Statistical means should not automatically define a concept’s membership midpoint.

Evidence

The 0.50 membership value should not be automatically determined by the distribution of the data. In general, one would almost never use the mean of the data as the 0.50 level in fuzzy-logic analysis. For instance, one would almost certainly not use half the seats occupied as the 0.50 semantic value for the concept of full planes. There are various ways that one might do this, a number of which are discussed in the next chapter. Again, the key point is that mindlessly choosing the middle of the data scale or the statistical mean or median of the actual values is not necessarily a good idea, and in fact usually is not. These can inform the choice, but they should not completely and automatically determine it. — Gary Goertz, Social Science Concepts and Measurement, 652

Categorization

• SEED: Article angle: the quiet mistake of letting the distribution set the threshold — how ‘average’ gets smuggled in as ‘meaningful’ across indices and rankings.
• CANDY: The half-full plane is a crisp, quotable illustration that statistical middles and semantic middles are different things.