#as sampling biases go

[hauntoblogical]

I have seen three different "how much of this site is queer" polls this evening poking fun at staff's 1/4 assertion and not a single indication any of them understand statistics. One someone even replied "This might have some sampling bias" what do you mean might what do you mean SOME

[kvothe-kingkiller]

[bioethicists]

polls are fun + maybe i just have research poisoning but every time i see someone who appears to be earnestly trying to get generalizable data from their tumblr polls ("everyone reblog this but don't tag it or you'll ruin the sample!") i'm just like hhhh... no....

[lousycapy]

How statistics can easily be manipulated to fit a certain agenda in Formula 1

Hello! I’ve seen a lot of f1 fans or media sources bring up statistics incorrectly to prove their points recently, which is really getting on my nerves. So here is a comprehensive guide, with examples, of how statistics work and why they are not the be-all and end-all some people might think them to be. This is a pretty long post, so the explanations are all below the cut. With that, I hope you find this useful!

Multiple factors come into play when analyzing a statistic, so I’ve separated them in different categories: what data set is used to make the stat, how the stat can be interpreted and how being factually correct doesn’t equate a valid argument :

THE DATA SET

To make a statistic, you first need values which correspond to a data set. What said data set is made of is very relevant to the exactitude of the stat and how much regard should be given to it.

For example, to determine the average lap times of a driver over a stint you would need to divide the sum of all lap times by the number of laps executed. Which means that theoretically you could use a single lap as an average, e.g. 1:57:325/1 which gives an average lap time of 1:57:325s.

However, as you might imagine this stat is not representative of a driver’s stint, since the lap chosen to be analyzed could very well be an outlier. That’s why sample sizes matter, the more values make up your data set, the more representative of reality the result obtained is.

It is also important to know what the data set consists of. Let’s reuse our average lap times of a driver over a stint example, are outlaps/inlaps included? Is it based on clean air, dirty air? Are there laps excluded due to driver mistakes (e.g. going off track)? A stat being presented without any explanation of how it was calculated is absolutely worthless.

Finally, comprehension of the data set is very valuable as well.

Let’s imagine this fictional scenario where Ferrari makes Charles and Carlos compare average lap times. They both use the same car, on the same track, on the same tires, at the same time, for a stint of a total of 10 laps. Both drivers average a lap time of 80.125s over their whole stint, so is the conclusion that they have both done the exact same thing accurate? No!

Despite having the same average lap times in this scenario, the data set suggests a different conclusion, and different trends. Considering stints in a race are going to be longer than 10 laps, it can be assumed that Charles would average better lap times thanks to his consistency compared to Carlos, who would get worse lap times as time passes as can be observed thanks to the trend line in his graph.

2. INTERPRETATION OF THE DATA

Now that we can recognize the importance of the data set and its constituents, it is time to understand how the data provided can be used to make a statistic.

More than one answer can be correct based on the same sample of data. Despite using the same set, depending on how the data is used it can lead to different statistics that drive different arguments both being factually correct.

For example, I’d like to refer to the wonderful basspro24chevy World’s Destructor Chamionship from Brazil 24 on Reddit in an effort of determining who is the most destructive driver. Here is a chart I’ve made which also includes number of races each driver took part in (Ollie not included I was too lazy to recalculate how it affects the drivers he’s replaced’s damage bill) and the average cost of damage per race of each driver.

Based on these statistics, both arguments could be made to justify either Checo being the most destructive driver, since he’s the one who’s cost his team the most damage over the whole season, or Franco, since he’s the one who on average costs the most for his team per weekend.

Depending on someone’s biases, they could make some drivers look better than others despite using the same data set as another person, and depending on how their argument is justified even if they end up with a different conclusion it doesn’t mean they aren’t right as well.

3. FACTUALLY CORRECT ≠ VALID ARGUMENT

Even if you are factually correct with your statistic’s interpretation, and it is based on an acceptable data set, it doesn’t mean it has a direct link of causality with your argument and provides validation to the point you are trying to make.

For example, someone could argue that Checo is a safer pair of hands in races than Pierre, because over the course of the 2024 season he has DNFed 2 out of 21 races, meanwhile Pierre has DNFed 3 out of 21 races. However, the point being argued here is which driver is a safer pair of hands, and other variables than the drivers come into play when discussing those two’s DNFs.

Indeed, Checo drives a RedBull with a Honda engine, whilst Pierre drives an Alpine with a Renault engine. Out of Checo’s 2 DNFs, 2 were caused by driver mistakes. Out of Pierre’s 3 DNFs, 3 were caused by engine issues. The World’s Destructor Championship can also be used as a counterpoint to Checo being a safer pair of hands than Pierre by comparing damage bills.

Thus, instead of the conclusion being that Checo is a safer pair of hands than Pierre, the DNFs statistic is more appropriate to conclude that the Honda engine is more reliable than the Renault engine.

Which means that to make a valid argument, you need to be able to explain why the statistic presented is relevant and what it suggests. Alleviating circonstances also need to be taken into account to solidify the point being made.

