I think hesitation and thoughtfulness is probably the correct initial reaction. People frequently pay less attention to situational context and more attention to internal dispositions than is justified*, so making sure that your environment is doing what you want it to do is generally a great idea. There are probably two radically different questions here.

First, there's the question of whether or not you think "rationality" is likely to help you. If you think it is a good idea to take beliefs and theory literally, act on their basis, with updates every time you learn something or mess up so that you are "more effective" in the future, then doing things to become more effective in this sense at a faster rate might be a good thing to try. If it fails, you can always update yourself out of "living according to theory". There are good reasons to think that mostly you'll do stupid things and hurt yourself initially. If any person doesn't think that they'll ultimately benefit from it, then maybe you should pick someone crazy to live with. Or, taking the fear-of-belief seriously, maybe you should simply leave roommate selection to nonreflective impulses?

But if you do think that rationality is something worth actively cultivating, even taking the likely initial costs into account, then the question is more "economic". How much will living with a creationist teach you? How much will living with a schizophrenic person teach you? How much will living with a school teacher teach you? How much will living with a group of committed rationalists teach you? I think the "lessons per day" and perhaps also the "aggregate value of the lessons per day" is likely to be highest if you're in a community that's consciously optimizing itself along that axis.

And yes, groupthink is a potential worry, but if you were (rationally) going to take them seriously you might simply read about it and talk about it and try to develop cultural forms that make it more difficult. For example, someone in the house could take responsibility for giving one third of the house members a groupthink questionaire (see appendix E, D, and A in the link) and keep a chart over time of the scores. Just being mindful enough to do something like this would probably go a long way towards ameliorating the problem. And if there is a problem then the measurement might reveal it and allow it to be squarely addressed.

Some people might think this is a waste of time, and presumably they'd have arguments for why this would or would not be so... but that just gets back to the core question of whether or not people's theories (even our own) can really be trusted to serve as the basis for planning and action.

Orgies.