Content area
Issue Title: Special Issue: Part 2: Simulations, Norms and Laws
Management of the renewable natural resources in Madagascar is gradually being transferred to the local communities, particularly that of forest resources. However, these local communities are struggling to assess the consequences of management plans that they themselves must develop and implement on ecologically, economically and socially sustainable grounds. In order to highlight key aspects of different management options beforehand, we have developed MIRANA, a computer model to simulate various scenarios of management plan implementation. MIRANA differs from other simulation models by not only taking into account individual practices and economic exchanges, but also by accounting for the applicable regulations. These regulations are taken into consideration by means of a multiplicity of normative structures within a spatial context. The objective of this paper is to describe the representations of institutions, norms and territories proposed by MIRANA and to discuss these representations in relation to the state of the art in the field of normative multi-agent systems.[PUBLICATION ABSTRACT]
Artif Intell Law (2013) 21:4778
DOI 10.1007/s10506-012-9133-8
Sigrid Aubert Jean-Pierre Mller
Published online: 2 November 2012 Springer Science+Business Media Dordrecht 2012
Abstract Management of the renewable natural resources in Madagascar is gradually being transferred to the local communities, particularly that of forest resources. However, these local communities are struggling to assess the consequences of management plans that they themselves must develop and implement on ecologically, economically and socially sustainable grounds. In order to highlight key aspects of different management options beforehand, we have developed MIRANA, a computer model to simulate various scenarios of management plan implementation. MIRANA differs from other simulation models by not only taking into account individual practices and economic exchanges, but also by accounting for the applicable regulations. These regulations are taken into consideration by means of a multiplicity of normative structures within a spatial context. The objective of this paper is to describe the representations of institutions, norms and territories proposed by MIRANA and to discuss these representations in relation to the state of the art in the eld of normative multi-agent systems.
Keywords Multi-agent simulation Institutions Norms
Environmental modelling Agent behaviour models
S. Aubert J.-P. Mller (&)
CIRAD-GREEN, Campus International de Baillarguet, 34398 Montpellier Cedex 5, Francee-mail: [email protected]
S. Auberte-mail: [email protected]
S. Aubert
ESSA, Ampandrianomby, 853-99 Antananarivo, Madagascar
Incorporating institutions, norms and territories in a generic model to simulate the management of renewable resources
123
48 S. Aubert, J.-P. Mller
1 Introduction
Management of renewable natural resources (RNR) in Madagascar gradually is being transferred to local communities, particularly the management of forest resources. Conditions for contractually managing the States forests are dened by the GELOSE legislative and regulatory framework. Legislation 96-025 of 30 September 1996, pertains to the secure local management of renewable natural resources and its implementation directive no. 200112 of 14 February 2001, pertains to the contract-based management of forests.
In this context, the contractualized management of forests is based on a decentralized system including controls and sanctions, as well as economic incentives. Natural resources use, logging and conservation are organized through a zoning of local community territory. However, given the complexity of the system, local communities are struggling with the challenges of developing management plans and with foreseeing the consequences of their implementation (Montagne et al. 2007). In addition, there is no consensus among those either directly or indirectly involved with RNR management regarding the comparative advantages of unregulated exploitation, regulated exploitation and pure conservation of timber resources for any given piece of land (Porter-Bolland et al. 2011). Local communities involvement in biodiversity conservation thus is being called into question.
In order to highlight the benets (or disadvantages) of possible local management options, we have developed the MIRANA computer model (Aubert et al. 2010), which simulates at a local scale the implementation of RNR management plans according to various scenarios. This model renders it possible to test a set of local natural resource management tools in various contexts for a given scenario. It assesses the effectiveness of these tools by observing simultaneously the impact of human actions on the forest ecosystem, on meeting the basic needs of the local community and on the creation and distribution of incomes resulting from forest-related activities. We therefore consider that the model is capable of treating ecological, social and economic sustainability at the same time.
MIRANA differs from most other simulation model programs of this type (Parker and Filatova 2008; Farol et al. 2010). Through a multiplicity of differentiated normative structures, it not only takes into account individual cases of natural resource harvesting, people settling on land plots, and related economic activities, but also the regulations likely to be applied to these activities.
From a methodological point of view, we distinguish between the development question of concern and the resulting research questions. The development question is:
What is the impact of the local communitys management plan on the ecological, social and economic sustainability of the local socio-ecological system (SES)?
The resulting research questions concerning computer modeling are:
How may the various regulatory mechanisms with their multiplicity of overlapping levels be represented?
123
Incorporating institutions, norms and territories in a generic model 49
How are these regulations considered during the decision-making process, both at individual and collective levels?
The objective of this paper is rst to present the conceptual analysis we carried out, and second to present the resulting formal account of this conceptual analysis using concepts developed in the eld of multi-agent systems.
In the following sections, we outline our modeling methodology as a cycle of conceptual modeling, formalization and implementation. Next, we introduce the conceptual model that describes the system with the notions devised by the specialists in the eld involved in the modeling process. We then present the formalization using multi-agent systems concepts. Terminological distinctions between the conceptual model and the formalization are proposed to avoid any confusion among the levels of description. For example, we distinguish the stakeholder in the conceptual model from the agent in the formal representation. Finally, we conclude by presenting the achievements and future perspectives of our research.
2 Methodology
Sociology, anthropology and philosophy works have been sources of endless inspiration and metaphors for multi-agent systems design, as they always have been for articial intelligence. However, the imported concepts acquire a technical meaning in the multi-agent systems eld that does not always retain the semantic subtleties of the original eld. What is a useful source of innovation can also be a source of confusion when the goal is no longer to use sociological metaphors for designing multi-agent systems, but to model as precisely as possible social and human sciences theories in order to be able to run useful simulations. To overcome this risk of confusion, we propose to distinguish between the theory as expressed by the specialists of the eld to be modeled, the model using proper multi-agent systems concepts and formalisms, and the implementation of this model in a computer program (see Fig. 1). We furthermore propose to make this distinction by using different vocabularies wherever possible for each of these three levels.
The theory is the explanation of how things work from a thematicians point of view. It is mainly expressed in natural language (even in physics). The proposition of Mller (2007) is to use ontologies to represent the content of the theory as expressed in natural language. An ontology in computer science is the specication of a conceptualization of a domain (Gruber 1993). It is specied, i.e. formalized, using description logics. In the following, however, we shall use unied modeling language (UML) class diagrams as advocated in Bommel and Mller (2007). UML class diagrams provide a standardized graphical representation that uses squared boxes for concepts and attributes and arrows for taxonomies and relationships. UML
Fig. 1 The three level modeling process
123
50 S. Aubert, J.-P. Mller
class diagrams have been proved to be easier to understand than mere unstructured lists of logical formulas. This formalization of the theory, using ontologies, will be called the conceptual model.
The model is expressed in a formalism or a set of formalisms such as differential equations, automata, multi-agent systems (as long as they do not reduce to a programming framework), etc. In our case, the multi-agent systems community provides a number of formalisms of multi-agent systems in general, agent architectures of various sorts (belief-desire-intention, economic agents, etc.), up to the normative (Boella and Van der Torre 2004) and institutional (Dignum 2004) multi-agent systems. In this paper, we will provide an adaptation of multi-agent systems formalisms that are capable of expressing a thematicians theory. The formalization of the conceptual model, using multi-agent systems, will be called the formal model.
Finally, the program is the implementation of the model. This implementation is expressed in its own formalism, i.e. a programming language or any language for which an operational semantics is dened. We will only dene briey how the implementation was made.
We now will formally introduce the notion of ontology because we will use it both to represent the conceptual model and as a formalism at the model level to formalize the notions of institution and agents. For the latter task, we also need to introduce the contextual ontologies to represent the multiplicity of points of view introduced by legal pluralism (Merry 1988).
3 The contextual ontologies
We dene the contextual ontologies by introducing a family of ontologies Oi. Each individual ontology denes a language Li Ci; Pi; Oi; Ii
h i where: Ci is a set of concept names; Pi is a set of relation names1; Oi is a set of individual (or object) names; Ii is a set of ontology names.
We obtain the derived concepts by the usual constructors: :c; c1 [ c2;
c1 \ c2; 9r:c; 8r:c; i : c where c; c1; c2 2 Ci; r 2 Pi and i 2 Ii. i:c denotes the con
cept c in the ontology i, which allows one to designate concepts dened in other ontologies.
