Integer valued means (Q744037)

This is a very interesting paper not least because of the historical overview of the classical two variable means that it contains. The basic problem that the paper confronts is that if the set to which the means are applied is in \(\mathbb Z\) the answer can be rather absurd, as in the average family has 2.7 children! Are there any reasonable means that can give integer answers to integer inputs? By a mean the authors intend a quantity that satisfies the Kolmogorov axioms with the strict monotonicity replaced by monotonicity, and of course the dropping of the continuity requirement. The surprising answer is that the only such means are determined completely by the largest and smallest term of the set of quantities whose mean is being sought. That is if \(a_1, \ldots ,a_n\) are integers then such means of these numbers that are themselves integers are functions of \(a_1\) and \(a_n\) only, such as \( \max\{a_1,\ldots, a_n\}\) or \( \min\{a_1,\ldots, a_n\}\) .