SHIC pig disease MCDA · worked example

How your votes set the weights

Not a simplified analogy — this is one of the real questions. PAPRIKA only ever asks about two criteria at a time, with everything else held equal, so a two-criterion example is the genuine article with the other six criteria switched off. Answer the trade-offs and watch the weights, and the ranking, move.

The setup

Two criteria, four levels on one and five on the other. Level 1 is always the least concerning end, exactly as in the real framework.

Criterion A

Herd mortality

Deaths in the affected herd.

StepValue

Criterion B

Human illness severity

How bad the disease is in a person.

StepValue

A criterion’s weight is the whole distance from its worst level to its best — it is not shared out, one portion per level. The worst level scores zero, so a criterion with four levels spends its weight over three steps, not four: 0.500 ÷ 3 = 0.167 a step. The five-level criterion spends the same 0.500 over four smaller steps of 0.125. Adding a level does not give a criterion more weight — it cuts the same weight into finer slices.

Before anyone votes, every criterion is weighted equally. That is an assumption, not a finding — it says a step from the bottom to the top of A is worth exactly as much as a step from the bottom to the top of B, and that the levels in between are evenly spaced. It is what the matrix uses today, and the whole purpose of the trade-off session is to replace it.

The trade-offs

Each question shows two hypothetical agents and asks which is the higher research priority. There is no right answer — the answers are the weights. What each answer does is rule out a range of possible weightings, so the feasible band narrows question by question.

These are the actual questions. In the full model the two agents differ on two criteria and are identical on the other six, which is what makes the comparison answerable — you are never asked to hold eight things in your head at once. Everything below therefore behaves exactly as the real session will.

What the answers leave possible

A worth nothing

A worth everything

Resulting weights

A weight is simply what it is worth to move that criterion from its worst level to its best. The two add to 1, so they are shares of one decision.

These numbers are the MIDPOINT of the band above, not a measurement. No single answer produces a weight — it rules out a range. The first question is worth 0.500 either way, so answering it only says which half you are in, and the midpoint of that half is 0.250 or 0.750. Later questions split the band unevenly. Which is the point: one answer halves the possibilities, and it takes about forty to pin a weight down.

Or explore by hand — drag to overrule the answers

What it does to the ranking

Five agents scored on those two criteria. The score is just the two level values added together — that is what additive means, and it is the property the whole method rests on.

Agent	A	B	Equal weights	After the trade-offs	Moves