For example, let’s imagine a scenario where Fernando is 1.235s off Lance during a qualifying run. To use this stat in an argument, you need to be able to justify why he was so far off. Was it genuine pace? Did he make a mistake which ruined the lap? Were they on the same tires? Was it track evolution? Are they on the same setup? Did Fernando come across traffic? Did Lance get a significant tow?

Contextualization matters twice as much as the actual statistic being presented, because the statistic without context can easily be manipulated in a way to drive a certain agenda.

4. CONCLUSION

All in all, what I’m trying to say is that even maths can be used to drive agendas. Statistics can not be taken at face value, because there are multiple factors that can influence their relevance. I hope you found this little guide helpful, and that it will help you analyze better the information you see online on how drivers are performing (or argue better with crazed fans, you do you 🫡)

Thanks for reading and have a good day!

[saxamophone]

reading, lately

Go give all these a look, they're phenomenal.

Eat Your Death, Draco by @sillywives T | 3k

What if Draco was really excited to join the Death Eaters but then found out they eat death? Literally. On a plate. Draco being Draco is hilarious in this one. (Written for @hp-abandonshipfest)

Seven Year Itch by @mintawasalreadytaken E | 5.6k

Please read @vukovich's prompt for this at the start of the fic, and then oh my god. What minta does with it is so, so good. Ugh! I don't want to say much because I don't want to spoil it. But it will hurt so good. (Written for @hd-hurtcomfort-fest)

Crush by @citrusses E | 8k

Harry's a mess but someone has a crush on him. What's the best way to ensure he can keep their attention? Asking Draco Malfoy for help, of course! Gosh, I love this. Funny and sweet and sexy. I love everything citrusses writes.

When the Flood Comes by @academicdisasterfic E | 10.3k

Draco is a lawyer, back in England to stop oil drilling legislation in the North Sea. Harry is a stoic Auror, set on seeing something through. There's political manouvering, powerful women, and men who are terrible, as always. But Harry's one of the good ones, as always. I love this fic sm, and maybe I'm biased (I am, but I'm also right) but you should read everything by academicdisaster. (Written for @hd-wireless)

Rookie Moves by peu_a_peu E | 75k

Harry and Draco are Auror partners. Harry attempts earnestness, while Draco remains Draco ('Get fucked, loser,' is an amazing response to when someone tries to shake your hand.) Hilarity ensues. Ron and Hermione blame themselves. Need a sample? @littlewinnow drew an amazing rendition of the above-mentioned 'Get fucked, loser.'

Also not a rec, per se (because it's widely known) but my comfort fic lately is Draco Malfoy and the Mortifying Ordeal of Being in Love. Hilarious, smart, and full of the best romcom tropes.

[pineapple-frostyfruits]

May I be annoying and add on, transjinx?

... Better to beg for forgiveness than to ask permission-

So:

I was tagged by @theangelyouknew

Put your 4 favorite characters from 4 pieces of media as options and let your Tumblr pals decide which one most suits your vibe then tag 4 people

npt: @lombolica @idkaguyorsomething @wibbly-wobbly-blog @sugarplumanderson

[nkjemisin]

Hey there. I'm writing a story set in New York City and am not American. I have few characters, but most of them are arab or white. I can't help but feel a bit wrong about it, given that America is much more diverse than that, and NYC being an emblem of that. Do you think I should force myself to include more representation or should I just tell my story, and leave that more diverse cast to some other story I could write? I know this is a neverending debate and there are many opinions about it, but I've always agreed with everything you've said in matters of representation in fiction, and so I'd be curious to know your personal answer on it.

I'm a little confused by how you're using "representation," here. It sounds like you think representation = "randomly sticking BIPOC everywhere." I think when most people use that word, it means something more like "create an accurate or at least plausible depiction of a group or place." In actual New York, there are plenty of Middle Easterners and white people who live in relatively homogeneous small communities where they might only see someone of a different ethnicity on the subway. If your story is set in one of those communities -- and you do stick some random BIPOC in that subway scene, because that's plausible -- then it sounds like your characters might be an example of good representation.

(Note: if you're not writing something set in the real world, but it features human beings, it needs to represent humanity as a whole, unless there's a good in-world reason not to. But if it's our world? You can get specific.)

Here's the catch, tho: plausibility is relative. If you've absorbed some biases and haven't done enough research, then you might end up writing something that feels plausible to you, but which isn't actually representative or plausible to anyone else. The way to avoid this is to do the research and check (to the best of your ability) your biases. For example, you aren't American, I assume you've at least visited NYC? If not, you should. You can visit some of the communities I mentioned! You can eat in restaurants, visit mosques, have conversations with actual real people who are living the life you're writing about! If you don't have the time, money, or spoons to do that, there are other ways to do good research -- films and YT/Tiktok videos made by people from the communities in question, for example. But you'd need to watch a lot of them to get a good representative sample.