As mentioned in Sect. 2, UML is sometimes more understandable than a list of axioms. Figure 2 illustrates some equivalences using boxes to represent concepts, white triangle arrows when deriving concepts, and simple arrows or attributes for relations. Figure 2 e) illustrates a concept dened by the list of its individuals, called an enumeration.
1 In descriptive logics, they are called roles, but we do not use this term here to avoid confusion when roles is used in the institutional sense.
123
Incorporating institutions, norms and territories in a generic model 51
Fig. 2 Some equivalences between description logics and UML. a C _
A _
The terminological axioms of denition and subsumption, of the form c1 _
c2 and c1 c2, form the set Ti, called the Tbox (terminological box).
The assertional axioms (or assertions) of the form c(o), o1 = o2, and r(o1, o2) where c e Ci, r e Pi and o, o1, o2 e Oi or of the form i:o where i e Ii, form the set Ai, called the Abox (Assertion box). i:o designates the individual o in the ontology i, which allows us to designate individuals dened in other ontologies. Hence, the equality axiom allows one to express the same individual under different names in different contexts. As we need to express quantities of resources, and not only individual objects, we add to the usual names the pairs c; q
h i where c e Ci is a
concept and q is a quantity. Thus, we can designate ve trees Arbre; 5
h i or 2.5 kilos
of rice Rice; 2:5 kg
h i. A formal account of this extension can be found in Mller
et al. (2011).Finally, we dene an ontology as a triple Oi Li; Ti; Ai
h i where Li is its
language, Ti is the set of terminological axioms and Ai is the set of its assertions. Additionally, given an axiom ax, we note Oij ax if ax is deducible from the
axioms of Oi.
The semantics of a family of ontologies Oi is dened by giving a family Mi
Di; pi
h i of local interpretations,2 where: Di is the domain of discourse; pi is the semantical function dened as follows:
pi c 2 Ci
Di
pi r 2 Pi
Di Di
pi o 2 Oi
2 Di
pii 2 Ii 2 fMig
pi:c fx 2 Dij:x 2 picg
pic1 [ c2 fx 2 Dijx 2 pic1 _ x 2 pic2g
pic1 \ c2 fx 2 Dijx 2 pic1 ^ x 2 pic2g
pi9r:c fx 2 Dij9y; x; y
h i 2 pirg
pi8r:c fx 2 Dij8y; x; y
h i 2 pirg
pii : c pjc \ Di where pii Mj
This last denition allows us to dene the semantics of the reference with an expression in another ontology. It depends on the possibility to locally name this ontology and to at least partly share the domain of discourse.
2 It is mainly this locality that denes the contextual feature of these ontologies.
A \ B, b D A [ B, c A _ 9r:B, d
8r:B, e E _
a; . . .; z
f g
123
52 S. Aubert, J.-P. Mller
Finally, an interpretation Mi is a model for an ontology Oi under the following conditions:
Mij c1 _
c2 if and only if pic1 pic2;
Mij c1 c2 if and only if pic1 pic2;
Mij co if and only if pio 2 pic;
Mij o1 o2 if and only if pio1 pio2;
Mij ro1; o2 if and only if pio1; pio2
h i 2 pir:
The semantics is made slightly differently than in Grossi (2007) where the semantics of an axiom is the set of all of its possible models. It is easy to prove that it is equivalent.
In the next section, we mainly will use the UML equivalent of the ontological axioms to represent graphically the discourse we elaborated to represent various regulatory mechanisms and to assess them through the impact of the management plan that we simulate. We then also will use the contextual ontologies to account for this discourse in formal terms.
4 The conceptual model
To account for the impact of a management planput into operation by a local community (VOI for VondronOlona Ifotony in Malagasy)on a given socioecological system (SES) (Rives et al. 2012), the MIRANA (Aubert et al. 2010) conceptual model was designed around four primary notions: institutions, stake-holders, resources and territories. This conceptual model was developed according to the theory of the specialists in the eld (the so-called thematicians), in particular, legal anthropologists. It also was inspired by the work of Ostrom, which reects an institutional economics perspective. Ostrom proposed a conceptual framework founded on notions of governance systems, users of natural resources, resource system and resource units (2009).
In the MIRANA conceptual model, the institutions and stakeholders respectively constitute the macro and micro levels of description. The concept of institutions allows us to describe the behavior of groups of agents who, through members of these groups, come to comply with and/or recognize a given set of norms. In consequence, they assume one or more roles while carrying out their activities. The stakeholders respectively represent individuals (natural person) or legal entities (moral person) performing their daily activities (Mller and Aubert 2012). We will present these concepts in turn before introducing the resources and, in particular, the territories which are shaped by the stakeholders and institutions.
4.1 The institutions
For Kirk (1999), who takes a legal perspective, an institution encompasses both rules and organizations that shape and enforce these rules. For many authors, however, an institution is the set of the humanly devised constraints that shape political, economic and social interaction (North Douglass 1991). These
123
Incorporating institutions, norms and territories in a generic model 53
constraints do not need, and generally are not, deliberate creations. An institution is the context in which people learn their roles. This notion is even used to designate not only the regular, patterned behavior of people in a society, but also the ideas and values associated with these regularities (Neale Walter 1994). Searle (1995) insists on the importance of both constitutive (or denitional) and regulative norms in institutions. Ostrom proposes several distinctions in Crawford and Ostrom (1995) and Ostrom (2005). First, she makes the distinction between an institution as dened above and an organization as a concrete realization of an institution by a group of people. Moreover, she makes a distinction among institutions-as-equilibrium, institutions-as-norms, and institutions-as-rules. An institution-as-equilibrium is composed of strategies obtained from individual optimizing behaviors. An institution-as-norms is composed of deontic modalities on strategies (what ought to be performed or not) and institutions-as-rules add the expectation that proscribed behaviors will be sanctioned. These denitions introduce a norm as a principle of right action binding upon the members of a group and serving to guide, control, or regulate proper and acceptable behavior (Merriam-Webster Online Dictionary). Generally, we consider that a juridical norm is concretized by rules which dene explicit sanctions. Finally, the Italian lawyer, Santi (2002), associates the institution with the notion of legal orders, and develops the idea of multiple, co-existing legal orders. Legal anthropologists furthermore consider three different normative orders to capture regulation mechanisms in a given society: State law (including essentially written rules), customary law (including essentially oral rules), and empirical/practical law (including essentially informal habits) (Le Roy 1999).
In MIRANA, we decided to consider an institution as a set of constitutive and regulative norms. The organization is a representation of a concrete group of people on whom the institution applies. We sometimes associate organization to a moral person. This moral person reies, and is in charge of, the activity of enforcing and shaping the behavior of people concerned by at least one institution. We chose these denitions to model social regulation as a mechanism regardless of the normative orders it realizes, and regardless of the way the enforcement is undertaken.
Therefore, an institution is made of regulative norms expressed in a vocabulary dened by constitutive norms (Fig. 2). The norms generally are materialized in rules that can emanate from collective settlements (imposed on everybody), agreements (imposed only on people concerned by the agreement) or decisions (imposed only on people concerned by an authority decision). Each previously identied normative order uses more or less these three modalities. Thus, the norms introduce the constraints and/or the opportunities affecting the realization of stakeholder actions carried out to achieve their goals. Accordingly, a community of individuals adopts traits of group behavior when dealing with an institution, such as the market (legal or informal) or when establishing a rice eld in an area where such activity is prohibited. In both cases, the institutions enable us to dene the roles that the stakeholders, as subjects of normative orders, will assume. The notion of institution is thus associated with an organization, which facilitates the emergence of a coherent set of norms within a given society (these norms are not necessarily consistent when they originate from two distinct organizations). A stakeholder may belong
123
54 S. Aubert, J.-P. Mller
Fig. 3 The denition of an institution
simultaneously to several organizations that are recognized as regulatory bodies by
those third parties with whom he may or may not have connections (Fig. 3).
