Comment on They Fell in Love With A.I. Chatbots — and Found Something Real

etherphon@midwest.social ⁨1⁩ ⁨day⁩ ago

• spit_evil_olive_tips@beehaw.org ⁨1⁩ ⁨day⁩ ago

  One in five are you god damn fucking serious?

  yeah…they call it “a recent study” but don’t bother to cite their source. which I find annoying enough that it nerd-snipes me into tracking down the source that a reputable newspaper would just have linked to (but not a clickbait rag like the New York Times)

  this article from a month ago calls it “Almost one third of Americans”. and the source they link to is…a “study” conducted by a counseling firm in Dallas. their study “methodology” was…Surveymonkey.

  this is one of my absolute least favorite types of journalism, writing articles about a “study” that is clearly just a clickbait blog post put out by a business that wants to drive traffic to their website.

  (awhile back, a friend sent me a similar “news” article about how I lived near a particularly dangerous stretch of I-5 in western Washington. I clicked through to the source…and it’s by an ambulance-chasing law firm)

  but if they had used that as the source, they probably would have repeated the “almost one third” claim, instead of “one in five”, so let’s keep digging…

  this from February seems more likely, it matches the “1 in 5” phrasing.

  that’s from Brigham Young University in Utah…some important context (especially for people outside the US who may not recognize the name) is that BYU is an entirely Mormon university. they are very strongly anti-pornography and pro-get-married-young-and-have-lots-of-kids, and a study like this is going to reflect that.

  a bit more digging and here’s the 28-page PDF of their report. it’s called “Counterfeit Connections” so they’re not being subtle about the bias. this also helps explain why the NYT left out the citation - “according to a recent study by BYU” would immediately set off alarm bells for anyone with a shred of media literacy.

  also important to note that it’s basically just a 28-page blog post. as far as I can tell, it hasn’t been peer-reviewed, or even submitted to a peer-reviewed journal.

  and their “methodology” is…not really any better than the one I mentioned above. they used Qualtrics instead of Surveymonkey, but it’s the same idea.

  they’re selecting a broad range of people demographically, but the common factor among all of them is they’re online enough, and bored enough, to take an online survey asking about their romantic experiences with AI (including additional questions about AI-generated porn). that’s not going to generate a survey population that is remotely representative of the overall population’s experience.

  source

  • Powderhorn@beehaw.org ⁨16⁩ ⁨hours⁩ ago

    I mean, it absolutely would be news if a dangerous stretch of I-5 popped up in Eastern Washington.

    source
  • etherphon@midwest.social ⁨1⁩ ⁨day⁩ ago

    Thanks for the deep dive, yeah that seems really suspicious, the pdf of that study definitely looks like it was paid for by someone with an agenda. I feel like anyone who’s up for taking a survey on AI is already into AI so it’s pretty skewed. I suppose one thing from that study did make some sense:

    Many people who are using AI relationship technologies report that they prefer AI communication over engaging with a real person in a genuine relationship.

    Sure, everyone loves a sycophant who hangs on their every word and agrees with them all the time… for a while at least.

    source
• XLE@piefed.social ⁨1⁩ ⁨day⁩ ago

  Considering the article has selected people who claim chatbots have fixed their marriages… This is a puff piece.

  source