I recommend this book to all the writing students I've taught at Clarion, and other writer workshops: Writing the Other, by Nisi Shawl and Cynthia Ward. There's a particular part of it that seems relevant here, which is a kind of hierarchy of "appropriate" appropriation, I think first mentioned by Diantha Day Sprouse but included in Writing the Other. Basically it says that if you want to write about a culture that isn't your own, you can learn about that culture in one of several ways: a) You can be an Invader, and just go take whatever intellectual and artistic tidbits from that culture that you want, regardless of how damaging this might be to members of that group. Example: non-Indigenous people who write about actual secret practices, or who encourage the desecration of sacred places. b) You can be a Tourist, in which you're still mooching from that culture, but at least you're figuratively paying someone for it and accepting tidbits that the culture has chosen to sell. Example: getting a sensitivity reader. Or c) you can be an Invited Guest, who brings in as much as they take out, and who has formed relationships that are beneficial to all involved. Example: being part of an exchange program, both as a student and later as a host, and maintaining those friendships outside of the program.

The goal is to be an IG, but that isn't always possible. Tourist is still better than being an Invader. (...I feel like I'm leaving out a category. It's been a while since I read the book; any more recent readers want to check me here?) But the closer you can get to actually participating in that culture, the more your work will be informed by reality instead of biases or misinformation, and the more likely your work will read as plausible not just to you, but to your widest possible audience -- people familiar with the culture and people who aren't.

(I'm a little concerned about your phrasing of "force myself to include more representation," note. Why would that need to be a forced thing? A writer's goal should be to write something that feels lived-in and authentic to [if it's a real place] most people's experience -- not to meet some arbitrary standard, but because that's how you master immersion and characterization. If good immersion and characterization feel forced to you right now, that suggests you need more practice. I recommend writing short stories!)

[local-hellhound-steals-spaghetti]

This is an embarassingly bad paper based on the use of two aversive based dog trainers who are clearly incompetant in the use of +R - yet the e collar training community is taking a victory lap.

The fact that this got past an animal ethics review is honestly appalling and the university should be ashamed.

Let's go over the main huge flaws with this study:

The positive reinforcement protocol they are using to compare to e collars is laughably incompetant - with no management or use of distance, desensitisation. Just use a leash? Maybe?

Use of aversives to stop chasing is, unsurprisingly, effective. Because, yes, painful consequences tend to work to suppress behaviour. That isn't the problem that people have with e collars.

The author seems to think that use of e collars is justified because it "gets results" in a short amount of time to stop dogs chasing - again, apparently we haven't heard of a leash.

"Aside from presumably pain-induced yelps in the dogs with e-collars when they received shocks, none of the dogs in any groups showed any signs of stress or distress." So we're saying a pain vocalisation isn't a sign of distress? That's your welfare parameter? Jfc.

Love how they didn't even try to get fecal cortisol baselines and just collected samples from a few dogs but not all of them. You tried, I guess?

The fact that the trainers for the study is "Ivan and Ivan's Student" is ridiculously biased. Acting as if both these trainers know how to apply +R effectively as opposed to their heavy aversive tool use is laughable.

"We designed our study’s methods around current trainer practices as well as the recommendations of our senior trainer while attempting to keep continencies as simple as possible." - yeeeah that's a conflict of interest. Especially when your "senior trainer" has an axe to grind with R+ people using scientific papers to support their training methods.

You're chucking treats in a bowl after a few sessions and you think that's going to stop a dog chasing a lure? Oh yeah, you got us. We just throw kibble in a bowl to stop dogs running into traffic. I am genuinely floored that that's how they think a +R protocol for chasing works. They have no business comparing the two when they can't even do one properly.

"We also did not test whether the food rewards we deployed were highly valued." Okay so the majority of my client dogs will go nuts for my meaty high value treats because they've only had store bought dry treats and the client is shocked at how food motivated their dog suddenly is..... So this study was relying on whatever the guardian of the dog brought with them lmao rookie mistake.

"Dogs that chase livestock may be shot or otherwise euthanized, and dogs that chase cars may be killed in traffic." Put. Them. On. A. Leash.

"The practical applicability of either the LIMA or LIFE models is likely limited, at least in the United States where dog trainers are not required to be licensed or certified. In a sample of highly rated dog trainers, more than half did not document any relevant education." Does that include your "senior trainer"? It does. Because he doesn't have any formal education in canine behaviour, welfare and science. You know this. You chose him to be your trainer for your study.

"The speed and effectiveness with which the e-collar inhibited the dogs in this study from chasing may justify the limited number of painful stimulations the dogs experienced if the object the dogs were chasing had been something that could directly or indirectly cause them serious injury or death." Or..... OR.... you could not use electric shocks on your dog and potential cause long term damage, fear responses ect.... And use a leash.

Just baffled by this whole study. My high prey drive whippet will go feral for a lure but only when she knows the context for when she's being allowed to chase. She can be at a lure coursing trial for the whole day not worrying about the lure because 1) I keep her at a distance where the stimulus isn't saliant enough and 2) she gets reinforcement for calmly watching from a distance.

I'm honestly getting second hand embarassment reading this study. It's just such bad science should never have gotten past an ethics board.

[ballsballsbowls]

I've seen a lot of posts saying "don't follow the polls because the margin of error is basically useless."

