Uhm, I seem to recall that social media was actually pretty good in the late 2000s and early 2010s. The authors used AI models as the users. Could it be that their models have internalized the effects of the algorithms that fundamentally changed social media from what it used to be over a decade ago, and then be reproducing those effects in their experiments? Sounds like they’re treating models as if they’re humans, and they are not. Especially when it comes to changing behaviour based on changes in the environment, which is what they were testing by trying different algorithms and mitigation strategies.
Uhm, I seem to recall that social media was actually pretty good in the late 2000s and early 2010s. The authors used AI models as the users. Could it be that their models have internalized the effects of the algorithms that fundamentally changed social media from what it used to be over a decade ago, and then be reproducing those effects in their experiments? Sounds like they’re treating models as if they’re humans, and they are not. Especially when it comes to changing behaviour based on changes in the environment, which is what they were testing by trying different algorithms and mitigation strategies.