With the MIRANA conceptual model, because each type of institution denes its own constitutive norms, in the form of a specic ontology, an entity can belong to several categories. The structure of these institutional points of view with ontologies corresponds to Searles framework (1995) in its construction of social reality. Searle explains how reality is socially appropriated by introducing how parts of a system can count as an instance of given categories from the point of view of an institution. These categories dene both the way an entity is perceived and, simultaneously, the roles it plays. In our case, an area can be considered as all or part of a forest (for ecologist scientists), State land (for forest administration), a protected area (for park administration), a municipality (for territorial administration), a community associations territory (for VOIs) or customary land (for traditional communities). Similarly, a tree can act as a keystone species (for ecology scientists), as a principal product of a forest (for forest administration), as a target species for conservation (for park administration) Finally, a stakeholder successively can take the role of lumberjack (licensed or illegal), inhabitant, member or user. Therefore, our ontologies are the denition of how stakeholders, resources and space can count with respect to an institution.
Within an institution, a norm establishes a relationship between two roles that could be attributed to subjects of normative orders (the stakeholders) or to objects of normative orders (the resources) as illustrated in Fig. 4. In our conceptual model, status thus is established by the connection between the subject of normative orders and an institution. This rst connection legitimizes the action (or use) of an individual, belonging to a community of individuals, who recognizes the institution in question, provided that it concerns an object of normative orders (support) as previously dened. The denition of the object of normative orders is established by means of the role that it is attributed by an institution, but also through its location on the considered territory (Boaventura De Sousa 1987). This description of a norm corresponds to the ADICO framework as proposed by Crawford and Ostrom (1995). ADICO stands for A as Attributes that correspond to our roles, D as Deontic, I as Aim that corresponds to our activity, C as Condition, and O as Or-else specifying
123
Incorporating institutions, norms and territories in a generic model 55
Fig. 4 The norm as a relation between two roles
the sanction. In our case, the conditions are not explicit but will be represented in our formalization. The Or-else could take the form of a ne or conscation if dened by an explicit rule, but this is not always the case. Indeed, during the simulations, we are interested rst in seeing whether rules are applied or not; not if sanctions (that depend on controls) are effective.
If regulated by a moral person, an institution provides not only a regulatory framework, but also the means to enforce it. The control function must, however, be distinguished from the sanction function. The latter should only be called upon in the case of norms being violated. These functions can be generally achieved through social control (Castelfranchi 2003), or explicitly attributed to a stakeholder who, thanks to the authority delegated by an institution, assumes responsibility to achieve the control objectives.
Thus in this conceptual model, the forest administration, park administration, commune, as well as the VOI and traditional communities (organized around the notion of lineage in the Malagasy case), respectively constitute the institutions of an authoritarian structure, realizing several normative orders and power structures of constraints (Fig. 5). This gure illustrates another UML convention: the use of the dashed line for the instance of relationship (the VOI is an instance of a legal institution). In the context of MIRANA, it is worth noting that the object of study for the legal anthropologist concerns the socio-ecosystem, based on the laws of biodiversity (Aubert 2009; de Sadeleer and Charles 2004). Therefore, it also is necessary to represent the ecosystem with its own natural laws. The ecosystem also would guide stakeholder actions. We decided to view the ecosystem itself as an institution, insofar as the relevant units of the ecosystemthe natural resources (representing animal or plant species)are also subject to natural laws (reproduction, migration, distribution etc.). This representation enables stakeholders to understand ecological dynamics from the ecologists point of view. In Fig. 5, we also mention employment contracts as temporary institutions between the VOI and the stakeholders for production or surveillance, and the illegal market as an empirical institution.
In the following section, we will use the MIRANA conceptual model to assess the impact of a management plan adopted by a VOI in a context of legal pluralism
123
56 S. Aubert, J.-P. Mller
Fig. 5 Some institutions
(Le Roy 2012). This use of the model will determine whether the opportunities, or limitations, of the norms of the adopted management plan take into account the existence of normative systems produced by cooperative or competitive institutions. It is therefore by the means of normative systems that one can assess the sensitivity of the SES to the new elements introduced by the development plan.
4.2 The stakeholders
For MIRANA, the stakeholders are exclusively active, human, individual or collective, decision-making parties with objectives. A tree cannot be a stakeholder, although its interests can be defended by an NGO as a stakeholder. Stakeholders act to meet their basic needs (objectives dened in the context of a subsistence economy), to make prots (in a market economy) or to manage natural renewable resources (i.e. to achieve collective goals). Like institutions dening their norms, a stakeholder denes his own ontology to describe his activities (Fig. 6).
From a legal anthropologists point of view, all of these stakeholders are subjects of normative orders. A stakeholder can be a human being acting as a natural person or represent a legal entity as a moral person. In our case, we dene a household unit as an individual stakeholder (natural person). Even if the household is made up of individuals, we consider them to be a single decision-making unit rather than taking into account the behavior of each member. Households are classied according to three types: migrant households, households established in the forest, and those established on marshlands. The previously identied institutions are associated with legal entities (moral person) as illustrated in Fig. 7. The legal entities considered are: local community associations (VOI), the forest administration (State administration), the commune (territorial administration) and the park administration. A legal entity always is associated with an institution that it manages. It should be noted that we did not identify a customary authority in our example, although we could have done so. In this case, customary authority would have been a legal entity from the point of view of the customary normative order. Each legal entity has different objectives, related but not reducible to the collective goals of the associated institution, and organizes its activities accordingly. In addition, it enforces the norms of the associated institution using social control up to control and sanction (Conte and Castelfranchi 1999).
123
Incorporating institutions, norms and territories in a generic model 57
Fig. 6 The stakeholder structure with some considered activities
Fig. 7 The stakeholder structure with some moral people
For an ecologist, a species (population) or an individual animal or a plant also could be considered to be a stakeholder. However, we have chosen not to incorporate this option as it is not sufciently relevant to our question. It is not necessary for us to consider the objectives of those that win the competition to reach the light or those who associate with others in order to attain the necessary nutrients for survival to portray the ecological dynamics that we consider through the MIRANA model.
The impact of the management plan adopted by the VOI will be assessed according to the actions that the stakeholders accomplish to achieve their objectives (realize contract, subsistence or protability). These objectives, however, may diverge from the management objectives identied within the VOI framework. This dimension of the model allows us to evaluate the ecological, social and economic constraints to which the management plan (adopted by the VOI) is subject. Beyond the rules promoted by the decentralized management system, this dimension makes it possible to estimate the vulnerability of the SES to unforeseen threats (Rives et al. 2012). To satisfy their objectives, all stakeholders can indeed do something illegal
123
58 S. Aubert, J.-P. Mller
from the point of view of one institution or another, even if they rst try to respect the rules of the organizations they belong.
4.3 The resources
The resources are material objects (animals, plants, ores, money, land, etc.) or intangible assets (knowledge, rights, labor). Objects of patrimonial laws, they can be counted and be potentially valued in monetary terms.
In our conceptual model, the resources could be attributed the role of object of normative order according to the norms of the various institutions and according to the activities that the subjects of the normative order wish to accomplish. Thus, relative to an institution, a rosewood tree could be considered to be a possession or a property. Similarly, a plot of land could be seen as an object of use, an area holding certain resources which may be seen as possessions, or even as property, depending on the normative act considered. Thereby, that plot may be subject to a number of restrictions related to its settlement. Relative to the stakeholders activities, the same rosewood tree can take the role of timber for construction, rewood, or as an individual member of the rosewood population, depending on the situation. Therefore, the constitutive norms of the institutions and the ontologies of the stakeholders dene the roles that can be attributed to the resources. Dening the notion rewood in an ontology consequently introduces the possibility for a given resource to count as rewood from this ontology perspective (Searle 1995).
The resources exist as sets and/or quantities, i.e. in resource stocks or populations. Accordingly, they can be created, destroyed, shifted from one stock or patrimony to another and are themselves transformed by means of stakeholders activities. These resources may also have their own dynamics (they reproduce, distribute themselves, disappear and grow) according to their assigned roles and their connections from an ecological perspective.
The resources we use in MIRANA are illustrated in Fig. 8. Note that the stocks of material resources are situated on land. The immaterial resources are associated with stakeholders. The land plays an important role not only as a support of the material resources, but also for the activities of the stakeholders and for the regulations of the institutions. In Fig. 8, resources are decomposable at will depending on the descriptions it supports, thus allowing for multi-scale structures. The networks are considered as pieces of land with a particular (linear) topology and differentiated functions.