And I've seen a lot of posts saying "don't follow the polls because they're being entirely run and paid for by the most partisan people imaginable running garbage data."

and I've even seen, "don't follow the polls because the sampling is really biased."

but I've not seen a single one going, "don't follow the polls because even in their extremely small and biased samples, some of these people just straight up fucking lie to them?"

I cannot be the only person in the entire United States who has extended family members who lie to political pollsters for funsies and then boast about it.

[miserye]

under the cut

Okay perfume talk, I wouldn’t consider myself to be a serious fraghead but I’ve been in this for a couple of years maybe 3 or 4 so I’m not a novice either atp,, Every holiday season Sephora and ulta release a perfume/cologne set of tester samples that you can also use to redeem a full or travel size of perfume which is a pretty good deal because getting perf samples is a pain in the ass esp if you’re looking for more elusive perfume, usually goes around $4-6 on average so getting this collection of say (for this year, it varies per year I believe the average is around 7-8 samples for the travel pack and usually something similar for the full/mid size set) 8 samples for a value of 34 is really good plus you get a travel size on top of that for free because again, if each of those samples you’re eyeing are even just 4 bucks a piece, that’s still 32 and you’re out a travel size so these are really good value deals.

Now here’s the issue. Every single year without fail, there are the sample fucking scents EVERY SINGLE TIME yeah yeah cult classics or whatever but it’s fucking boring and it lowers the value of the set because we already tried them the previous year and if you just want to try like say three scents, it’s better to just go in store try them and buy a normal travel size, which are typically from $28-34 depending on the brand but this is a typical average for travel sizes. It’s always the most BORING scents they choose to re-use because we’re stuck in like 2015 or something idk but usually you can see D&G’s light blue, herrera’s good girl, Valentino bright crystal, ysl black opium AND libre, Viktor & Rolf either flower bomb or good fortune. Here’s the other issue with these sets, 90% of these scents are strong vanilla or heavy sweet floral which is fine EXCEPT that’s most the set. It’s just not very versatile usually, and caters to a specific group of people who enjoy more overpowering scents because, again, that’s what perfume is known as typically. Kinda strong vanilla (specifically vanilla base) and usually a floral like rose or peony.

ALSO my other issue with these sets is that I KNOW the other perfume in them are ass because I’ve smelt them before. The other ones they add in are not ‘indie’ perfumes per say but they’re not super popular such as Juliette has a gun which HAS GOOD SCENTS BUT they ALWAYS ALWAYS ALWAYS put in not a gun which is a layering scent and is lame and is not for the average normal people consumer because most people don’t think to layer their perfume when they’re testing out new scents so I think it’s not great in these kinds of sample sets because people want to take each scent in as an individual.

Regardless, they always choose some of the most popular scents and some mid scents from other brands which are ‘trendy or people have their eyes on’ but typically smell mediocre if I haven’t mentioned that. That brings me to this year where they’ve really stepped it up. This year, they changed the typical line up to the ‘one off’ scents where a lot of big frag brands will create a ‘twist’ on their scents and sometimes it does make a difference, but sometimes not.

For example, black opium from ysl has like three iterations where there’s the classic, there’s one called over red, and there’s one called like illicit green or something WHCIH is not to be confused with edp, edc, and edt that’s different <- this is the strength of the fragrance concentration (side tangent, perfume is comprised of two main elements, the alc which diffuses the perfume and is the reason why when you first spray something it smells like alc and the actual fragrance oils, each of those have varying quantities of each, the strongest smell and amount of fragrance oil to alc concentration goes eau de parfum, followed by eau de cologne, then eau de toilette <- the toilette is a reference to the bathroom aka toilet, meaning you’re supposed to refresh this perfume when you go to the bathroom so it’s intentionally a perfume that doesn’t last).

Anyways. This year, the sample set from Sephora has more ‘new’ scents that step away from the cookie cutter cult classics and also have favored the new ‘twist’ perfumes on the actual classics, typically as a means of promotion for the new version of the scent which usually isn’t as well loved/recieved aka Herrera’s new version of good girl which is good girl blush. They also have included perfume oils which is really new and fresh because those have been increasing in trend recently and I really appreciate it. Anyways, there’s a new massive set for 95? That has 15 samples and includes a free MID sized perfume, it says full size but it’s actually the size in between full and travel, it’s usually around a two oz or so less which comes out to around 30-50 mL usually (depends on the brand, some do both 30 and 50 if they’re more expensive) compared to a full which is usually around 98sih mL. The line up was pretty good!! I was happy about the mix of older and newer scents that they chose, though I’m not happy about the appearance of Charlotte tilbury into the perfume world, I haven’t tried any but just looking at them they look cheap and gross in terms of scent. While the trend to cater towards sweet and floral hasn’t changed, I’m glad they started branching out of the usual mainstream scents and I really do think that even the smaller full and travel sets they came out this year are a lot more unique than usual.

