I know i said earlier i back up from here but , this being my field, obviously i cannot keep my nose out from the subject so …
It is not quite this simple. Social interactions in the game are very complex thing that is affected by many things. What current research and discussion suggests, it is not about personality types or proximity princible or mere-exposure effect, but it is more depending if the game is designed in the way it supports different trust levels between players and gives opportunities to make bonds with good trust. This all needs to be done within limits of our cababilities to maintain different levels of friendships (Dunbar’s layers are usually used as measure). So in short, game activities have to support cognitive social cababilities of the players and create repeating positive loops that are needed in relationship building.
You seem genuinely interested on subject, so here is one link to report from last year about this subject from convention where game makers from different game houses have yearly think tanks about different aspects of developing better games. I can look you some more too if you wish, but i think you will also find those easily yourself, just starting from references section on end of this report.
Group Report: Design Practices for Human Scale Online Games
https://www.projecthorseshoe.com/reports/featured/ph18r8.htm