These companies should be criminally liable for training their models on child porn.

you are viewing a single comment's thread
view the rest of the comments
[–] 5 points 1 day ago* (last edited 18 hours ago) (4 children)

Aside from the other good replies you have around the subject, i'll address something more fundamental to your argument.

you can’t just dismiss an argument because the study is small.

You absolutely can dismiss a study for being too small. In some cases you really should.

If you can't procure and retain a meaningful sample size, then you shouldn't be running the experiment (or more usefully, designing an experiment that works with smaller sample sizes, or one that uses a demographic where you can get adequate sample sizes).

It's one of the reasons why participant gathering is usually done up front.

  • source
  • parent
  • hideshow 4 child comments
  • [–] 2 points 1 day ago (3 children)

    True. But over 1500 responses is a massive sample size for a survey like this.

    I haven't read the actual study yet, but I would assume they filtered out things like one person submitting a thousand responses.

  • source
  • parent
  • hideshow 3 child comments
  • [–] 2 points 1 day ago* (last edited 18 hours ago) (2 children)

    True. But over 1500 responses is a massive sample size for a survey like this.

    Genuine question, based on what?

    Like, I'm just seeing a number, I've no idea if it's big or small in the context of this area of study.

    I'd be interested to know why that's considered a good result.

    I haven’t read the actual study yet, but I would assume they filtered out things like one person submitting a thousand responses.

    Agreed, I'd assume a basic level of scientific rigour.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 4 points 1 day ago (1 child)

    https://www.research-advisors.com/tools/SampleSize.htm

    I don't know what the total population is here (people interested in CSAM? everyone on earth?) but 1500 responses gets you a 2.5% margin of error with 95% confidence on a population of 300 million people (a bit less than the entire US).

  • source
  • parent
  • hideshow 1 child comment