Each respondent was randomly sorted into one of two scenarios (an A/B experiment), then shown 15 random faces and asked the same yes/no question about each. The faces are real photos drawn from an age-guessing survey; the x-axis here is each face's perceived age — the average age strangers guessed it to be.
🧛 Casual sex arm (the “vampire” one-night-stand). Intro: “You're single. An immortal vampire appears in your window at night. She'd like to have a one-night stand. There will be no consequences, and nobody will ever know. Do you say yes?” Then for each face: “This 200-year-old vampire would like to have a one-night stand. No consequences, nobody will know. Do you agree?”
The vampire framing is deliberate: she's actually 200, so the face's apparent age says nothing about her real age, and there are no consequences or social judgment — the aim is to isolate raw physical attraction from everything else.
💖 Date arm (serious relationship). Intro: “Imagine you're single, available, and looking for a relationship. A person approaches and asks if you want to go on a date. She's looking for something serious.” Then for each face: “She's looking for something serious, and wants to go on a date with you. Do you go on the date?”
👪 Public-wife arm (sexless public marriage). For each face: “If you press the button, she becomes your public-only wife. You are never intimate with her, but everyone else believes you are married. (You're cursed to stay single forever otherwise.)”
This arm isolates whether a face is someone you'd want as a public partner / social companion, with physical intimacy explicitly removed. (Added later, so its sample is smaller than the other two for now.)
Answer options: No / Borderline-maybe / Yes. The default score metric counts Yes = 1, Borderline = ½, No = 0, averaged across everyone who saw that face.
Also asked: own age, number of sexual partners, gender, and whether attracted to women. By default this explorer shows only men who said they are attracted to women (toggle it off above to include everyone).
Respondents were told the photos were randomly generated/selected by an algorithm, not hand-picked.
temptation-survey.csv anywhere on this page to refresh. Error bars are Wilson 95% CIs; n per face is small early on, so they'll be wide.