In terms of ulta, they’re lacking a bit as per usual when it comes to these things, but they also included a new like diversity set so there’s one with all the cult classics and there’s a different one where most are relatively newer scents which I like, also the price point is crazy it’s like 17 and 14 scents (the sets have different # of scents, the classics is 17 the new one is 14) for 20 plus a 15 gift card so the ‘essentials’ they have is good for people getting into perfume and those that want the normal popular scents and the newer one are more relevant scents to 2022ish and while it’s still a sweet vanilla and floral set, I’m happier about the ones they chose to include. Though the thing with these sets is that there’s always a chance they’ll give you perfumes that are empty/half full just because this is something third parties put together. There’s less of a risk with regular sample sets that the brands put out because it’s from the brand, but Sephora and ulta have a history of having crappy sets with less than full bottles, however if you’re just wanting to sincerely try out a scent like once or twice, it’s whatever. Honestly I don’t need most of my samples past one use because I can tell when I like or hate something pretty easy. Anyways, that’s my thing about perfume LMFAO sorry this was long

do you guys want to see my literal essay i just sent to my coworker about perfume

[am-i-the-asshole-official]

AITA for creating and running a gimmick blog?

A few months back, I was bored and decided to hop on the trend of gimmick blogs. Since I don't have any crazy skills like identifying cars or programming bots, I settled on something I thought was extremely simple: correcting typos. So I'll sometimes reblog posts that have typos in them and comment with corrections. I would only do this on posts that were already lighthearted or joking in tone; I would never derail a serious post with it. And I really haven't used it very often - maybe a couple times a week at most, just when I happen to notice a good candidate.

Well, recently I was inundated out of nowhere by a bunch of anons telling me this was a horribly offensive idea. According to them, I was insulting dyslexic people, non-native English speakers, people without access to education, and a whole bunch of other groups with these unwarranted corrections.

I had honestly never considered that angle, and I've paused using that blog so I can try and reflect on it. But when I replied to a couple of the asks asking for a better explanation of exactly how this was harmful - because I genuinely wanted to be informed - the vast majority of the replies, with few exceptions, were obscenely rude to me. I've been called classist, ableist, racist, and a lot of much worse words I don't care to repeat here. I blocked exactly two people because they were being extremely hateful in my notes, while still trying to engage with the more polite ones, but of course I still got accused of blocking and ignoring everyone I disagreed with.

The truth is, I'm still not sure whether or not I disagree with any of them on the actual subject at hand; I just can't deal with people being bitter and rude and assuming the worst of me. I tried to make it clear that I was more than willing to listen and have a conversation in good faith, but that has proved impossible.

So now I'm really hurt and really, really confused. I'm not going to just blindly trust a small group of hypocrites on the internet who claim they're worried about people's feelings while at the same time trying to completely villainize me as if I don't have feelings too. But I also understand that they might have a point. Sadly, politeness is not always correlated with correctness.

I absolutely do not want to continue running this gimmick blog if it's truly harmful and offensive to people. I've just never encountered this take before, and it was delivered with such vitriol that I had to take a break from tumblr entirely just to recover my sanity. So I'm hoping a much broader and less biased sample size will help to clear this up. I know an AITA poll isn't perfect, but it should do.

If I get a YTA verdict, I will delete the typo-correcting blog and stop immediately, no questions asked. If not, I'll know I just angered an extremely vocal minority that has no idea how to deal with conflict.

AITA?

What are these acronyms?

[ladyshinga]

One of the biggest reasons I loathe "AI" things like its "art" and chat bots and shit? Because humans don't understand computers and y'all start thinking they're all-knowing GODS who can DO NO WRONG

Imagine a law bot is fed lots of past cases in order to determine punishments for a prisoner. I know we're not there yet, just bear with me for an example.

Thing is, all the past human judges were RACIST, and their punishments were disproportionate - white prisoners get lighter sentences, everyone else (especially black people) get worse.

An AI isn't gonna have morals or ethics. ALL IT CAN DO... because again, it's a COMPUTER PROGRAM and not an ACTUAL Artificial Intelligence... is read back over all the example cases it's been given to come up with something similar. It sees a very "white" name? It'll give out those same judgements because that's what it's sampling from.

Humans are programming this shit, and HUMANS have biases. Computers aren't smarter than us, they aren't wiser than us, they will make some one's bias so much worse BECAUSE other humans shrug and go "well a computer said it so it must be true" - it becomes much harder to argue a point when you're arguing against an algorithm and not a person with discernment and a real thought process.

Consider the TERF woman-only app called "giggle" that determined who was "allowed" on the app based on a selfie and an AI that could "read" some one's bone structure and "tell" if they were biologically female. Guess what! All the history of "bone structure" arguments for biological sex, racist science! Amazingly, it's WHITE cis women who had the easiest time getting on this app because the AI is ONLY basing it off of a CERTAIN subset of white women to determine WHAT femininity IS. And that's the whole history of male vs female "science", it's HEAVILY filled with white-focused traits that ultimately end up punishing, say, black women whose facial traits might not look like what the AI thinks is "feminine"

Stop trusting computer programs that were made by flawed humans. Stop thinking we're in this amazing future where this is real actual AI and we can trust these programs to be logical and non-biased. It's a fantasy

[youcouldmakealife]

KS Fill: Mike/Liam; happy place

For the always welcome prompt: Anything where Mike and Liam are happy

Teensy bit NSFW because, well, it's Mike and Liam.

