[W]hich bit exactly is racist, calling blacks and mexicans stupid or disagreeing with the folks at princeton who also think they are stupid but believe this should qualify them for special privileges at the expense of more deserving candidates?

Attribution theory's fundamental distinction leads quite naturally to its fundamental rule: When a behavior occurs in the presence of a sufficiently strong, facilitative force, an observer should not infer that the actor is predisposed to perform that behavior. Just as one should not conclude that a balloon that rises on a windy day is filled with helium, one cannot make unequivocal inferences about the abilities of an athlete, the convictions of a politician, or the mental health of a student when poor lighting, a roomful of opinionated voters, or sudden bad news may have induced their behaviors. In other words, one should not explain with dispositions that which has already been explained by the situation. This logical rule was first formalized by Jones and Davis (1965) as the law of noncommon effects and later extended and codified by Kelley (1967) as the discounting principle, which warns observers not to attribute an effect to any one causal agent (e.g., a disposition) when another plausible causal agent (e.g., a situational force) is simultaneously present. In other words, when people do precisely what the physical environment or the social situation demands, dispositional inferences are logically unwarranted.