Attempts at interface agent personalization are usually aimed at helping the user perform a task or service. For example, scheduling of appointments, inspection of messages, discovering items of interest and different forms of negotiation. While this is very noble undertaking, it makes assumptions about the level of trust and credibility a user may place in such an agent in a real world setting. If Microsoft’s experiments in social user interfaces teach us anything, it is that a "one size fits all" solution does not truly engage the user and encourage reuse. This "relationship management" between the user and the character begins when the two first meet. As with human-human relationships, first impressions are essential. Instead of looking at the functional aspects of the relationship, we believe the characters embodiment is the best place to start the personalization process. We describe a study in which participants from several different age ranges, genders and ethnic groups were asked their preference of anthropomorphic character, based on a cooperative computer task. We found that participants generally selected characters from the same ethic group as themselves and that almost all participants selected a young character (instead of a middle aged or elderly character). No significant preference was found for character gender.