Actually, the problem is not with labeling or with generalizing, per se.

There are actually two problems.

Problem 1a.    When one treats a valid generalization as a "rule" so to speak, one tends to view all the data that doesn't fit the generalization as a remarkable exception. Unless one is dealing with something that has only a very limited number of attributes (say one to three attributes), however, the generalization is equivalent to an idealized representation and no one member of the class of whatever it is that is being generalized about is likely to exhibit all of the attributes. This means that variance from the generalization is the rule most of the time. People tend to not recognize this. (That last statement is a generalization, and the use of the word "tend" signifies I am aware it is a generalization.)

Problem 1b. Failure to recognize assumptions and beliefs for what they are in one's own thinking processes. There is no inherent problem with making assumptions or having beliefs. The problem in thinking is the failure to recognize them as such - to mistake them for truth or fact.

Problem 2. Value judgement. This is related to 1b. above, but separate from. Value judgements are based on assumptions and beliefs, but not all assumptions or beliefs are value judgements.