ingenious methods. I turned a calloused, exact, question-asking machine. My phrasing might cut up any hair, and I began failing much less and fewer steadily. Now, my crying-in-corner-agony per query fee might be lower than 1%.
It seems that, if you get information, it’s a must to work out tips on how to interpret it. Since I’m a homeschooler with no faculty diploma, I began wrangling good folks to inform me how the hell a correlation labored, what p-hacking was, and tips on how to do issue evaluation. I’m a sluggish however thorough learner, and finally received to the purpose the place I used to be fairly stable with easy statistics.
However my datasets have been getting big, as a result of my survey design was getting higher at seducing folks into taking them. My surveys, initially getting 10,000 responses, not too long ago began to blow up into the a whole lot of hundreds. You possibly can’t course of big information in Google Sheets, so beginning in 2021 I begrudgingly discovered tips on how to program in Python.
My viewers expanded previous simply porn customers into different, often nerd-adjacent, platforms. Typically my surveys would go viral elsewhere, bringing in novel audiences — not too long ago, principally younger, feminine social media customers, totally on TikTok. I began shopping for randomized samples from the research recruiting web site Positly to match in opposition to my very own, to attempt to perceive the diploma to which the respondents I drew from social media is likely to be unrepresentative.
I’ve been saturated on this planet of rationalists — an mental neighborhood shaped across the group weblog LessWrong — for eight years or so, and from watching them viciously tear aside different research, I’d gained a way of errors to not make: Don’t management for too many or too few issues, do not belief your intuitions on how stats ought to work, don’t keep away from reporting issues that contradict your outcomes, don’t examine for brand spanking new methods to analyze information simply because it didn’t match your expectations, and report the null findings. I received odds-ratio-pilled and used these over p-values once I might; I revealed uncooked information; I shared my code; and I attempted to maintain my evaluation and interpretation separate.

Ana Galvañ
Lastly, I’d write up my findings and put up them on-line. I’m not an professional — there’s a ton of stats stuff I don’t know — however I attempted to maintain it easy and simply say what I discovered.
It seems that lots of people actually hate this. Individuals say I’m not reliable as a result of I’m a intercourse employee and thus biased, as a result of I don’t have a consultant pattern, as a result of I don’t have credentials. My work isn’t peer reviewed, I haven’t weighted the info to make sure it’s consultant, and I’m an “OnlyFans lady simply polling her sexy viewers.”
So when the man on the convention provided funding, I used to be excited to do official, actual work. I’d dreamed that someday, if I had cash, I’d do a proper, grown-up research, with verified members. It’d be dependable and like the best way professors do it and make all of the offended web folks embarrassed for ever doubting me. I fantasized about repentant emails drizzling into my inbox. “You have been proper all alongside,” they might say. “The foolish on-line correlations you discovered do the truth is match the official severe enterprise ones too! You’re an actual scientist and we belief you.”
However as I went about the remainder of the convention, attempting to plan how I’d do my analysis respectably, it slowly dawned on me that … there truly wasn’t a clearly higher choice to recruit folks to speak about kinks. Every potential research idea I labored by suffered from equal, if not larger, vulnerabilities than mine did.
For instance, let’s say you need to research people who find themselves actually into vore, a fetish round swallowing somebody entire or being swallowed entire, sometimes by a a lot bigger creature. Perhaps you need to discover out if it correlates with issues like childhood abuse or psychological sickness — widespread theories folks have concerning the origin of fetishes. Let’s say you need round 300 folks to examine.
Firstly, how do you discover these folks? Vore, like most fetishes, is uncommon. For those who wished to do a survey the place you randomly ask folks about their childhood, and likewise in the event that they’re into vore, you’d should ask hundreds to get the 300 individuals who mark “sure.” So how do you discover your 300 body-swallowing lovers? You most likely want to take a look at some boards for vore lovers.
Already, we’re introducing a nonrandom pattern — we’re selecting from folks energetic on a vore discussion board. Perhaps these individuals are much less ashamed about their fetish, possibly they’re lonelier in actual life, possibly they’re rather more into vore than the precise inhabitants of individuals into vore. These individuals are most likely uncommon in rather more particular methods than the class of “folks simply shopping TikTok,” for instance.
