The whole idea some people have of meeting a cofounder a few times (especially in an artificial environment like a hackathon) and then making an all-of-nothing decision is flawed, IMO. The best process is to spend a LOT of time with the person, and work with him on progressively larger projects. Yes, how they interact with people is an important factor, but I'd try to spend many months (or years) on the process, not a few meals.