# 220

Dear Bartosz Paradowski, Bartłomiej Kizielewicz, Wojciech Sałabun

I read your paper

Equal Criteria Influence Approach (ECIA): Balancing Criteria

Impact in Multi-Criteria Decision Analysis

My comments:

1- You say “This paper introduces a novel iterative method called the Equal Criteria Influence Approach (ECIA) to tackle this problem”

Are you sure it is novel?

The MEREC method (Keshavarz-Ghorabaee) was published in 2021 and yours is dated September 2024. It appears to me that your method has a very strong resemblance to MEREC in progressively eliminating criteria, run the software and compute the difference between the precedent alternative and the new coefficients generated by the elimination of one criterion at a time. Unfortunately, the two methods are not able to measure correctly the significance of criteria, because they do not preserve the same structure of the decision matrix, and both are solving, not a problem, but a series of problems, all different. You can also check this, consulting AI on this procedure.

2- You are not iterating, you are methodologically repeating a sequence, itmathematically, is not iteration. In iterations each step builds on the previous result. In your method each removal is independent of the precedent.

3- “of each criterion and assigns them corresponding weights based on their perceived significance”

I do not think that perceived significance is enough or real to evaluate alternatives, not only because their arbitrariness - which is enough reason to reject the process - but also because criteria weights cannot evaluate alternatives (See Shannon). They are only trade-offs among criteria. You admit this when talks about potential errors. To measure an error, you need something to compare how much you deviate from a standard or benchmark, as a specification. Which is your standard here? You do not have any.

4- “to allow for a complete consideration of the impact of individual criteria on the outcome of the decision”

What criteria impact are you talking about? An impact produces an effect on something, I do not see how a criterion can produce that impact, at least not in any MCDM methods. You can have an impact when you change values in another criterion in a cause-an-effect relationship, something that no MCDM method has. Why? Because for that you need an intersection structure that none of them has.

5- What is it that you call ‘content of a decision’. I know what it means but I do not understand what you identify with that expression, it could refer to many things

6- “ is based on MCDA sensitivity analysis, wherein individual criteria are systematically excluded from the decision matrix”

The Sensitivity Analysis (SA) is performed to assess the strength of a solution, but it does not eliminate anything from the decision matrix. If an alternative depends say of criterion C5, and this has not or very small variability, the corresponding alternative is at risk, because a little variation of C5 will change the position of the best alternative or produce a reversal of rank, which is not related to rank reversal (RR), when alternatives are added or deleted

I believe you are confused. In your system you remove a criterion, which is unknown in SA

7- Which are the base line preferences? You do not explain either how to compute alternatives weights. You also speak about impact. There is not such. If you remove a criterion it disappears from the scenario. What really happens is that you are solving a new problem which has differences with the former.

As a trivial example:

8- Suppose you are baking a cake and taste it; because you don’t like it, you remove one of the ingredients, say vanilla, and cook the mixture for the cake. Most probably you will notice a change but it is not attributed to any action; it is simple other type of cake.

9-If you prefer, we can talk mathematically. You know that criteria form a criteria space that contains all alternatives and which boundaries are the criteria. If you remove one of them the criteria space is modified because a former boundary does not longer exist. Since the new space does not have memory, it does not even take into account the change. It is a new problem with a new space

These are my comments. Hope they can help

Nolberto Munier