4.4 Land and territories
The notion of territory is highly ambiguous. It refers generally to a geographic area under the jurisdiction of a governmental authority, or, more simply, an administrative subdivision of a country (Merriam-Webster online dictionary). However, in geography, broader denitions are used, such as:
Layout of material and symbolic resources able to structure the practical conditions of existence of an individual or of social collective, and to inform
123
Incorporating institutions, norms and territories in a generic model 59
Fig. 8 The resources in MIRANA
back this individual or this collective on its own identity (translated from Debarbieux 2003, p. 910)
or the following, from political science:
Territoriality will be dened as the attempt by an individual or a group to affect, inuence or control people, phenomena, and relationships, by delimiting an asserting control over a geographic area. This area will be called the territory (Sack 1986, p. 19)
Here, we adopt the broader denition of a geographic area delineated by the activities of a stakeholder or the norms of an institution when the norms are dened over spatial areas, as is the case for territorialized institutions.
In our case, we only take into account those territories associated with previously designated institutions. While it is conceivable that an institution may not be attached to a territory, such as an NGO or an association, we do not consider this case scenario with MIRANA. All of the institutions that we consider are territorialized. Additionally, a territory may be composed of several distinct areas or zones.
Being a territory or a subdivision thereof is a role attributed to a piece of land as a particular resource by an institution. As such, the constitutive norms of an institution regarding land or pieces of land describe the territorial structure related to the institution. Figure 9 illustrates the territories dened in MIRANA. The ontology of an institution thus always includes the concept of territory that can be decomposed in various zones like conservation areas or building lands. These zones can themselves be composed of land plots. Consequently, the institutions norms apply to these zones as well.
Fig. 9 The territories of MIRANA
123
60 S. Aubert, J.-P. Mller
As a consequence, in the same manner that a stakeholder, in the social dimension, can participate simultaneously in several institutions, a stakeholder could be located in a number of territories at the same time. For example, a household might settle somewhere and realize it can carry out its activities simultaneously at home (in the house or eld), on a customary territory (that of his or anothers lineage), on the VOIs land, on communal land, or within a State Forest etc. While performing his activities, an individual therefore is subject both to the norms of the institution to which he belongs and, in addition, to the norms of the institutions claiming the territories on which the individual carries out an activity.
4.5 Illustration
To illustrate the notions of an institution, a territory, a stakeholder and a resource used in the MIRANA model, we employ four distinct concepts that enable us to depict the interactions that establish themselves between the various SES entities depending on the management scenarios envisaged by the VOI:
The VOI as an institution, bringing together several members of the local communities (identied as members of the VOI),
The VOIs territory, that is to say the land that is concerned by the management transfer and,
The VOI as a stakeholder, and legal entity (moral person) responsible for signing the contract to transfer management control of forest resources. This contract is established with the forest administration and the commune. Moreover, the VOI is in charge of implementing the regulations.
The VOI institution as a resource allowing members to acquire some rights also can be considered.
The VOI is legally considered in State law as an ofcial decision-making body for which conservation and sustainable development are primary objectives. This body can dene zoning boundaries, establish quotas, and grant permits and employment contracts. Furthermore, they may establish contracts of payment for environmental services, levy taxes, carry out inspections and sanction punishments as necessary.
From the notions of institution, stakeholder, resource and territory, the impact of a management planadopted by a VOIcan consequently be assessed from an ecological, social and economic angle by means of our computer model. This model incorporates the decision-making process of the individuals (natural persons) concerning the renewable natural resources in a context of legal pluralism.
4.6 The dynamics
How to represent the dynamics of the system starting from the thematicians discourse, and in a way that is understandable to him, is an open question. An ontology is clearly not enough to do so. In the following, we will use a mixture of natural language descriptions, UML activity and swimlane diagrams to make some descriptions more precise, in particular in terms of interactions. The biophysical
123
Incorporating institutions, norms and territories in a generic model 61
dynamics is limited to the species populations and fertility. Species population dynamics are limited to continuous growth up to a maximum that depends on the species and the land cover. No migration is taken into account in our model. Fertility also grows up to a maximum and declines when the corresponding land plot is cultivated. We now shall describe in more detail the behavior of the households and the VOIs.
4.6.1 Household behavior
Households are characterized by an available workforce and a set of annual needs. These needs include quantities of food, money, rewood (for cooking and heating), construction wood, medicinal plants and so on. Each year, each household plans and executes its activities (see Fig. 10).
A household starts its cycle by trying to sell all or part of its workforce by requesting contracts (contract request in Fig. 9) from the VOI. The planning thereafter is composed of three phases:
1. If the contract request is accepted (get request), it receives one or more contracts (contracts) for lumber jacking, planting or controlling to detect possible norm violations. The household consequently has to plan the related activities and evaluate the remaining workforce. The objective is to sell its workforce to earn income to cover some of the households needs.
2. Then, the household plans how to satisfy its needs given its available workforce. The Usage permits regulate access to the resources needed to satisfy these needs. Therefore, the household asks for the number of permits that match the available quantities of resources in the area considered. The objective is to meet the households needs;
3. Then, if some workforce remains, the household plans the production of goods to sell on the market. Here also, exploitation permits regulate production and, consequently, are requested. Here, the objective is to maximize household income.
The three phases produce sequences of actions to perform. These actions are added to a global household plan (global plan). Notice that the behavior of the households does not reduce only to income optimization because we take into account two additional important dimensions of human behavior:
The possibility of selling the households workforce through employment (all contract offers are accepted, even if opportunity costs are not relevant);
Auto-consumption that is not based on optimization but on satisfaction only.
After this planning phase, the planned actions are executed and the results delivered to the employer, consumed or sold depending on whether they were produced for contracts, satisfying needs, or sale. The employee gets paid on delivery and the production sold on the legal market is submitted to a tax. At the end of year, every resource that has not be delivered to the employer or been consumed is converted into money by being legally or illegally sold, and constitutes the annual nancial result of the household.
123
62 S. Aubert, J.-P. Mller
Fig. 10 Household behavior
We now describe the regulation of the households activities by the institutions. However, beforehand, we shall make three remarks:
1. Each contract constitutes itself a small institution with a limited duration (1 year in our simulations). Each contract denes the role of employer and employee with the associated rights and duties in terms of the delivery of goods or services and the payment. In our case, the contracts are made with the VOI, which delegates the role of license holder for lumber jacking, the role of police for control and the role of environmental service provider only to its members;
123
Incorporating institutions, norms and territories in a generic model 63
2. Part of the regulation is achieved externally by a control mechanism. The households in charge of control dedicate a part of their time to monitoring the actions of others. If a violation is observed, a ne is applied and the resulting resources are conscated and given to the VOI. Actually, we consider that surveillance activity uses up some of the workforce and increases the probability of a violation to be detected in a given territory.
3. Each household in its decision mechanism internally achieves the other part of the regulation. The result depends on whether the household is legalist or not and will be described hereafter.
At the planning level, each activity has to take place in a certain land area. Therefore, part of the planning phase consists in choosing a place to carry out the activity. The place to be chosen depends on whether the household is legalist or not. If the household is legalist, the activity can only take place on a land plot where it is authorized from the points of view of all of the dened institutions. This authorization depends on the norms applicable to the corresponding territories or zones concerned by them. If the household is not legalist, it may consider carrying out the activity in places that are not allowed from the point of view of one or more institution. The characterization of the households behavior is decided before the simulation to give the possibility to test the management plan in the (improbable) situation that all households will respect the rules establish by the VOI.
At the execution level, the execution of the planned actions to satisfy the needs or the commercial production depends on the various permits issued by the VOI. If the permit is not granted and the household is legalist, the corresponding action will not be executed, otherwise, it will. If the action is illegal and the violation is detected, a ne has to be paid and the corresponding resources are conscated.
This behavior will allow the checking of the impact of the imposed regulations on nancial results (economic sustainability) and household satisfaction (social sustainability). If all of the households are strict legalists, the level of satisfaction of annual needs will be a good indicator of the sustainability of regulations. If none of the households are legalist, the number of violations (detected or not) also will constitute a good indicator of the pressure imposed by regulations. Another indicator could be the relative importance of the goods sold on the formal or informal market.