This concern plagues most current sexual fetish analysis. Pedophilia is probably the most excessive instance. In a society with such large penalties for even suspected pedophiles, it’s apparent that any pedophile-heavy supply you go to goes to be massively affected by choice bias — e.g. intercourse offender registries. One study recruited pedophiles from an “outpatient facility for sexual offenders” and — shockingly — discovered that they’d compulsive behaviors and anti-social persona traits.
However again to our comparatively harmless vore lovers. Nonetheless we discover them, we have to incentivize them to take part. You possibly can pay members in cash or reward playing cards or, in case you’re a professor, with grades. This would possibly work. On paid survey-taking websites, you’ll be able to attempt to enhance the standard of your pattern by taking a look at issues like how briskly it takes them to do the survey, are their solutions constant, have they got top quality scores by earlier researchers, and so forth., However you can’t filter by “Do they like vore?” — that’s not a default demographic that the majority companies present.
Let’s say you may have a survey that takes a half-hour, which suggests you is likely to be paying $4 a survey. If you would like 300 people who find themselves into vore, and let’s say ~4% of the inhabitants likes vore, this implies you’re going to wish to pay roughly 7,500 folks, which makes this a $30,000 research. And consider you don’t know the vore prevalence beforehand, as a result of this information isn’t accessible in conventional journals, as a result of no person has achieved a high-powered common fetish prevalence research earlier than. Besides me, which is why I do know that quantity.
So let’s restrict ourselves to confirmed vore fetishists. In that case, what’s to cease folks from filling out the survey dishonestly as a way to get the cash? Perhaps they wouldn’t, however “research achieved on folks attempting to get $20” is a generally mourned drawback in analysis, and it appears affordable that area of interest fetish lovers are even extra inclined to be dishonest than regular in the event that they imagine researchers are going to be scrutinizing the outcomes.
And we would like this to be a real-life survey anyway. To do that, we have to get our vore buddies into an workplace, confirm their identification, after which have them fill out a survey in alternate for cost.
Uh, that is horrible. How might we presumably belief any outcomes we get from this?
Issues like “How do you discover a not-super-selected supply, confirm nontrolls are taking the survey, and get them to be trustworthy?” exist in a whole lot of fields, however they’re particularly tough in sexual fetish analysis. The complete factor is finding out uncommon traits that individuals are incentivized to lie about!
I not too long ago seemed on the the ratio between the frequency of fetishes in my pattern and the diploma to which individuals have been conscious of these fetishes in others. The least seen merchandise I measured was BDSM. And the more uncommon the fetish, the more taboo, and the larger the stress in opposition to sharing it.
Genital response measuring is an possibility — however we’d actually be deciding on for people who find themselves okay with an invasive measuring course of, the place your genitals are exhibiting precisely how turned on you might be by, e.g., an anime of a younger boy screaming as he slides down the huge gullet of a half-lizard girl in a microkini. There are advantages to this — it is likely to be good for checking correlations throughout a lot of fetishes — however there’s no approach this pattern will be random.
So, given all this problem, it is smart that the majority fetish analysis includes the situation I described above — going to numerous fetish communities to ask them questions, often within the type of visiting a discussion board or membership devoted to this fetish and posting a survey there. Contemplate the next pattern of research revealed between 1980 and 2020:
Sexual Fetishes: Sensations, Perceptions, and Correlates (316 individuals who took an nameless web survey posted in fetish communities)
Personality Characteristics of Fetishists, Transvestites and Sadomasochists (543 individuals who took this questionnaire distributed in particular person to fetish golf equipment)
Forbidden Games: The Construction of Sexuality and Sexual Pleasure by BDSM ‘Players’ (343 who responded to requests for interviews on Yahoo, sexuality, and BDSM communities)
BDSM: A Subcultural Analysis of Sacrifices and Delights (73 folks, from a put up to a BDSM e-mail dialogue group)
BDSM Role Fluidity: A Mixed-Methods Approach to Investigating Switches Within Dominant/Submissive Binaries (202 folks, from posting to web fetish communities and flyers in in-person fetish meetups)
The Psychology of Gay Men’s Cuckolding Fantasies (580 individuals who responded to an web advert asking about homosexual males’s cuckolding fantasies, posted by — I presume David Ley’s — social media channels and Dan Savage’s podcast)
If the Shoe Fits: Exploring Male Homosexual Foot Fetishism (262 respondents from a corporation for foot fetishists)
To be clear, I nonetheless suppose these research are respectable! I don’t imply to say a research is bunk simply because the researchers behind it have restricted entry to populations – however the technique appears ripe for unintentional correlations.
