The paper examines how classical mathematical statements about averages and means behave when you require that everything be explicitly computable. In ordinary mathematics, you can freely ask whether two different ways of averaging a set of numbers give the same result, and the answer is either yes or no. Constructive mathematics adds a stricter requirement: every mathematical object must come with an actual procedure for computing it, and every proof must provide an explicit method for finding the answer. The paper works within this stricter framework, using real numbers that are defined by computational procedures rather than just assumed to exist abstractly.
The central finding is a negative result: there is no general algorithm that can always determine whether two means of constructive real numbers are equal to each other. This is a computability impossibility result, meaning the problem is not just hard in practice but genuinely unsolvable in principle for arbitrary inputs. The authors also establish various inequalities between different types of means in this constructive setting, which requires more careful argument than in classical mathematics because you cannot simply assume that two quantities are either equal or unequal without having a procedure to decide which.
The one important exception is reassuring and intuitive: if the input numbers are rational rather than arbitrary real numbers, the problem becomes fully solvable. Rational numbers are finitely describable and their arithmetic is exactly computable, so an algorithm can always decide equality in that case. This contrast between rationals and real numbers highlights a fundamental boundary in computation, where moving from discrete, exact inputs to continuous ones can turn a straightforward decision problem into something provably impossible.