4.6.2 VOI behavior
The VOI has the objective, through its associated institution, to guarantee a sustainable use of the renewable resources on its territory. As a stakeholder and moral person, the VOI is in charge of implementing the norms of the corresponding institution. It assumes the role of manager within the VOI institution. It is very important to understand that the VOI as an institution denes the norms, and the VOI as a stakeholder denes their implementation. This implementation relies on a number of tools:
granting lumberjacking contracts and exploitation licenses to implement the exploitation quotas (the quota is assumed to be dened on the basis of the renewal speed of resources renewal);
123
64 S. Aubert, J.-P. Mller
granting usage licenses to implement the usage quotas; granting plantation licenses to compensate for forestry resource losses, and consequently to restore the ecosystem;
grants for intensication of cultivation to increase crop productivity and possibly reduce the footprint on the ecosystem;
granting control contracts to implement the norm compliance by the households.
Finally, the VOI ensures its own nancial sustainability by gathering nes and taxes, as well as by selling the contracted production and conscated goods.
This behavior is summarized in Fig. 11 where no sequential order is given to activities because most are triggered by the arrival of requests, the order therefore is unimportant.
At this level, it is possible to parameterize the regulation policies by the institution norms, the quotas and the implementation policy and to assess the feasibility of the management plan. Therefore, we are globally able to assess the impact of the management plan on the ecological sustainability by indicators related to the ecosystem itself, the economic sustainability of households and the VOI, and the social sustainability of households. The coherence of public policy also can be appreciated through controls made by forest, park or commune administrations or through taxes or nes perceived by these administrations. These constitute the answers to the development question raised in the introduction. We shall next tackle computer modeling issues.
The interactions between a household and a VOI are summarized in Fig. 12. The
UML swimlanes render explicit the sequence of interactions during a one-year cycle.
5 The formal model
In accordance with our methodology, we will formalize the conceptual model described above by a multi-agent system to obtain a formal model. Because the formal model is based on a multi-agent system, we rst will present the most up to date concepts in the eld of multi-agents system modeling to take into account the notions introduced earlier. We then will present our synthesis and explain how the model satises our modeling objectives.
5.1 The state of the art
Generally, a multi-agent system is made of agents, objects, possibly embedded into an environment, and possibly with an organization (Ferber 1999). An agent is an autonomous entity able to perceive its environment and to act on it, as well as to communicate with other agents (Wooldridge 2009).
Organization-centered multi-agents systems were introduced to contrast with agent-centered multi-agents system (Ferber and Gutknecht 1998). Extensive literature has since been produced, of which a small part is dedicated to social system modeling (Castelfranchi 2003), and a majority to MAS engineering. Hbner et al. (2002) fall into the latter category with Moise?, models focused on the distribution of tasks within an
123
Incorporating institutions, norms and territories in a generic model 65
Fig. 11 VOI behavior
organization with three specications: the structural specication that denes the structures roles, the functional specication that denes a hierarchy of goals partitioned into missions, and the deontic specication that links the missions to the roles. Esteva et al. (2004) with AMELI species electronic institutions imposing protocols on the agents interactions. A more sophisticated version is proposed in OPERA (Dignum 2004). To describe the semantics, various formalisms are proposed, using Object-Z (Lopez y Lopez et al. 2006), temporal logics (Jonker et al. 2007) or input/output logics (Boella and Van der Torre 2004). Generally, the specications are based on the notions of role and norm. As the literature mainly is focused on agent interactions, the roles often are limited to the agents roles.
123
66 S. Aubert, J.-P. Mller
Fig. 12 Interactions between the VOI and the household
Generally, norms are represented by deontic operators (permission, obligation, etc.) on a variety of contents. Hence, in Lpez y Lpez et al. (2002, ibidem 2006), the norms apply to goals, in Hbner et al. (2002) to missions (sets of coordinated goals), in Esteva et al. (2004) to actions (in particular, speech acts). Regarding the constitutive norms, an account of Searles count-as has been formalized by (Grossi 2007) and (Grossi et al. 2008) using epistemic logics and contextual description logics. For the latter, only the TBox has been used accounting for conceptual relativity. In particular, they cannot account for different points of view regarding a single object or geographic area.
In multi-agents system in general (Hbner et al. 2002), and in electronic institutions in particular (Esteva et al. 2004), the possibility of a violation (and therefore of a sanction) is not taken into account. Indeed, norms are used to specify a nominal behavior of the multi-agent system. Therefore, the resulting implementation must not violate the corresponding constraints. It is called norm compliance
123
Incorporating institutions, norms and territories in a generic model 67
by regimentation in Grossi et al. (2007). It is possible to violate norms at compile-time even if, as far as we know, these have never been described explicitly. At execution time, an agent developed separately from the institution can violate the norms (i.e. the imposed protocols), however the detection is systematic for integrity and security reasons. Only Lopez y Lopez et al. (2006) and Boella and Van der Torre (2004) describe how agents can violate the norms. Lopez y Lopez et al. (2006) introduce the reasons to comply with a norm, namely the contribution to an agent goal, the consequences of a possible sanction, and/or the associated reward. Psychological proles include opportunistic (when a norm does not directly contribute to a goal) or fearful (even if the sanction is not harmful) behavior. In any case, the norms are managed externally, not by the agents.
The space in which agents are situated is called the environment. Souli and Marcenac (2000) introduce for the rst time the idea of embedding agents simultaneously in different environments that are vary according to the different perceptions held by agents of the space. However, as in most classical multi-agents system, the environment is nite for practical reasons but the border has no meaning. Ferber et al. (2005) introduces AGRE (Agent-Group-Role-Environment), an extension of AGR in which an agent can have several so-called modes in different spaces. A mode corresponds to a role in a social space, or to a body in a physical space. This extension entails a multiplication of spaces similar to that which we proposed for territories (the physical spaces) and institutions (the social spaces).
From this quick overview, we retain the possibility to formalize our conceptual model using multi-agent systems with a notion of institution that corresponds to our conceptual denition. This notion will be complemented with contextual ontologies to account for the constitutive norms. We will show that this extension nicely ts our need to account for multiple points of view, roles and territories.
5.2 The multi-agent system
The general idea consists in using the contextual ontologies as introduced in Sect. 3 to account for various points of view:
The global point of view of the multi-agent system in which an environment supplies a multiplicity of interpretations from the various agents and institutions,
The agents points of view, which depend on the activities they are carrying out, The institutions points of view as described by the constitutive norms that provides the vocabulary in which the regulative norms are dened.
First, we introduce the notion of a multi-agent system that exists in only one exemplar to reify the global point of view.
Denition 1 A multi-agent system description is a tuple MASD OMAS; AMAS;
h
IMASi where:
OMAS is an ontology LMAS; TMAS; AssMAS
h i,
AMAS is a set of agents,
IMAS is a set of institutions,
123
68 S. Aubert, J.-P. Mller
Globally, OMAS is the description by an ontology of the multi-agent system from the point of view of the modeler. AMAS and IMAS are the concrete agents and
institutions the ontology will, among others, talk about. LMAS denes the vocabulary used by the modeler to describe the system. TMAS gives the denitions of this vocabulary and AssMAS is the concrete state of the multi-agent system at a given time. As an ontology is a static description, the dynamics will be dened later.
To account for the multi-agent systems specicities, we will decompose the ontology used by the multi-agent system, the agents and the institutions (or more precisely L for each ontology) in:
a set ARole of concepts for the agents, a set RRole of concepts for the objects, a set Act of activities which are a sequence available to operators, and a set Loc of locations that are spatial areas (as sets of places).
For example, LMAS = RRoleMAS[ORoleMAS[ActMAS[LocMAS.
From a multi-agent systems point of view, we recognize the environment described by the axioms of OMAS dened on LocMAS, the objects as passive entities that will be placed in the environment described by the axioms on RRoleMAS, the
agents as decisional entities that will also be placed in the environment and are described by AMAS and the axioms on ARoleMAS, and nally the institutions that structure the (both passive and decisional) entities interactions dened by IMAS. Our
aim is to use them to model respectively the land, the resources, the stakeholders and the institutions as sketched out in the Table 1.
Both the agents and the institutions dene a specic ontology. This ontology is considered as the vocabulary (i.e. a set of names) used for constructing the so-called social reality (Searle 1995) and, therefore, account for the constitutive norms (for example, that a particular tree species counts as fuel wood from a given point of view), and the regulative norms which dene the rules of behavior (for example, that someone counting as a license holder has the permission to cut what counts as fuel wood). Therefore, each agent or institution denes a point of view and its associated terminology that is what something counts as for him.
Additionally, an agent is endowed with a set of goals while an institution type denes a set of regulative norms. We come up with the following denitions.
