With cardinally encoded preferences, we can develop a utility function that represents our preferences and adheres to the common-sense criteria listed above.

But should we? What about utility monsters?