Then if you're just looking at the viewer metrics, it looks like the step sibling stuff is really popular. So the algorithm prioritizes it further, so people watch it more, etc.
This is all speculation of course, but it's a cautionary tale about why statistical and causal reasoning is useful even if you're in industry, as long as you actually care about delivering a good data product.
I recall reading something years ago that said some porn site that started to do data-driven analytics and found that even though "step-sibling/parent fantasy" stuff wasn't generally popular, but it was particularly good at attracting paying subscribers. I kinda wonder if that effect was overstated, but all the porn sites started cargo-culting that finding as "conventional wisdom" and made the genre a thing.
> So much porn is perfectly normal-looking based on the thumbnail/preview other than the creepy title, so maybe people just watch it anyway if it looks good, even if the plot premise is offputting.
I think of lot of that particular genre isn't normal-looking, for instance by casting size differences to creepily create the impression of an illegal age difference:
https://www.vice.com/en/article/aeygae/porns-latest-trend-is...