Denition 2 The description of an agent a is a tuple ADa Oa; Ga
h i where: Oa is an ontology La; Ta; Aa
h i,
Ga is a set of goals stated as an assertion to make true.
Denition 3 The description of an institution i is a tuple IDi Oi; Ni
h i where:
Oi is an ontology Li; Ti; Ai
h i,
Ni is a list of regulative norms of the form ar; mod; act; or; l; q
h i where: ar is an agent role, mod is a deontic modality (obligation, permission, prohibition), act is an activity, or is an object role on which the activity is applied,
123
Incorporating institutions, norms and territories in a generic model 69
Table 1 correspondence between the higher-level concepts of the conceptual model and the MASD constructs
l [ Loc and q is a quantity.
A regulative norm states that a stakeholder counting as playing a given agent role (or) has the obligation, permission or prohibition to perform the activity act on the quantity (q) of an object counting as a given resource role (or) in a place counting as a particular location (l).The regulative norms capture exactly the notion of norm as described in Sect. 3.1. For example (see Fig. 3), we can introduce the roles of User (User [ Role) and Thing (Thing [ Act), as well as the activity ToUse (ToUse [ Act)
and introduce the norm User; permission; ToUse; Thing; Territory; 1
h i. The name
Territory is used instead of everywhere because an institution is supposed to have authority only on his territory. ? means that there is no restriction on quantity. We will see in the next section how to represent that a particular agent is a User, and a particular object is a Thing and therefore, that the norm applies. To simplify, we do not consider conditional norms or time restrictions although the latter is taken into account in the actual implementation.
The natural order on the quantities, a natural order on the deontic modalities (obligation [permission [ prohibition), as well as the containment relationship ( induce an order on the norms that is given by the following denition.
Denition 4 ar;mod;act;or;l;q
h i ar0;mod0;act0;or0;l0;q0
h iif andonlyif ar ar0;
mod\mod0; act act0; or or0; l l0 andq\q0:
Given that ( is a partial order, B is a partial order as well. This denition is very important to compute the rights to do something somewhere. Intuitively, if we take a set of norms, all of the minimal elements of the resulting partial order dene the actual norms to be applied on agents actions.
5.3 The dynamics
The structure of the multi-agent system, the agents and the institutions naturally induces the set of possible elementary state transitions. The elementary state transitions are called the operators. Therefore, we have to dene three sets of operators, respectively OMASS, OA and OI, for the multi-agent systems, the agents
and the institutions. They will be described using transitions.
However, these operators have to be applied and sequenced coherently to produce the global dynamics. We propose to formalize the global dynamics using DEVS (Zeigler et al. 2000) and more precisely the extension of DEVS proposed in
MIRANA concepts MASD concepts
Stakeholder Agent (A and ARole)
Institution Institution (I)
Resource Object (RRole)
Behavior Activity (Act)
Land Location (Loc)
123
70 S. Aubert, J.-P. Mller
Mller (2009) in order to deal more naturally with multi-agent systems. In the following section, we rst will describe the operators and then introduce DEVS and its use for specifying the global dynamics.
5.3.1 The operators
The MASD is a triple LMAS; TMAS; AssMASh i; AMAS; IMAS
h i of which everything can
change with the exception of LMAS and TMAS because they constitute the vocabulary of the modeler. Consequently, there are three categories of operators, OMASS:
the creation or deletion of the MASD itself, the addition of an element x into AssMAS, AMAS, or IMAS,
the removal of an element x from AssMAS, AMAS, or IMAS.
They can be formally described as follows (where ) denotes a transition rule): creation of a MASS: ? ) LMAS; TMAS; AssMAS
h i; AMAS; IMAS
h i
deletion of a MASS: LMAS; TMAS; AssMASh i; AMAS; IMAS
h i ) ?
addition of an element: assi x LMAS; TMAS; AssMASh i; AMAS; IMAS
h i ) LMAS;
h
h
TMAS; Ass0MASi; AMAS; IMASi such that Ass0MAS0 AssMAS [ fassigAss0MAS0
AssMAS [ fassig, and so on for AMAS, or IMAS deletion of an element: assi x LMAS; TMAS; AssMASh i; AMAS; IMAS
h i )
LMAS; TMAS; Ass0MAS
; AMAS; IMAS
such that Ass0MAS0 AssMAS fassig, and
so on for AMAS, or IMAS
These operators are used to realize the global dynamics of the system.
In the same way, the possible operators OA include changing the goals and beliefs encoded in the ontology. The formalization goes as follows:
creation of an agent: ? ) La; Ta; Aa
h i; Ga
h i
deletion of an agent: La; Ta; Aah i; Ga
h i ) ?
goal adoption: gi x La;Ta;Aah i;Ga
h i ) La;Ta;Aa
h i;G0a
such that G0a Ga [ fgig
etc.
Notice that the change in La and Ta corresponds to the acquisition or forgetting of vocabulary and denitions, leading towards conceptual knowledge revision.
Finally, the possible operators on the institutions OI concern institution creation or deletion as well as changes in the constitutive and regulative norms. The related changes are:
creation of an institution: ? ) Li; Ti; Ai
h i; Ni
h i
deletion of an institution: Li; Ti; Aih i; Ni
h i ) ?
norm adoption: ni x Li; Ti; Aih i; Ni
h i ) Li; Ti; Ai
h i; N0i
such that N0i Ni [ fnig
norm release: ni x Li; Ti; Aih i; Ni
h i ) Li; Ti; Ai
h i; Ni
h i such that N0i Ni fnig
etc.
goal release: gi x La;Ta;Aah i;Ga
h i ) La;Ta;Aa
h i;G0a
such that G0a Ga fgig
123
Incorporating institutions, norms and territories in a generic model 71
5.3.2 The dynamics
To specify the global dynamics, we use the DEVS (for Discrete EVent System) formalism, and more precisely the DEVS extension as described in Mller (2009). This extension has been shown to be particularly suited for describing multi-agent systems. We will introduce simplied denitions3 of an atomic DEVS model as elementary building block, and then the denition of a coupled DEVS model to recursively combine the building blocks.
An atomic DEVS model denes a generalized automaton in the following way:
Denition 8 An atomic DEVS model is a tuple X; Y; S; dext; dint; kext; kinth i where:
X is the set of incoming events, Y is the set of outgoing events, S is the set of possible states, dext is the response to the incoming events as a state transition, called an external transition: S 9 T 9 X ? S. The transition depends on the duration (T) since the last transition,
dint is a spontaneous state transition, called an internal transition: S ? S. kext is the output function denes the outgoing events to issue before an internal transition: S ? Y,
kint is the time advance function and denes the duration until the next internal transition: S ? T.
A DEVS model has been proved sufciently general to be able to describe any kind of behavior in terms of event exchanges. The behavior itself can be described directly using DEVS as an automaton specication or with any formalism ranging from differential equations up to BDI architectures.
Denition 9 A coupled DEVS model is a tuple N X; Y; D; fMdg; fIdg
h i where X is the set of incoming events, Y is the set of outgoing events, D is a set of component names, for each d [ D, Md is a DEVS model (atomic or coupled), for each d [ D [ {N}, Id is the inuencer set of d (i.e., the models connected
to d).
Therefore, a coupled DEVS model is a set of interconnected DEVS models. Further connections go in and out of the coupled DEVS model, making it appear as an atomic DEVS model from the outside. These ingoing connections are dened by making N [ Id, and the outgoing connections are dened by IN. It allows a recursive construction of DEVS models starting from a set of atomic DEVS models.
A coupled DEVS model can have a dynamic structure with the possibility to add or remove DEVS models, as well as to change the connections. The usual method (Zeigler et al 2000) is to add a particular model called an executive, which entails the topology of the coupled DEVS model, such that any change in the executive
3 Some details have been omitted because they are not necessary for this presentation.
123
72 S. Aubert, J.-P. Mller
state is mapped into a change of the coupled DEVS structure. In the following, we assume that this executive model always exists.
Having respectively dened the possible states of a multi-agent system, of an agent and of an institution, the embedding of these states within dynamics specications using DEVS is straightforward. The following denitions introduce the complete models for an agent and an institution.
Denition 10 An agent model (AM) is a tuple X; Y; AD; dext; dint; kext; kinth i.
Denition 11 An institution model (IM) is a tuple X; Y; ID; dext; dint; kext; kinth i.
