Today I would first like to begin with error of classification and then generalize it more to error of losing detail. One common mistake people make is the error of classification. The error is so:
- All elements of class X has Y and Z property.
- All elements in the world with Y property must belong to class X and hence have Z property
No other post explains this idea better than this one : http://lesswrong.com/lw/e95/the_noncentral_fallacy_the_worst_argument_in_the/
However that post is too specific to errors of classification with respect to logical arguments. It applies very well to other domains. Moreover errors of classification has bigger problems because the classification itself can be wrong.
I will give you some examples:
Everyone knows that a smart person will talk well about what they are good at, barring artificial constraints such as language or mental illness. But we incorrectly assume that everyone who talks smooth is smart which in retrospect looks ridiculous.
I have also heard of people claiming their employers trying to build slides and provide bean bags for them because companies like Yahoo, Facebook, Google etc seems to do them. After all bean bags and slides are what makes these company cultures so great, isn’t it ?
Another example is exams. Here is where the problem of loosing detail can really be problematic. Most people who are good at understanding the basics of what they study can score reasonably well in exams. However, everyone who scores well does not have to be a good student. They don’t even have to be very smart. It is possible to tune the your study habits specifically to beat the system.
This creates a negative incentive system. If you consider plaudits from other people as an incentive ( which in most cases is a reasonable assumption for most mortals ) then there will be a lot of people trying to tune their study habits to beat the system than do what is profitable to the whole world, i.e, students who go into schools should try to gain knowledge and improve their analytic skills.
This would be an interesting problem to solve : to create a culture that values detail and makes correct classifications.