Plenty of current analysis (possibly because of the necessity of utilizing energetic neighborhood teams to get your pattern) additionally focuses on motion versus fantasy, which conceivably biases the pattern. A number of issues unrelated to a fetish most likely have an effect on whether or not somebody acts on it. Are they in a liberal tradition that celebrates their love of sitting on desserts? Are energetic cake sitters extra unpleasant than those that wistfully dream about someday icing their very own buttocks? Much less afraid? Then we’d anticipate any research achieved on cake sitting to seek out the cake-sitter neighborhood to be extra liberal, unpleasant, and daring than the true inhabitants of individuals aroused by cake sitting.
Given all this problem, the best kind right here might be an nameless survey the place individuals are correctly incentivized to be trustworthy — and what does this higher than giving them data again about themselves? It’s good to give folks a purpose, and that purpose ought to be as carefully aligned with reporting the reality as you can also make it. Telling folks what you’ve came upon about them based mostly on that survey appeared like an apparent reply.
Principally: an elevated tackle the traditional BuzzFeed quiz, “What Princess Are You?” In fact this isn’t excellent — one can simply think about somebody intentionally placing in odd solutions to see what this does to the outcomes — however I feel this nonetheless beats out different strategies of analysis for many functions.
Plus, it’s straightforward to disincentivize folks spamming humorous solutions into your kind for enjoyable — simply make your survey actually friggin’ lengthy.
The Actually Friggin’ Lengthy Survey
I’d gotten a yr’s price of funding from the great Invisible College in 2021, and so I cracked my knuckles, stop OnlyFans, and settled all the way down to work.
I wished to make a survey concerning the origin of sexual fetishes as a result of we all know virtually nothing about it, which might be as a result of that is actually arduous to do, for the explanations I’ve outlined above. Fortunately I didn’t totally recognize the issue once I began on this drawback, or I would by no means have tried.
My objective was easy — I’d decide a fetish, ask if folks have been into it, after which ask them a bunch of questions on childhood. This is able to permit me to find out, for instance, in case you received spanked in childhood, did that correlate with curiosity in being spanked as an grownup? This was fairly straightforward for a slender fetish like this, however I wished an enormous pattern for the explanations I defined above — looking out particularly for folks all for spanking appeared like it will do the entire “skewing outcomes ’trigger you’re drawing from the spank-specific neighborhood” factor, so I wanted to get a bunch of individuals to reply for another purpose after which occur to ask them about spanking. This meant a extremely massive pattern.
And my entry to massive samples was restricted. My viewers — probably the most energetic of which was round 150,000 followers on X, although my whole was round one million distributed throughout numerous platforms — would solely take so many surveys, and I didn’t need to ask the identical background questions time and again for every new fetish survey. It could be a lot extra environment friendly to mix all the fetishes into one massive survey!
And whereas we have been there, we must always throw in additional questions — what about porn use, intercourse fee, BMI, IQ, Huge 5 persona traits? You possibly can examine correlations for every further query with every thing else, so the extra questions you add, the extra precious the info turns into. As my checklist of questions elevated, the extra compelling it turned so as to add on one more query, and this was a nightmare snowball.
Ultimately, I included 61 demographic, childhood, and different questions; 35 “common intercourse and porn” questions, and 206 fetish questions, for a complete of 302 questions. However many of those have been checkbox lists, the place I requested folks to “mark all that utilized.” For those who rely these, the entire quantity of things I measured climbed to round 1,000.
I don’t know if you know the way survey-taking norms work, however attempting to get somebody to reply 1,000 objects is totally unhinged. It’s like asking somebody to fulfill for espresso after which forcing them to remain for 12 hours of small speak. And the ultimate cherry on the highest of this sundae of horror was that the dimensions of pattern you have to make findings vital within the conventional sense will increase with the variety of questions you’re asking (or, extra particularly, correlations you’re checking). So I wanted an enormous pattern measurement — many hundreds, a minimum of. However how do you get many hundreds of individuals to sit down down and reply a thousand questions?
It is dependent upon the kind of query, however on common I’ve discovered folks take round two to 6 seconds to reply a survey query, which meant it will take roughly between 30 and 90 minutes to finish my survey.
