When someone else’s actions are significantly unlike our own, we are forced to build a new model, not based on ourselves. That model – almost inherently – is based on very little information. We take the most obvious bits and make our models out of that.
Take a child’s drawing of their parents. In one of my son’s drawings of our family, the main difference between my wife and I is that I have a goatee. Obviously, there’s a LOT more differences – and more have shown up in his drawings of us – but this is some anecdotal evidence of the effect in action.
This leads to its own problems. Stereotypes, for example, are a textbook example. While stereotypes are inherently bad (IMHO), it’s sometimes hard to quantify why. And that’s when you get to security issues – and profiling.
The police point to profiling (racial or otherwise) as being accurate. Yet many innocent people get pulled over for driving while black, or simply speaking Arabic on a plane, or having an unusual name. Both are true – because of a failure of Meadian empathy. By creating a model of “the bad guys” based on a limited amount of highly visible data, there’s a loss of accuracy and sensitivity in any police operation.
Put another way – while the DHS folks are patting down someone whose crime is looking Arab, a white terrorist (remember Tim McVeigh?) could get by relatively unexamined. Want to make sure your mule gets across the border? Make sure they look and act like a middle class American.
You don’t need to resort to sociobiology, functionalism, or any particular ideology. Through the way we learn about others, through the way we appear to develop empathy is a mental explanation of stereotypes, and a compelling argument that they need to go away. Not just for some fuzzy-squishy reason, but for the very practical reason of our own security and safety.