The kitchen’s always been Mike’s favourite room of the house.

Well, Liam would like to think it’s his second favourite room, but it probably isn’t. It’s not like they keep shit confined to the bedroom or anything anyway. There isn’t a surface in this house that Liam hasn’t gotten bent over at some point, though Mike was very, very thorough about disinfecting the kitchen counter after. Totally anal about it.

“Why are you snickering to yourself,” Mike says.

It doesn’t even come out suspicious, he’s gotten into such a rhythm cutting little stars out of a sheet of cookie dough. He probably wouldn’t admit this even to himself, but while Mike may like to cook, he loves to bake. Only ever does it for others, though, like the baking itself isn’t unmanly but eating the results is.

That means Liam usually only gets to sample his baking one filched cookie at a time, which is bullshit. But also, probably good for his career. Mike’s just as good at baking as he is at cooking. Possibly even better, but Liam’s got a sweet tooth, so he’s probably biased.

“Liam,” Mike says, and Liam realises he hasn’t bothered to answer.

“Thought of a sex joke,” Liam says.

Mike snorts. “Of course you did,” he says. “Do I want to hear it?”

“You always do,” Liam says. Mike secretly thinks he’s hilarious.

“Go ahead,” Mike says, trying and failing to sound long-suffering. Cutting little stars is clearly his happy place. Liam’s going to go completely nuts online later buying cookie cutters. He wonders if they have little skulls and crossbones or rockets or something. Maybe Mike will let them keep a batch of cookies if they’re badass enough. Probably not, but a man can hope.

“Remember how we fucked in the kitchen?” Liam asks.

“Which time?” Mike asks.

Liam waves a hand. “Irrelevant to the joke.”

“Okay,” Mike says. “Yes, I remember.”

“So I was thinking about how you were, you know, super thorough cleaning the counters off after?” Liam asks.

“I prepare food on these counters,” Mike says. “They need to be—“

“Not saying you shouldn’t have been,” Liam says. “Just, you know. That was kind of anal of you.”

Mike doesn’t stop himself from snorting in time. “That was bad,” he says, a second too late. “Completely juvenile.”

“I know,” Liam says cheerfully. Made Mike snort, though, so what does that say about him?

Mike looks down at his handiwork, a sea of stars across the baking sheet, enough for every member of the roster.

“You’re making two batches, right?” Liam asks. It’s just the team for this one, no significant others or kids, so technically only one batch is needed, but one per person doesn’t work out well. There are thieves on that team. He should know: he’s one of them.

“Three,” Mike says. “Make sure there’s enough for the support staff.”

Once again, Mike is going to be the most popular person at the team Christmas party. Third straight year. Liam thinks they should make him a trophy or something.

“Can we decorate them to be North Stars?” Liam asks. “Or is that not Christmassy enough?”

“Why do you think I picked stars, Fitzgerald?” Mike asks, and Liam grins at him. “Why don’t you make yourself useful, find the green food colouring for me?”

Liam has no idea where the fuck it would be, but he has eyes and hands and enthusiasm, and the full awareness that Mike will gripe at him from across the room like a backseat searcher until he locates it.

“Why would it be in there?” Mike asks, as Liam opens the cupboards, and Liam grins at the mixing bowls. The griping’s already begun.

“Baking stuff,” Liam says.

Mike sighs, and Liam moves one over.

“Liam,” he says, and Liam grins wider. “Really?”

“Tell me where to go, and I’m there,” Liam says, but instead Mike pins him against the counter, basically surrounding him. Liam leans back into him, closes his eyes when Mike wraps an arm around his chest, brushes his lips against Liam’s temple, the kind of thing he’d deny was a kiss if asked. ‘What? Kiss you? My mouth just happened to be in the area, don’t flatter yourself’.

Liam tilts his head up, and he gets a proper kiss for that, though Mike pulls away just as he’s starting to get a crick in his neck. He always seems to know, somehow.

Mike presses another not-kiss to his cheekbone, then smacks his ass. “Go find me the food colouring,” he says.

It takes ten minutes, and Mike almost throws the cookie cutter at him, but eventually Liam locates it.

“Can I help make the frosting?” he asks.

“You just want to lick the beaters,” Mike says, but he doesn’t say no.

[oflgtfol]

SHUT DA FUCK UPPPPPPPPPP

sincerely,

all the queer women who work at my astronomy job

"i, as a woman, feel like i don't have autonomy in a patriarchal society. oh i know the solution - let's say my lack of autonomy comes from the fundamental fucking universe itself instead of just being a social construct!! that solves everything!!"

[tsspromptmonth]

Sleepy Bean Fanfic Café Pre-Orders Open 10/30!

Hey, Babes, get those comments going and your orders ready! The Sleepy Bean Fanfic Café is opening for pre-orders on Wednesday, October 30th.

We'll accept orders from Oct. 30th - Nov. 30th. Starting today send your Asks with your requested drink and screenshots of you comments!

