In this paper we propose a computational approach based on multi-agent systems to study the multiple forms of cooperative and competitive relationships within Industrial Districts (IDs). In particular, we develop a computational model by using AgentBuilder software, which is referred to a specific kind of cooperative relationships, namely that aimed at both balancing the utilization of supplier production capacity and minimizing the customer unsatisfied demand. Then we carry out a simulation analysis to prove the benefits of the selected kind of cooperation for the IDs and to evaluate the benefits of the cooperation in different competitive scenarios and diverse ID organizational structures.

Keywords:

Agent-based Computational Approach; Coordination Mechanisms.; Industrial Districts

Introduction

1.1

An Industrial district (ID) is a socio-territorial entity characterized by the active presence of both a community of people and a population of firms in a naturally and historically bounded area (Becattini, 1990). In other words, IDs are geographically defined productive systems characterized by a large number of small and medium sized firms that are involved at various phases, and in various ways, in the production of a homogeneous product. These firms are highly specialized in a few phases of the production process, and integrated through a complex network of inter-organizational relationships. A close relationship between the social, political, and economic spheres further characterizes the IDs (Pyke and Sengenberger, 1990)^[1].

1.2

The literature on IDs has widely stressed that the contemporary presence of competition and cooperation is one of the most important feature of the inter-organizational networks within IDs. This is also considered as a critical factor for the IDs' competitiveness (Becattini et al., 1990; Piore and Sabel, 1984; Porter 1998).

1.3

Due to the complex nature of IDs that integrate social, geographical, political and economical aspects, the literature on ID involves different streams of study, namely Social Sciences (Bagnasco, 1988; Granovatter, 1985), Industrial Economics (Becattini, 1990; Piore and Sabel, 1984; Storper and Harrison, 1991), Regional Economics (Digiovanna, 1996; Harrison, 1993), Economic Geography (Amin and Thrift, 1992; Markusen, 1996; Sforzi, 1990), Political Economy (Fuà, 1983), and Industrial Organization (Porter, 2001; Baptista and Swann, 1998). These studies have developed key notions and models to explain the reasons of the ID competitive gain, such as: the flexible specialization model conceptualized by Piore and Sabel (1984); the localized external economies concept anticipated by Marshall (1920) and further stressed by Becattini (1987) and Krugman (1991); the industrial atmosphere notion conceived by Marshall (1919); and the innovative milieux notion developed by the GREMI (Maillat et al., 1995). The resulting economic notions and models have been mainly based on empirical research (case study and survey), whereas computer simulation (i.e. computational) tools have been rarely adopted.

1.4

As a research tool computer simulation is attractive for different reasons. First, it allows the behavior of the 'real world' systems to be examined by developing simplified analogous models of the real systems. Then, it can be used to explore the behavior of 'artificial' systems in order to predict what might happen should such a system come into existence in the real world (Goldspink, 2002). Furthermore, simulation permits the researcher to study processes in ways nature prohibits, given that it can be run many times with the values of the model parameters modified in each run and changes observed in outputs (Carley and Prietula, 2000; Berends and Romme, 1999). According to Axelrod (1997), computer simulation can be applied for different purposes, namely prediction, performance, entertainment, training, education, confirmation and discovery. Of these, in this paper we use simulation mainly for confirmation and, to a lesser extent, prediction. In fact, we develop a computational model to study the inter-firm coordination within IDs, with the main aim at testing the existing theory on the benefits of the cooperation within ID but also making predictions about the possible outcomes of this theory under different scenarios.

1.5

In the paper we adopt a multi-agent systems (MAS) approach (Ferber, 1999; Weiss, 2000) to model the coordination mechanisms based on cooperation and competition within IDs. In particular, the MAS approach is used to model the coordination mechanisms observed in real IDs and it does not model how emergent behaviours arise from the agent interactions as global properties of the system. