Fuzzy Information Axiom

The Fuzzy Information Axiom is a ranking method for when both the performance range an alternative actually achieves and the acceptable design range are given as fuzzy numbers, and you want to find which alternative meets that range with more confidence. It is the fuzzy-data adaptation of the Information Axiom from Suh's 1990 axiomatic design theory; Kahraman and Kulak defined this adaptation in 2008 with manufacturing-system-selection examples. The output is an information content figure for each alternative and a rank derived from it; the smaller the figure, the more confidently the alternative meets the functional requirements.

Suh's axiomatic design theory holds that a good design must satisfy two conditions: the requirements must be independent of one another, and the design must meet those requirements with minimum uncertainty. The Fuzzy Information Axiom carries the second condition into situations where measurements are not a crisp figure but a range, a triangular fuzzy number. Think of an archer's arrow landing on a target. What matters is not the single point where the arrow lands, but how much the region the arrow could plausibly land in overlaps with the target's acceptable region. A small overlap means the knowledge of hitting the target is uncertain, that is, there is a high chance of surprise; a large overlap means little uncertainty. The philosophical consequence of this approach is exclusionary. If an alternative's range does not intersect the design range at all, that alternative is judged unsuitable for that requirement regardless of how well it performs on other criteria.

The method proceeds through five steps.

First, gathering the system and design ranges. Every alternative's performance on every functional requirement is expressed as a triangular fuzzy number (lower bound, most likely value, upper bound); this is called the system range. The decision-maker's accepted design range is given in the same form, as a triangular fuzzy number. Where linguistic expressions are used, these are first converted into such triangular numbers.

Second, computing the system area and the common area. The method computes two areas for every cell: the area under the system range's own membership function (the system area) and the area of the region where the system range and the design range intersect (the common area). For a triangular fuzzy number, the system area is half the width of the range. The common area is the geometric area of the region where the two triangles overlap; this step is not fuzzy arithmetic but the measurement of the intersection region of two triangles.

Third, computing the information content. For every cell, information content is the base-2 logarithm of the ratio of the system area to the common area. If the common area is zero, that is, if the two ranges do not intersect at all, information content tends to infinity and that alternative is judged wholly unsuitable for that requirement.

Fourth, finding the total information content. Each alternative's information contents across the functional requirements are multiplied by the requirement weights and summed. If no weights are given, all requirements are treated as equal.

Fifth, ranking. Alternatives are ranked from lowest to highest total information content; the alternative with the smallest information content is the design that meets the requirements with the most confidence.

The formulas behind each step, the intermediate tables and the citation formats are given on the DecisionMind method page; this card carries no formulas.

Information content shows how little uncertainty an alternative carries in meeting these design ranges, and nothing more. The figure itself is not a probability or a percentage; it is used only to rank alternatives against one another. An information content close to zero shows that the alternative's range overlaps almost entirely with the design range. A large figure shows that the overlap is small, that is, that the alternative meets the requirement only marginally. Infinite information content means the alternative is wholly excluded for that requirement; regardless of how well this alternative performs on other requirements, it is treated as excluded overall.

A small information-content gap between two alternatives can reverse if the ranges are shifted only slightly, because the common-area calculation is sensitive to the range boundaries. The report should show how sensitive the ranking is to the boundaries of the design range.

Thus instead of writing:

"The Fuzzy Information Axiom found the best design"

the report should read:

"With these design ranges and these weights, the alternative carrying the least uncertainty is this one; the ranking is sensitive to the boundaries of the design range"

The Fuzzy Information Axiom works with triangular fuzzy numbers: every cell consists of a lower bound, a most likely value and an upper bound. At the time of writing, DecisionMind records no extension to this method beyond the base method. If your data is crisp, it must first be converted into a triangular fuzzy number, using a linguistic scale or an expert-supplied range.

You need the following: a system range for every alternative on every requirement, a design range for every requirement, direction information for every requirement, and, preferably, requirement weights; if no weights are given, requirements are treated as equal. A minimum of two alternatives and two requirements is required; the recommended size is two to twenty-five alternatives and three to ten requirements. Requirements are assumed to be independent of one another; where strong dependence exists, this must first be addressed with the Independence Axiom.

The Fuzzy Information Axiom is a suitable choice when your performance measurements and your acceptable range are not a crisp figure but a range or an expert estimate, and you want to ask how confidently the design meets the requirements. Its typical territory is manufacturing-system selection and engineering decisions that call for linguistic expert assessment.

The case where it should not be used arises when there is strong dependence between requirements; the method treats requirements as independent, and if this assumption does not hold, the result is misleading. If your data is already crisp and uncertainty does not matter, the classical Information Axiom or other crisp ranking methods are simpler.

Performance and acceptance range are fuzzy, requirements are independent → Fuzzy Information Axiom

Data is crisp, uncertainty does not matter → classical Information Axiom or methods such as TOPSIS, SAW

Requirements are strongly dependent → redesign first with the Independence Axiom

No overlap at all is acceptable on one requirement → methods built on elimination logic

The method's chief strength is that it brings uncertainty explicitly into the calculation instead of hiding it at the data-collection stage. It flags, explicitly, with infinite information content, the case where an alternative does not overlap with the design range at all; this produces an exclusion that cannot be offset by good scores on other criteria. Because it rests on axiomatic design theory, it aligns directly with design-discipline principles such as the independence of engineering requirements.

Its limitations stem from the same structure. First, the common-area calculation is sensitive to the range boundaries; shifting the boundaries only slightly can change the information content and, with it, the ranking. Second, the assumption that requirements are independent is not always realistic; on strongly dependent requirements the method can give a misleading result (Suh, 1990). Third, converting linguistic expressions into triangular fuzzy numbers requires a separate expert judgement, and this conversion directly affects the result. Fourth, infinite information content demands special handling in practice; if more than one alternative takes an infinite value, no ranking can be made between them.

The most common mistake is reading information content as though it were a probability or a percentage; the figure is used only to rank alternatives. A second mistake is setting the design range narrower than it really is and thereby excluding alternatives needlessly; the range is a separate engineering decision and must be justified in the report. A third mistake is continuing to apply the method despite knowing that dependence exists between requirements. A fourth is failing to state, in the report, the scale used to convert linguistic expressions into triangular fuzzy numbers; a different scale gives a different result.

The governing principle is this:

The Fuzzy Information Axiom result is a reflection of the design ranges you set and the assumption of requirement independence; if the ranges are contested, the ranking is contested too, and the report must show this.