Social medias are applications and platforms that enable different users to engage with one another. Popular social medias include Facebook, YouTube, WeChat, Instagram, Tiktok, Twitter, Reddit, Parler and Tinder.
The dynamics of social medias is an important research topic for AI ethics, as the ethical behavior of algorithms seems to strongly depend on their interactions with users and groups of users.
Academic research on social medias is unfortunately usually limited because of multiple factors. Perhaps most importantly, social medias are highly customized to the users, and the way they are customized to a given user usually vastly exposes the users' personal data. This opaque customization makes external analyses of social medias largely incomplete.
It is also noteworthy that external researchers cannot run randomized experiments. Their analysis must then account for confounding variables to avoid Simpson's paradox, which makes their reliability limited. Finally, part of the research relies on user surveys, which makes data vulnerable to reporting biases.
Given such limitations, Aral-20 calls for more transparency from private companies to enhance our understanding of social medias, especially regarding AI ethics concerns like disinformation, misinformation, radicalization, manipulation, cyberbullying, hate, addiction, attention span, loneliness, depression, suicides, biases and deplatforming. In fact, social medias' data have been argued to represent important public health HoangFE-21, science communication Hoang-20 and alignment opportunities StrayAH-20.