Why is it that we do generalize? What the use? If every event or every part of every event is considered to be unique and not within the context of a generality, then we could not use the knowledge of the event in the future because we would be unable to link the event to any future event. In the generalities lie the predictive power which makes the knowledge of an event useful. In my mind this is why it seems to me that humans spend a large portion of their brainpower on the process of correlation, or generalization. Correlation is also part of root process of data compression.

Now, that gets particularly problematic when we carry our generalizations to the level that we assume them to be correct under all conditions, without requiring any further inspection.

I would agree, but I would say that the problem is not in the generalization in general ( yeah that hurts I know ) but rather in the degree of generalization. If I generalize too greatly in the topic of color say, I may be left with only the ability to distinguish between black and white. This is a problem because (absent their locational positioning ) I would be unable to determine if a traffic light is red or green. In this example, over generalization is clearly not a helpful extreme.

Predictive power is a function of generalization. At the two extremes of generalization ( none, or complete ) there is little predictive power. No generalizations provide no predictive ability since you can't correlate two events. Over generalization obscures further detail which might allow further generalizations and hence further predictive powers. However in between there lies an amount of generalization which provides the maximum predictive power. So to utilize the information of an event maximally, one must take care to generalize, but not to over-generalize.

Whose ideas do you trust? Usually the people who've given you good ideas in the past. How do you check them? By using them in your everyday life. So the symbolic representations (generalizations ) which are passed around among humans can be checked. By correlating good ideas with certain sources of information you can begin to trust the sources of information. The process of generalization allows you to judge a source of information. On the other hand, you can over generalize here as well. An individual might be excellent at passing along ideas about solving computer problems. If you generalized too far and said "this person only gives out great ideas." then you might have judged wrong, because the person might be really horrible about helping with stickly social problem. ( Ok, that's the computer nerd generalization... ;-) )

The world we live in is inhabited by other adaptive systems and hence it is dynamic. To never revisit your generalizations and test them could lead you astray because what was once a valid generalization may no longer be.

I guess my general point is that generalization are not bad things, but they are dangerous if used either too much or too little.