I might pay folks to take the survey, however that may get actually costly. If I have been paying folks $8 per hour to do that, and the survey optimistically took half-hour to do, then 10,000 responses would value me $40,000.
I needed to get inventive.
First off was shortening the period of time to take the survey, which sounds easy however was agonizing. I couldn’t let go of any of my treasured questions. Every query I thought of slicing meant I used to be releasing the entire different correlations I requested about into the wind. I felt like a hoarder on a TV present, wailing as I watched Marie Kondo slowly strategy my entrance garden.
However I might condense them. After a lot of experimentation, constructing and testing out numerous survey variations, I settled upon a gated technique. I’d ask folks the primary two classes — private stuff and common intercourse stuff — like a traditional survey, however the fetish half could be gated. I’d checklist a bunch of vaguely described fetish classes, like “Humiliation (defilement, impotence, cuckoldry, ridicule, and so forth.)” or “Transformations (progress/shrinking, bodyswapping, furries, and so forth.),” ask them to examine off those they thought would possibly apply, after which I’d feed them related questions from the classes that they checked.
However this meant I wanted to outline fetish classes, and there’s no good analysis on this wherever. I needed to invent them. So I went and scraped the web for all of the fetish lists I might discover, compiled them right into a spreadsheet, and painstakingly tagged all of them. There have been practically 900 fetishes, and I needed to individually analysis most of them by wading into fanatic boards as a way to know tips on how to tag them. I as an individual actually lack willpower (or self-discipline usually), and so right here the one factor that stored me going was intentionally, frothily fantasizing about getting all the info in the long run.
This nonetheless wasn’t ultimate. My private categorization of those fetishes was going to artificially cluster the info. For those who clicked on, say, the “clothes” class, you’d be uncovered to extra clothing-specific questions, which could make you extra prone to mark “sure” to stuff you’re solely mildly all for, whereas somebody who didn’t click on that class wouldn’t see these questions in any respect.
And I wasn’t that assured in my fetish clustering. What if I put “cake sitting” into the “humiliation” class as a substitute of the “sensory” one, nevertheless it turned out cake sitters didn’t care in any respect about humiliation? Regardless of my rigorous analysis, there’s no approach I used to be going to have a 0% error fee.
To assist compensate for this, I picked just a few of the gated fetish questions and repeated them exterior the gate, so I might examine the solutions between the entire survey inhabitants and individuals who expressed an curiosity in a particular cluster to see simply how a lot my survey design was affecting responses. This nonetheless wasn’t excellent, although!
The perfect query format was a Likert scale — “I discover gentle bondage to be:” on a “not erotic” to “extraordinarily erotic” zero-to-five scale. With this data, I might assist differentiate between responses like “meh I assume” and “that is important to my sexual perform.” However scales take some time to finish, and so this ended up as a sacrifice on the altar of time sensitivity. I reserved Likert scale questions for common classes, like “garments,” after which condensed most of the subcategories into more and more nested checkboxes. I get to see how a lot you want the final class, after which binary responses to objects inside that class. For those who like garments, does that imply particular clothes textures, states, or objects? If textures, is it latex? Leather-based?
The complete course of was a collection of tiny, painstaking tradeoffs. Every time I made a decision to make a query right into a scale, this boosted the entire time it took to finish the survey by just a few seconds. And I needed to care about these few seconds.
Ultimately, the common time it took customers to finish the survey was 40 minutes. That is actually lengthy for a survey! You actually should get folks to care about it as a way to end.
So: How do I get them to care?
I made a decision to present customers a rating on the finish — How freaky are you in comparison with others?
This was watery-eyed frustration all by itself, as a result of how do you measure freakiness? Certain, you could possibly do a survey to get folks to fee how taboo numerous sexual issues have been — and that is exactly what I did — however this nonetheless leaves the issue of weighting. Necrophilia is basically freaky, positive, however what if somebody is just barely into it? Is that freakier than somebody who’s extraordinarily into voyeurism? After which there’s the truth that freakiness can be a perform of gender norms. For those who’re a extremely submissive man, that’s actually culturally freakier than being a dominant man, so I needed to calculate totally different scores independently for women and men.
Scoring the survey took virtually as a lot time as truly constructing it did.