More details below the cut

Menus

Sample Order

A few words on Comments

Have more questions? Hit us up!

[cybertied]

How I Make These: Part 1

General tips and learnings

There’s a limit of 30 images per post so I’m splitting this into 3 parts:

Part 1: General tips and learnings

Part 2: Sample prompt development

Part 3: Sample prompt development (continued)

First, a word of caution

If your prompts get blocked too much you can start to get account suspensions in increasing severity (ranging from 1 hour to 24 hours to indefinitely). If you care about your account, be careful with this. This guide runs through my process and learnings but I don’t take any responsibility if you run into any issues with what you do with it.

Intro

These are my learning and tips for making AI images of guys tied up in different bondage situations. If you’re reading this, I assume you’re into something similar. Everyone has their own preferences and this is biased towards mine but if you like something else, maybe you can use the tips here to generate the kind scenarios you’re interested in.

I’m constantly getting messages about how I make these and can’t spend so much time responding to each one individually anymore so hopefully this helps you or is maybe just interesting. I know this is rather long, but if you can’t handle going through this, you may not have the patience for dealing with Bing.

Also, if you’re from Bing and find this, please don’t use this info to make the system smarter and prevent my tips from working 😁. But please do see my note at the end about my feelings towards Bing’s content policies which basically shame certain lifestyles. 

Some quick FAQ responses

What platform do you use? I use Bing for everything. It’s the only generative AI platform I’ve messed with. The quality is good and the interface is easy but the content filter is really sensitive and incredibly inconsistent.

Will you make me an image of (specific scenario)? Likely, no. This stuff is personal and everyone has their preferences. It takes so much time and trial and error to get good images and you only get so many prompts each day that I don’t want to spend time on things I’m not interested in. But maybe these tips help you make your own!

What’s your prompt? I get asked this a lot without a specific image cited. I’d need a specific example to answer that. But I don’t save all my prompts and I’ve made thousands of images so it’s really tough to go back and find stuff. I can maybe try to give pointers but I probably can’t give out exact prompts.

How do you get past the content filters? That’s really tough to answer because it’s so contextual for each image (as you’ll see in my tutorial later) and it takes a ton of trial and error for each one. But the example process I show a bit later highlights a lot of the principles and my process. It really depends on what you’re trying to do but I’ve shared a lot of my main strategies here.

How the content filter (maybe) works

It’s important to understand how things get blocked so you can work through it. Here’s how the content filter seems to work after entering a prompt from what I can tell: 

If you immediately get a blocked message, you either used a keyword it doesn’t allow or your phrasing wasn’t accepted. Variants of the word “hogtie,” for example, frequently get blocked immediately. But sometimes it gets through if it’s within certain contexts so it doesn’t seem like an absolute rule where certain words are never allowed (I’m sure some words never are but I don’t get that explicit with stuff).

If the image starts to process for a bit and then it gets blocked with a message, it seems like a bit more analysis is performed and your phrasing was deemed bad in some way.

And finally, if you get the yellow image of a dog, I believe the images got generated but some final image analysis step was performed to check for questionable material (this comes up later in some of my tips) and blocked it.

On a related note, you’ll get 1-4 images from a prompt. I could be wrong but I think when you get fewer than 4 images, the omitted ones were blocked. And you get the dog if all were blocked. But this is just my assumption. They throttle traffic occasionally if too many people are using it so maybe fewer images get generated during peak times to save on computing usage. 

And like I said, it’s really inconsistent. Things get through with one nationality of rugby player but not another. Or a gag will work on a football prompt but not rugby. Or it’ll work one day and not the next. So it all takes a ton of trial and error. Even just changing the ordering of words can impact things going through or getting blocked. 

General tips

Be careful not to get blocked too much if you care about your account. You could get suspended for an hour, then 24 hours, and then permanently (which can be contested). If I’m getting blocked too much, I’ll just start doing safe prompts for a bit and then try again later.

I’d recommend starting with a prompt that creates your scene and character but without any bondage elements. Sometimes even physical attributes (like “shirtless” or “leather”) can get blocked so you can at least perfect the general look and remove variables to start isolating which words or phrasing causes blocks. If you write out a full long prompt and it gets blocked, you’ll have no idea which parts might be objectionable.

Once it generally looks how you want, slowly introduce whatever bondage elements you want one at a time. This allows you to see what gets through and what gets blocked.

Context really seems to matter. If you provide a reason for a guy being tied up, even if it’s ridiculous, it has a much better chance of going through. For example, “a guy is sitting on a wobbly chair so he’s tied to the chair to prevent him from falling off.” Or, “he’s sitting on a high rise patio so he’s tied to the chair to keep him safe from falling.” Or, “he broke the good gymnasium rules so the referee tied him up.” The reason can be nonsense but if it fits the context, it frequently works.

Be descriptive. The word “gag” I assume will always get blocked. But describing one frequently works. I have some tips for this later. 

Similarly, I’ve noticed that just “rope” gets blocked a lot. But adding descriptors that fit the context might get through such as “wrestling rope,” “athletic rope,” “tactical rope,”or even just “jute rope”. 

