I read a great article in the Journal of Management Information Systems yesterday. The paper was about how team members develop trust in one another, but is applicable to trust in any inter-personal relationship.
The findings show that there are two fundamentally different mechanisms through which we decide if we trust someone. At first, before we have any personal experience with an individual, we generate a sense of trust based on that person's membership in a group (demographic, organizational, etc). So for example if we think of Hispanics (or accountants, or Celtics fans) as being trustworthy, then we will instinctively trust a new Hispanic (or accountant or Celtics fan) person that we meet. This is mediated by a propensity to trust in general, which makes some of us more trusting than others across the board. This is called "swift" trust.
Then, once we gain personal experience with the individual, the stereotype-based swift trust disappears and our trust is based just on the personal experience. This is called "knowledge-based" trust.
They also found that confirmation bias impacts how the swift trust becomes knowledge-based trust. If our stereotype tells us that a person is trustworthy, the fundamental attribution phenomenon tells us that the trustworthy things they do are in character and increase our knowledge-based trust. It also tells us that the untrustworthy things they do are out of character so we discount it. So our stereotypes influence how we perceive the real experiences that come later.
The paper describes some management practices that are important for team performance with this in mind. It is critical to generate some swift trust at the beginning of a new team to make sure that the bias is towards mutual trust. So the team leader wants to emphasize the group memberships that are trustworthy and de-emphasize any others. Team building exercises may be helpful at the very beginning, but not later.