The final half was straightforward. I assembled a listing of fictional and well-known characters — like SpongeBob, Hitler, and Bambi — and requested folks to rank them on how taboo they assumed their sexuality was. I then matched them up, based mostly on these rankings, to the scores that folks received within the Huge Kink Survey. I personally received Home, from that TV present some years again, or roughly Eighty fifth-percentile kinky. The character that folks assigned the very best kinkiness rating was Willy Wonka.
As of my penning this, the survey has round 550,000-700,000 respondents, relying on how aggressively I’m cleansing the info. Most respondents are of their early 20s, and round 70% of respondents are girls — a mirrored image of the demographics that the majority love taking web surveys (as verified by my good friend who runs one of many oldest, greatest personality-testing web sites on the web).
One beauty of my pattern is that I requested about a whole lot of area of interest fetishes with out having to enter a single niche-specific neighborhood and ask its members to take my survey. The survey outcomes for the fetishes I’m asking about are as near an unbiased pattern as we are able to moderately get given you’re attempting to get an enormous pattern with low value.
I feel details about your persona is likely one of the absolute best incentives to finish a survey. Sure, your solutions are most likely warped a bit by your self-perception, however a minimum of it’s an incentive instantly associated to you, versus attempting to receives a commission.
And the size of the survey most likely not directly helped accuracy right here too — once I was youthful and took persona assessments for enjoyable, I’d retake them to see what altering little solutions did to my rating — however that is actually annoying to do if the survey is 40 minutes lengthy! This additionally made it resilient to intentional brigading or trolling — it’s merely tough to sink that a lot time into warping the outcomes, particularly to the diploma the place it will present up on my last analyses, and particularly with sufficient cleverness to get round all of the survey checks I included to catch unserious responses. And when taking a look at survey outcomes, I solely embody information from individuals who’d accomplished a majority of the survey.
The information I’ve gathered up to now appears to match up fairly effectively with current information, a minimum of for well-studied variables. As an example, my top and weight numbers correlate at round the usual 0.5, the speed of pedophilia in males is 3%, and the hole in curiosity between ladies and men in BDSM is clearly represented.
I don’t know why academia hasn’t tried to give you extra modern options for gathering giant quantities of survey information. Researchers are already doing web surveys, so why not sink some effort into advertising? Rent some influencers? Get on podcasts? I don’t know. I deal with my surveys not solely as a automobile to assemble information, but additionally as merchandise in themselves — to promote your survey, it’s a must to make respondents take pleasure in it. It’s important to make them like providing you with their time. It looks as if present academia — or a minimum of the intercourse half — views taking surveys as one thing like a chore, or like faculty. I’m unsure they’ve thought of making it enjoyable.
I don’t imply sacrifice the standard of the survey itself. I used to be painstakingly dry with the wording of the questions, as a result of I didn’t need a cheeky tone to prime respondents to present cheeky solutions — however outcomes on the finish are a clear option to make it enjoyable. I didn’t have to inform folks they have been the sexual equal of SpongeBob or no matter on the finish of this survey, however I did — and this brought about them to share it with buddies and on TikTok.
I come from a background of promoting intercourse work, the place I needed to claw my approach up the web flesh by sheer “how do I make folks need to ship this factor to their buddies.” And that is the attention with which I’m approaching analysis. I’ve achieved most of this with out funding, but have gotten samples that, in case you paid take a look at takers to do, would value upwards of $4 million.
I additionally make my analysis a neighborhood effort — not solely do I share my uncooked information and code, I often crowdsource questions from the general public about what to review subsequent. What hypotheses do folks have that they need examined? I do drafts of survey questions in X polls, to see how commenters will inevitably misread my wording and thus inform me on tips on how to write the query extra clearly sooner or later. I hope this course of helps vanquish the sacredness of analysis.
And extra particularly, I hope it normalizes folks with fetishes. I really feel such care and compassion for folks strolling round with these unusual arousal patterns of their head that always trigger such alienation. They’re shunned or ignored socially, but additionally by researchers — due to the logistical problem, as a result of institutional evaluation boards make approval arduous, as a result of sexuality is a topic rife with potential triggers, or as a result of folks merely don’t need to examine issues that aren’t stylish or socially sympathetic. I’ve received half one million information factors, with a person behind each, and I’ll preserve attempting to know all of them.