DEXi is a computer program for multi-criteria decision making (Figure 1) developed by Marko Bohanec (Jožef Stefan Institute, Ljubljana).

This software is designed for the interactive development of qualitative decision models and the evaluation of options. It is useful for complex decision-making tasks, when it is necessary to select a specific option from a set of possible options in order to meet the decision-maker’s objectives and preferences.

Figure 1: Screenshot of the MASC model open in the DEXi software

For more information and to download

Tutorial on how to use DEXi

Structure of DEXi Models

DEXi enables the transparent and participatory design of multi-criteria decision trees, with weightings that allow for the expertise and/or preferences of stakeholders to be taken into account. Since the models are qualitative, they can be used ex post (to diagnose situations) or ex ante (to choose among several options, particularly among several innovations).
These trees are hierarchical structures that break down the decision problem into subproblems, which are less complex and therefore easier to solve than the complete problem. Each subproblem is then represented by an evaluation criterion or input criterion (“tree leaf”) assigned a qualitative value (such as “low,” “medium,” or “high”). However, some input criteria may be assigned quantitative values (e.g., yield). The classes for these quantitative criteria are then defined using threshold values (Figure 2). The IZI-EVAL tool can be used to generate these threshold values, drawing on data from the Agrosyst information system, which is dedicated to characterizing innovative cropping systems on farms in the DEPHY network.

Figure 2: Discretization of quantitative results into qualitative categories (Source: DEXifruits User Guide)

Aggregate criteria (tree nodes), located upstream of these input criteria, progressively combine the information contained in the lower-level criteria on which they depend, weighted according to their importance, until a single evaluation criterion is reached that allows for a comprehensive assessment. Aggregations are performed for each criterion using “utility functions,” which take the form of tables populated by experts based on qualitative “if-then” reasoning, such as:
IF <criterion 1 is very low> AND IF <criterion 2 is low to moderate> THEN <aggregated criterion is very low>
When used in multiple locations within the same tree, the criteria (input or aggregated) are said to be “linked.”

The IZI-Eval tool also enables statistical analyses to assess the sensitivity of models built with DEXi. This allows for optimizing the tree structure and determining the required level of precision for the input data.

A Variety of Tools for Agriculture

DEXi has been used, among other fields, in pharmacy and medicine, and—since the early 2000s—to develop models and decision-support tools for agricultural issues: assessing the sustainability of innovative systems in various production chains, designing diverse grassland ecosystems, estimating biotic pressure on a crop, selecting tillage strategies, and more.

There are models and tools used for ex ante or a priori evaluation to aid in system design prior to implementation. These include, in particular, the DEXiPM arable crop and DEXiAF tools.
There are others used a posteriori to evaluate existing systems—MASC and DEXiFruits, for example.
Some tools and models can be used both ex ante and ex post.

See the DEXi tools available on the MEANS platform