A multiple criteria socio-technical approach for the Portuguese Army Special Forces recruitment

Abstract

Cat-SD (Categorization by Similarity-Dissimilarity) is a method developed for handling nominal classification problems in the context of Multiple Criteria Decision Aiding (MCDA). This paper describes the design and implementation of this method and an application dealing with a recruitment and selection process in the Special Forces of the Portuguese Army. Besides, it proposes interaction protocols to elicit the preference parameters of the method to facilitate the construction of a decision model when the analyst guides the decision maker. Cat-SD has been implemented in DecSpace, a user-friendly online platform for supporting decision aiding processes using one or more MCDA methods simultaneously. The study related to the Portuguese Army Special Forces recruitment and selection demonstrates how these protocols and a tool like DecSpace can facilitate the process of applying the method in real-world scenarios.

Appendix

DecSpace : Current main features and related functionalities

• Types of Users There are four types of users with different permissions:

  • 1. Developers They can implement and add new MCDA methods;

  • 2. Administrator She/he manages all users and projects, having permission to modify or delete any object;

  • 3. Anonymous users They can test the platform and explore it with the restriction of having temporary projects (they cannot be saved);

  • 4. Registered users They have an account and their work is persistent, that is, the users can create their own projects, having access to all the features that DecSpace offers.

• User registration and login To use the platform a user can be registered, providing an e-mail address and a password (a username is created based on the user’ e-mail), or login to access the personal area, if the user is already registered. It is possible to enter as a guest, by clicking on “Try it now!” (see Fig. 10 that displays the DecSpace homepage.

• My Projects In this area, a registered user can create projects (with a click on the “New Project” button), choose whether the project is public or private, and manage the projects, using a set of functionalities. Each created project has associated the following: Name, Privacy, Last Update, Created, and Actions. Besides the functionality of creating new projects, the following actions are available for each project: Open Project (go to the “Workspace”), Duplicate Project (create a project equal to the selected project), and Delete Project (permanently eliminate the project, after confirmation by the user). In addition, the default order of visualizing the list of projects can be changed by inverting the alphabetical order of the project name, and the number of rows per page can be chosen among the four options (5, 10, 25, and All);

• Settings In this area, it is possible to change the username and the password (exclusively available in the personal area of registered users);

• Public Projects This area contains all public projects that are shared by users (available to everyone), with the following information for each project: Name, Owner, Last Update, and Created. The same possible Actions as in “My Projects” are available (open, duplicate, and delete), as well as the possibility of changing the project order by inverting the alphabetical order by “Name” or “Owner”, and choosing the number of rows per page;

• Methods All the available methods are in this area. Each method has a short description, an example, and a step-by-step explanation, intended to provide some helpful information to the user;

• FAQ It contains commonly asked questions and the respective answers related to some features of the platform;

• Workspace This is the area where the users construct workflows by dragging and dropping, and properly connecting MCDA methods modules and data modules (“boxes”) in an intuitive graphical user interface. The following actions are available: Execute Workflow, Save Project, Import Data, Method Selection, Delete Workflow, and Project Menu (see Fig. 9). The data and the preference information can be manually provided in the methods modules. It is possible to import and export CSV and JSON files. A .zip file can also be imported, containing a workflow that was already used. The workflows can be executed, saved, and deleted.

Fig. 10

DecSpace homepage (Pre-Alpha Version)

Information needed for building a Cat-SD workflow in DecSpace

• (a) Data

  • Criteria In this table, seven columns appear by default:

    • 1. Name It is the name of the criterion;

    • 2. Description It is the respective criterion description or some related information (it is not mandatory);

    • 3. Direction It corresponds to the preference direction (the user must choose one of the two possible options: “Maximize” or “Minimize”);

    • 4. Scale Type It is related to the kind of data of the criterion performance levels (the user must choose “Ordinal” or “Cardinal”);

    • 5. Min If the scale type is “Cardinal”, then a minimum value for the performance levels should be provided;

    • 6. Max If the scale type is “Cardinal”, then a maximum value for the performance levels should be provided;

    • 7. Num Levels If the scale type is “Ordinal”, then the total number of scale levels should be provided;

  • Actions Data related to the potential actions only include:

    • 1. Name It is the name of the action;

    • 2. Description It is the respective action description (it is not mandatory);

  • Performance Table The rows correspond to the actions and the columns corresponds to the criteria. Performance levels on each criterion must be provided for each action (the platform verify that they fulfill the criteria scales characteristics);

• (b) Preference information

  • Reference Actions For each action, the following information is needed: the name of the category to be considered (Category), the name of the reference action (Name), and then the performance levels have to be fulfilled in each criterion column;

  • SD Functions For each criterion a set of rows with a value within the range $[-1,1]$ (SD Value) and the respective performance differences for which the function takes the SD Value must be provided in a form of a mathematical condition (Condition);

  • Weights For each category and each criterion, a value of the criterion weight must be given, i.e., a set of weights per category;

  • Interaction Coefficients Firstly, the user can choose a category among the predefined categories that appear as the possible options (Category). Secondly, for such a category, a first criterion can be chosen among the options, i.e., the previously defined criteria (Criterion 1), and then a second criterion, among the remaining criteria has to be selected (Criterion 2). Thirdly, the type of interaction can be chosen among the three options (Type). Finally, a value for the interaction coefficient must be provided (Value). This procedure has to be followed for all interaction coefficients considered in the model. Alternatively, the data can be previously organized in a file, and then imported and adequately connected, as for the input data and the preference parameters. The platform checks the non-negative condition (see Eq. 2 in Sect. 2.2 and notifies the user, displaying an alert message box, in case of the values do not fulfilled the condition;

  • Likeness Thresholds For each predefined category (Category) a likeness threshold must be defined with a value within the range [0.5, 1] (Value).