The social capital here is based on profit related to learning - and to getting better grades. The structure of the class is purposefully designed to encourage students to activate their ties and get the help they need. In this classroom environment, students are very likely to make judgments about their peers' reliability and ability, and they then use that to either help other students or to seek help from other students (analogous to the job seekers in the article by Sandra Susan Smith, "Don't put my name on it": Social Capital Activation and Job-Seeking Assistance among the Black Urban Poor, Amer. J. Soc., 111, 2005, 1-57).
The interplay of networks, social capital, and classroom structure lead to interesting questions that help us understand how students learn. Do students look to their peers for help? Which students willingly help other students? If they do, why - is it related to reputation or status? The same question can be asked if students don't ask for or give help. Do students identify certain other students - in their teams or in the whole class - as "smart", and how do they make this identification? Does assigning students to teams build a "better" network (activate more capital) than letting students choose their own teams? Finally, what about students who feel very different from their peers? Does requiring them to interact actually block learning for these students?
I don't have hard evidence (yet!) to answer these questions, but anecdotally I have found that most students do learn better and learn more deeply when I teach using POGIL than when I taught using traditional lecture. What is your experience?