Because it is more difficult for a human being to express something in numbers, the values for the attributes are given in natural language; they have to be translated: null = 0.01; very few = 0.1; few = 0.25; medium = 0.5; high = 0.75; very high = 0.9; total = 1. The output has to be interpreted: values over or equal to 0.5 mean `the person is a pet holder' and values less than 0.5 mean `the person is not a pet holder'.
E.g., it is now possible to describe the finder of a dog:
For these values the finder of a dog is a pet holder and therefore liable for the damage caused by this dog.
Whereas the input side uses seven values, the output side has only boolean values. So no `defuzzification' and approximate reasoning is necessary.
All rules at this point are a hundred percent certain (UC_FACTOR= 1). An uncertain rule could be created by: