The default model for EA community builders, especially college community builders, is to do most community building via formal programs such as intro fellowships targeted at students who the community builders have little or no particular connection to.  However, I claim that per unit time, EA outreach to your family/friends/associates (which I will refer to as your "social network") is much more effective.  This is for four reasons:

1. Even very smart/rational people are persuaded of things significantly on the basis of how much they like the person who is attempting to persuade them/how much they trust them.  As a community builder at a large college, random students who join a reading group have no particular reason to trust or like you.  However, presumably your friends/family/associates like and trust you a decent amount, making it easier for you to persuade them of things in general (and potentially to apply mild social pressure to them to take actions on the basis of the things you persuade them of).  Moreover, if you are known by your social network to be a more thoughtful/smart/good person than average (as I suspect that the average EA is), you will be even more able to influence them relative to a person who does not know you.
2. People in the social networks of EAs are likely to themselves have significantly stronger proto-EA attitudes than randomly selected college students, as people tend to befriend people who are like themselves, and people tend to be similar to their families for genetic reasons.  Given that whether a person has strong proto-EA attitudes appears to me to be the main determinant of whether or not you can successfully convince them to get involved in EA, this makes people in the social networks of EA community builders much more promising targets for outreach than people from many other groups.
3. You have much more information about the people in your social networks than you do about random college students.  This allows you to select members of your social network to pitch EA to who are particularly persuadable and promising, and also enables you to tailor your pitch more specifically to them than you are able to for a random college student.
4. In my experience, long-ish one-on-one conversations are remarkably effective as a community building tactic.  Generally, random college students will not be willing to have an extended one on one conversation about EA with you.  However, you can almost certainly have an extended one on one conversation about EA with someone in your social network.

Personally, as a community builder at Harvard EA, I believe I have had much more impact through my informal discussions with my friends and family members than I have through running formal programs and reading groups.  While I think I select my friends for proto-EA characteristics more than the median EA, my guess is that it would still be more effective at current margins for most EA community builders in college to spend a unit of time on this kind of informal outreach as opposed to a unit of time on formal community building.