In both cases, the possible states are nothing but the possible agent states and possible institution states, respectively. It is up to the transition and output functions to dene the actual behavior in terms of state changes. For example, a possible event ([ X) is the attribution of a contract to produce a given quantity q of timber for the
VOI. Accordingly, dext must be dened such that:
the goals to get the quantity q of timber is added to the list of goals C of the agent state;
the permission to cut the corresponding quantity of trees is granted (modication of R);
the obligation of delivering the quantity q of timber is given as well (modication of R again).
This example shows the possible treatment of an event, and its realization through a subset of the operators OA dened on the agent state. Additionally, the internal transition function dint is in charge of monitoring the goals of the agent, to modify the plan accordingly and, possibly, to execute actions. The actions are executed by issuing the corresponding events (kext). kint controls when the agent is going to plan and execute his activities.
Finally, the dynamics of the whole multi-agent system is formalized using the following denition.
Denition 12 A multi-agent system model is a tuple D; fMdg; fIdg
h i in which the
executive model Exe is of the form: X; Y; MASD; dext; dint; kext; kinth i.
Notice that a multi-agent system model is a coupled DEVS model without any incoming or outgoing event because the model is assumed to be closed at this level (unless we want to add interactions with the user). The executive model is dened for two purposes:
to maintain the structure of the multi-agent system: for each agent ai and institution ii, there is a corresponding model Mai or Mii. Accordingly, any change in the multi-agent system state is reected into the multi-agent system DEVS model.
to dene the dynamics of the environment. Moreover, all of the actions of the agents on the environment are converted to a corresponding event that is sent and dealt with by Exe. This distinction between an action and its actual execution by Exe reects the inuence/reaction paradigm described in Ferber and Mller (1996). The inuences are the events and Exe computes the reaction as a sequence of OMASS operators.
123
Incorporating institutions, norms and territories in a generic model 73
At this stage, only a part of the dynamics has been fully formalized, the other part having been directly implemented using the MIMOSA supporting language, i.e. Java. This implementation reects the behavior s described in the conceptual model and will not be elaborated further in this paper.
6 Implementation
The formal model described above, and consequently the MIRANA conceptual model, was implemented using the MIMOSA platform (Mller 2010). In MIMOSA, the modeler has to dene two kinds of models:
the conceptual model that corresponds to the T-Box of description logics, i.e. the concepts and their relationships. To each concept is attached a description of its dynamics using a set of available formalisms (differential equations, automata, scripting),
and the concrete model that corresponds to the A-Box of description logics, i.e. the instantiation of the conceptual model as a set of linked individuals. Additionally, for each instance, an initialization method as well as a set of observers can be dened.
These descriptions are used to generate a simulation model in the form of a DEVS (Discrete EVent System) model (Zeigler et al. 2000). The whole system is written in Java.
Consequently, the ontology described in the MIRANA conceptual model is directly entered using the graphical notation of MIMOSA (a kind of simplied UML class diagram). It then has been instantiated in the concrete model (using a kind of extended UML object diagram), and the simulation model generated and run. It is not the purpose of this paper to describe all of the details. We will just insist on the way the multi-agent system is initialized:
only the maps of the territories are dened. The space is generated from geographical information data by generating the minimal grid of cells of a given size (in our case 50 meters by 50 meters) covering the space in which we are interested (see Fig. 13). On the territories, we added the villages with their names and the map of the ecosystem habitats, considered as the territories of the species;
the institutions are described with the agent and resource roles, as well as their associated territory (and map). In particular, the intersection of each zone with each cell denes the pairs lri li; cih i (see denition 7) for each institution;
the household agents are generated from a description of their population given their annual needs and their initial memberships, and placed on their villages;
the species (considered as objects) also are generated from population descriptions and distributed on the ecosystems habitats as a function of the type of land cover.
Figure 14 illustrates some simulation results over 12 years (120 months) with non-legalist households and only a small fraction of the area with full conservation.
123
74 S. Aubert, J.-P. Mller
Fig. 13 The initialization of the MAS model space
Fig. 14 Some simulation results. a VOI nancial results, b habitat evolution
Figure 14a shows the VOI nancial results. The red line represents the tax incomes that are relatively constant over time, producing an increasing net income (green line). The initial negative result is due to the payment of the rst salaries. Figure 14b shows the evolution of the habitats as a percentage of the total surface. There is only a slow erosion of the primary forest. If the degraded land increases, there is similar growth of the secondary forest. The simulation over 60 years (tree growth cycle duration) shows some recovery of the primary forest.
7 Conclusion
We have presented the MIRANA conceptual model in which we have proposed a formalization, using ontologies, of the concepts of institution, stakeholder, resource and territory as dened by thematicians, more particularly in law anthropology. The aim is to understand the impact of norms in a context where human society structures a multiplicity of regulation mechanisms. At this level of description, we introduced the concept of institution as a general mechanism that shapes the spatial, biophysical and social reality and denes the regulative structures imposed on this reality using norms. The notion of role, usually only attributed to stakeholders, has
123
Incorporating institutions, norms and territories in a generic model 75
been extended to resources and portions of space (objects of normative orders), and taken into account using contextual ontologies. A consequence and contribution is the introduction of the notion of territory as an institution-based point of view of the space.
In the description of the system behavior, we introduced not only the possibility of norm violation, but also two complementary mechanisms for dealing with norms:
an external mechanism at the institution level using controls and sanctions, an internal mechanism within the stakeholders, using the norms as a resource for planning activities.
The latter mechanism is original because it is not only a lter of possibilities but also an enabler of further activities. Moreover, we made the distinction between the institution norms and their implementation. This implementation is realized by the moral person associated to the institution.
We also introduced a richer set of stakeholder behaviors. Indeed, behavior cannot be reduced only to means/ends planning or economic optimization, but a combination of these. In addition, the possibility to contract the stakeholders workforce is introduced.
We then introduced a new formal denition of an institutional multi-agent system using the concepts of agent, object, space and institution, and showed how to map the conceptual model description into this formal denition. Having encapsulated the institutional multi-agent system into DEVS, we were able to illustrate how to account for the dynamics described in the conceptual model. Conversely, we showed that the proposed institutional multi-agent system introduces a number of novelties, among which the natural treatment of a multiplicity of environments through the notion of territory seen as a way to account for a particular interpretation of a space considered as a resource. Finally, we partly described the implementation.
At the agent level, behavior has been abstracted within the DEVS framework. However, the specication of the transition functions themselves is still largely adhoc and should be made more generic than it currently is. For example, while the contracting dynamics actually are implemented, their representation as an institution has not yet been done. It would provide a more generic mechanism for contract negotiation and commitment. In particular, new contract types such as concession and incentive contracts would be easier to add to the system. Economic exchanges, as the market mechanisms, also could be implemented as institutions. Finally, households do not interact directly with exchange resources (including information). The addition of the concept of a social network, that is entirely missing now, is envisioned.
Regarding the application, i.e. the impact of the regulation (the VOI management plan) at the level of the local community on the ecological, economic and social sustainability of the SES can be analyzed and its vulnerability characterized. The satisfaction level of the households needs and evolutions in this level allows the impact on social sustainability to be assessed. Indeed, recurrent non-satisfaction would lead to increased illegal activities or the migration of the households towards other territories, either seasonally (town job), or denitively (emigration). This
123
76 S. Aubert, J.-P. Mller
dimension could be explored from a regional point of view, especially in the context of applying policies to reduced emissions from deforestation and degradation (REDD). The satisfaction level of the households and the VOIs nancial resources, its distribution and its evolution allows the impact on the economic sustainability of the new regulation to be appreciated. Of course, these two dimensions could be partly fungible because a better economic performance allows compensating, at least partly, the non-satised needs due to normative constraints and resource availability (but only if markets exist). Finally, the availability of the renewable resources, which is produced, in our case, by the ecosystem and its evolution, allows the ecological sustainability of the SES to be appreciated.
The VOI also regulates the system with legal tools that have an economic dimension (contracting, taxes, nes and subsidies). Hence, the collective economic sustainability also is measured by the capacity of the VOI to maintain itself nancially. It is possible to measure whether the ecological sustainability is due to the households regulated behaviors or to the possibility for the VOI to compensate losses. One can see that sustainable development is a complex interplay between individual behaviors, modes of collective organization and the dynamics of renewable resources (in this case, plants and animals because we did not consider water and biomass ows). Various sensibility analyses are ongoing to explore the multiple dimensions of this question.
