Extension of an order on a set to the power set: Some further observations (Q1104836)

Given a weak preference ordering (i.e., a reflexive, connected and transitive preference relation) over a finite set of alternatives \(\Omega\), the paper is about the extension of this ordering to the power set \(2^{\Omega}-\{\emptyset \}\). Two properties are introduced. The first one is the averaging measurement of extension, called Gärdenfors Principle (GP), which says that if an alternative is added to a set which is better (resp. worse) than the best (resp. least) element of the set then the expanded set will (resp. will not) be preferred to the original set. The other property is the additive measurement of extension, called monotonicity (M), which requires that if a set A is preferred to a set B and if a set C (which has no common element in either set) is added to both A and B then \(A\cup C\) will be preferred to \(B\cup C\). \textit{Y. Kannai} and \textit{B. Peleg} [J. Econ. Theory 32, 172-175 (1984; Zbl 0533.90005)] have shown that given a linear order over a set of at least six elements, it is not possible to have a weak order on the family of nonempty subsets that satisfies both GP and M. This paper shows that Kannai-Peleg's impossibility result does not hold for a set of five elements. It also establishes that in ranking A,B, where \(A,B\in 2^{\infty}-\{\emptyset \}\) the possibility of nonempty \(A\cup B\) in both GP and M is the critical factor that divides the line between possibility and impossibility results.