Consider workarounds. “Sweaty” or “wet t-shirt” might get blocked. But if you set the scene on a rainy field, a humid office, or a damp storage room, the result effectively will look the same. Or “wearing a speedo” might get blocked but if you create a situation where that’s likely such as “in the 1950s at a Palm Springs resort, a handsome guest sits by the pool” might result in a speedo image. (This is a hypothetical example but I had one like that work.)

If you’re getting that dreaded yellow dog image, sometimes crazier image effects can help get it through. Like I wrote earlier, I believe this occurs when images are generated but evaluated one last time for content. I’ve found that if an image has more visual “noise” going on, it can be harder for the system to detect issues. Dramatic shadows, water droplets, lens flares, vibrant high contrast lighting, etc., all seem to help make images more likely to render. Unfortunately I don’t always like the look of these things but it seems to help and I’ll take it.

Getting good hogtie images is probably the most challenging prompts I’ve tried and I only occasionally get lucky with it. Even when it accepts the word “hogtie,” the guy is usually not hogtied.

On that note, be aware that prompt descriptions are frequently ignored. For example, I’m constantly writing “…with his arms behind his back” and frequently his arms aren’t as you’ll see later.

Getting good images with multiple people is really tricky. Often, the same character will be used for multiple people in the image so you end up with what looks like twins or triplets. And large groups of people tend to look like weird and lack detail from what I’ve experienced.

One way I’ve found success with people looking distinct from each other is if there’s an inherent contrast between them such as a player/coach situation.

I assume the word “gag” is always blocked. But describing them works sometimes. And as you’ll see, sometimes I can’t get them to work at all.

Tape gags are probably the easiest to get.

Cloth gags such as bandannas or scarves also frequently work with with language like “he has a red bandanna wrapped over his mouth” but often it looks like western train robbers masking his face rather than being a gag.

For ball gags, I usually do some variant of “his mouth is open with a small (color) ball (held, strapped, buckled, or wedged) (between his teeth or inside his mouth)” and sometimes I’ll tack on “with leather straps”. Different combos of those words tend to work but some situations will block all of them. It’s really inconsistent.

Costumes can be useful. Sometimes it won’t generate a particular character. I was trying to make a Captain America image and kept getting blocked. Then, instead of referring to the man AS Captain America, I tried describing a man in a Captain America costume and it worked.

I’ve found that different elements of a prompt get mixed up and jumbled. Coaches tying up an athlete frequently have rope wrapped around their wrist. I was trying to make a Deadpool and Wolverine scene and Deadpool would almost always have Wolverine claws. It’s incredibly difficult to get images of football or rugby players ball gagged because the gag will just be a ridiculous large football or rugby ball. Or having a guy tied up with rope and ball gagged with a leather strap will frequently make the leather straps rope instead. I’ve been trying to make a good image with a cyclist in it but it always includes the bike in really awkward ways. Anyway, just be prepared for mix ups like that.

Related to that, keep in mind that a computer needs to interpret what you write. The ordering of words may really matter. Like, I’m sure if I wrote, “deep in the ocean, a man is on a submarine,” the AI would probably know to not literally put the man on top the submarine. But you could avoid that ambiguity by writing “inside the submarine.” Or, “atop the submarine” if you really did want him on there. I’ve occasionally run into issues where I write something like, “…he is tied up to the chair with rope…” and although he would be tied to the chair, there’d be extra rope hanging around him oddly or bundles of rope around him… he’s literally, physically “with rope.” You could avoid this by writing something like “he is tied with rope to the chair.” I probably forget to do this but sometimes when I notice some oddness, it’s from this type of issue.

I've had a really hard time getting good, nuanced facial expressions. Any descriptor I use tends to be comically over dramatic. Still working on this.

Also, the maximum character count is pretty limiting so you kinda need to decide which parts you want to specifically describe the most.

Some thoughts on Bing

Here’s my soapbox. It’s incredibly frustrating to use. I get that Bing doesn’t want to allow ANY kind of image to be generated. But honestly, it feels like it’s judging and shaming alternative lifestyles. I can only imagine how much more difficult it is generating images of women. And while I know this stuff might feel extreme or shocking to some, but it’s generally pretty vanilla within the world of BDSM. I’m not trying to do anything sexual or with nudity. Guys tied up and gagged have been portrayed in western, spy, police, historic, comedy, adventure, and action TV shows and movies of all rating levels for all ages since the start of the mediums. It’s ridiculous to block this stuff. And I’ve seen some questionable stuff get generated. I simply wanted the guy to have darker hair and skin so I used “Mexican” as a descriptor and the image generated was frankly a racist stereotypical depiction. My prompts are always describing grown adults and one image was a young child which was disturbing. I had another prompt that was working with “Scottish” and when I switched it to “Brazilian” to mix it up, that got blocked. Why does Brazilian automatically get blocked? While I understand the need for some moderation, it seems like they should focus on other areas. 

Now let’s try all this out to develop a sample prompt

Continued in part 2