The primary advantage of promoting a sociable problem solving is that it engages a higher proportion of students. My preference is to have them work in pairs on one piece of paper, but some students prefer to work alone. Usually this is respected.
Experiment with different social clumping. There is evidence (See this paper by Peter Liljedahl) that randomized student pairings is superior to either students or teachers selecting the pairings. The first time I mentioned this to a teacher she was unimpressed as there were combinations of students in her class that would be guaranteed to erupt… however the important thing is that the students perceive the pairings to be randomized. Cheat behind the scenes by prohibiting certain student combinations.