4.- DESENVOLUPAMENT I CONTROL DE LES OBRES Replanteig de detall de les obres

Once our editor is generated, we have developed a plugin. This plugin allows eclipse to recognize our language and thus facilitate the creation of a program in our language by a user. Without giving the details of the creation of our plugin here we implement the process of deforestation of Laf as defined in the chapter 4 (Section 6.1) and according to the syntax presented in chapter 5. Figure 6.2 shows the computational model obtained (See appendix C for more illustration of code).

This program highlights the relationships between the activities of timber traders, forest administration authorities and farmers’ options, who are primary actors of defor- estation. It presents in terms of action and option the different choices available to each class of actor in his level. We also observe that the key words of our language have a different coloring to the other words.

6.4

The resulting TiC for Laf Forest Reserve

Main functionality of the TiC simulator

1. Modify sentence types: for example to implement a new policy for the authorities. 2. Modify the distribution parameters for auto-generated object types. For example,

the number of farmers, or the property that farmers dispose of.

3. Transforms the domain concepts in the target language while keeping the logical construction of these concepts.

4. Generates a general structure of the model source code.

The TiC simulator user interface

Figure 6.3 and 6.4 show a sample screenshot of an early generated NetLogo code of Laf model from the TiC simulator (See Appendix B for more illustration of code). We observe in this figure an early implementation of the Laf model characterized by processing of parameters, creating different classes of agents, creating the body of all the functions and procedures that define the actions of the agents, the initialization of simulation environment, and a general scheme for the main program. As we see, there remains only to user to complete the code functions and procedure according to the agent’s goals.

The TiC tool offers the user two big advantages:

Data consistency from data collection phase to the code generation is therefore carried

out by the domain expert. The data transfer of one competence to another is no more useful. This reduces data loss.

Gain in time since code is generated directly from the conceptual model.

In addition, by integrating environmental problem analysis frameworks in our ap- proach, we improve usability since we separate domain knowledge from technical issues. That way, a domain expert gets a greater opportunity to make effective use of the techniques proposed.

Moreover, we compared our approach with other existing metamodeling approaches according to the four criteria proposes by Ferdjoukh (2016): generating instances, editing metamodels (or models), user intervention and error detection. Table 6.5 summarizes the result of this comparison.

Figure 6.2: Model of Laf deforestation process obtained from the developed language domain

6.4. THE RESULTING TIC FOR LAF FOREST RESERVE 83

Figure 6.3: Example of generating NetLogo code for parameter declarations

Figure 6.4: Example of generating NetLogo code for function declarations.

Author Tool Criteria for comparison

Generating Editing User Error instances (meta)models intervention detection Automatic Manual No Yes No Edited Parameter Automatic Manual Morvan et al. (2010) IRM4MLS/ • • • •

SIMILAR

Marcenac et al. (1998) GEAMAS • • • •

Solovyev et al. (2010) SPARK • • • •

6.5

Conclusion

This chapter highlights the substantial results related to the application of the tool TiC in the description of the problem of Laf Forest Reserve deforestation in the study area. Some methodological lessons can also be drawn from this experience. From the methodological point of view, the lessons learned relate to the description of the different models proposed in the approach, notably the natural language and its conceptual equivalent. The main advantage of natural language is to allow the domain expert to better represent the observed process while using the concepts that will be easily understandable by the system analyst who in turn can easily represent them using the technological tools . Moreover, the results of this chapter confirm the interest of using a contextual natural language. By taking the concept of the domain as a starting point, we can clearly represent the problem of the domain using the appropriate concepts and lead to a transformation to a conceptual model and a code conforming to the process studied.

Chapter 7

Validation of the approach in the

Laf Forest Reserve case

In the previous chapters, we have presented the application of a Multi-Level Agent Based Model (MLABM) to describe environmental problems based on the Problem-in-Context (PiC) and AiC (Action-in-Context) framework to describe and explain agent motivations and co- operation. In this Chapter, we show how the associated TiC (Tool-in-Context) developed in chapter 5 can help the domain expert to generate a simulation tool. This simulation tool consists in transforming the defined language model into a NetLogo code model, thus facilitating an early prototype application to be used by the domain expert. The resulting application is used to (1) explain the process of deforestation and (2) to analyze scenar- ios for a more rational management of the space around the Laf-Madiam forest reserve

(Réserve Forestière de Laf-Madiam), located in the drylands of North Cameroon. We

discuss the prototype resulting from our approach and show how it allows the domain expert to focus on the social causes of environmental problems, taking into account the biophysical processes.