References
Aubert S (2009) Lanthropisation: le rapport entre socit et nature au regard du Droit et Droits de proprit intellectuelle et biodiversit. In: Chretien-Vernicos G, Rude-Antoine E dir (eds) Anthropologies et droits: tat des savoirs et orientations contemporaines, AFAD, Dalloz, Paris, pp 342343, 353355
Aubert S, Muller JP, Ralihalizara J (2010) MIRANA: a socioecological model for assessing sustainability of community-based regulations. In: International congress on environmental modelling and software modelling for environments sake, Ottawa, Canada, July 2010, pp 18
Boaventura De Sousa S (1987) Law: a map of misreading; toward a postmodern conception of law. J LawSoc 14(3) (Autumn, 1987):279302
Boella G, Van der Torre L (2004) Regulative and constitutive norms in normative multiagent systems. In:
Procs. of KR04, pp 255265. AAAI Press, Menlo ParkBommel P, Mller JP (2007) An introduction to UML for modelling in the human and social sciences. In:
Phan D, Amblard F (eds) Agent-based modelling and simulation in the social and human sciences. Bardwell Press, Oxford, pp 273294Castelfranchi C (2003) Formalising the informal? Dynamic social order, bottom-up social control, and spontaneous normative relations. J Appl Log 1(12):4792Conte R, Castelfranchi C (1999) From conventions to prescriptions. Towards an integrated view of norms. Artif Intell Law 7(4):323340Crawford SES, Ostrom E (1995) A grammar of institutions. Am Polit Sci Rev 89(3):582600de Sadeleer N, Charles HB (2004) Droit international et communautaire de la biodiversit. Dalloz, Paris Debarbieux B (2003) Territoire. In: Lvy J, Lussault M (eds) Dictionnaire de la gographie et de lespace des socits. Belin, ParisDignum MV (2004) A model for organizational interaction: based on agents, founded in logic. PhD thesis, Utrecht UniversityEsteva M, Rosell B, Rodriguez-Aguilar JA, Arcos Josep L (2004) AMELI: an agent-based middleware for electronic institutions, vol 1. IEEE Computer Society
123
Incorporating institutions, norms and territories in a generic model 77
Farol S, Mller JP, Bont B (2010) An iterative construction of multi-agent models to represent water supply and demand dynamics at the catchment level. Environ Model Softw 25:11301148
Ferber J (1999) Multi-agent systems. An introduction to distributed articial intelligence. AddisonWesley, London
Ferber J, Gutknecht O (1998) A meta-model for the analysis and design of organizations in multi-agent systems. In: International conference on multi-agent systems, pp 128135, IEEEFerber J, Mller JP (1996) Inuences and reactions: a model of situated multiagent systems. In:
Proceedings of ICMAS96 (international conference on multi-agent systems), AAAI Press, Menlo ParkFerber J, Michel F, Baez J (2005) AGRE: integrating environments with organizations. In: Hutchison D,
Kanade J, Kittler J et al (eds) E4MAS 2004. Berlin, Heidelberg, pp 4856Grossi D (2007) Designing invisible hand-cuffs. PhD thesis, Dutch Research School of Information and
Knowledge Systems
Grossi D, Aldewereld H, Dignum F (2007) Ubi lex, ibi poena: designing norm enforcement in e-institutions. In: Coordination, organizations, institutions and norms in multi-agent systems II, pp 101114Grossi D, Meyer JJCh, Dignum F (2008) The many faces of counts-as: a formal analysis of constitutive rules. J Appl Log 6:192217Gruber TR (1993) Formal ontology in conceptual analysis and knowledge representation. Chapter: towards principles for the design of ontologies used for knowledge sharing. Kluwer Academic, DordrechtHbner JF, Sichman JS, Boissier O (2002) A model for the structural, functional, and deontic specication of organizations in multiagent systems. In: SBIAJonker CM, Sharpanskykh A, Treur J, Yolum P (2007) A framework for formal modeling and analysis of organizations. Appl Intell 27(1):4966Kirk M (1999) Land tenure, technological change and resource use: transformation process in African agrarian systems. Peter Lang, FrankfurtLe Roy E (1999) Le jeu des lois, une anthropologie dynamique du Droit, Paris, LGJD, col. Droit & socit, Srie anthropologiqueLe Roy E (2012) La terre de lautre, une anthropologie des rgimes dappropriation foncire. LGDJ, Paris
364 p
Lopez y Lopez F, Luck M, dInverno M (2006) A normative framework for agent-based systems. ComputMath Organ Theory 12(23):227250
Lpez y Lpez F, Luck M, dInverno M (2002) Constraining autonomy through norms. In: Proceedings of the international conference on multi-agent systemsMerry SE (1988) Legal pluralism, law & society review 22: 869896, Published by Blackwell Publishing on behalf of the Law and Society Association Stable. http://www.jstor.org/stable/3053638
Web End =http://www.jstor.org/stable/3053638 . Accessed20 Jan 2009, 05:36Montagne P, Razanamaharo Z, Cooke A (2007) Tanteza, le transfert de gestion Madagascar, 10 ans defforts, Antananarivo, Resolve/CIRADMller JP (2007) Mimosa: using ontologies for modeling and simulation. In: Proceedings of informatik
2007, 25 Sept 2007, Bremen, Germany. s.l.: s.n., [5] p
Mller JP (2009) Towards a formal semantics of event-based multi-agent simulations. In: Nuno D, Sichman JS (eds). Multi-agent-based simulation IX: ninth international workshop, MABS 2008, Estoril, Portugal, 1213 May 2008. Revised selected papers. Springer, Berlin [Allemagne], pp 110126. (Lecture Notes in Articial Intelligence, 5269)
Mller JP (2010) A framework for integrated modeling using a knowledge-driven approach. In: Swayne DA,
Yang W, Voinov AA, Rizzoli A, Filatova T (eds) Proceedings of the iEMSs fth biennial meeting : international congress on environmental modelling and software (iEMSs 2010), Ottawa, Canada Mller JP, Aubert S (2012) Une ontologie pour construire une reprsentation multi-niveau de et par les systmes sociaux. In: Rochebrune 2011, pp 111, Jan 2011 (Les Chemins de Traverse, to be published)
Mller JP, Rakotonirainy HL, Herv D (2011) Towards a description logics for scientic modeling. In:Proceedings of KEOD2011, Paris, France
Neale Walter C (1994) Institutions. In: Hodgson GM, Samuels WJ, Tool MR (eds) The elgar companion to institutional and evolutionary economics. Edward Elgar, Ann Arbor, pp 402406North Douglass C (1991) Institutions. J Econ Perspect 5(1):97112
123
78 S. Aubert, J.-P. Mller
Ostrom E (2005) Understanding institutional diversityGoogle books. Princeton University Press,Princeton
Ostrom E (2009) A general framework for analysing sustainability of socio-ecological systmes. Science325:419422
Parker DC, Filatova T (2008) A conceptual design for a bilateral agent-based land market with heterogeneous economic agents. Comput Environ Urban Syst 32:454463Porter-Bolland L, Eillis EA, Guariguata MR, Ruiz-Mallen I, Negrete-Yannkelevich S, Reyes-Garcia V
(2011) Community managed forests and forest protected areas: an assessment of their conservation effectiveness across the tropics. For Ecol Manag. doi:http://dx.doi.org/10.1016/J.foreco.2011.05.034
Web End =10.1016/J.foreco.2011.05.034 Rives F, Antona M, Aubert S (2012) Socio-ecological functions and vulnerability frameworks to analyse forest policy reform. Ecol Soc (no spcial public policies and management of rural forest: lasting alliance or fools dialogue ? submitted)
Sack R (1986) Human territoriality. Its theory and history. Cambridge University Press, Cambridge Santi R (2002) Lordre juridique. Dalloz, ParisSearle J (1995) The construction of social reality. The Free Press, New YorkSouli JC, Marcenac P (2000) A framework to model multiple environments in multiagent systems. In:
PRICAI 00
Wooldridge M (2009) An introduction to multiagent systems, 2nd edition. Wiley, New YorkZeigler B, Kim TG, Praehofer H (2000) Theory of modeling and simulation, 2nd edn. Academic Press,
New York
123
Springer Science+Business Media Dordrecht 2013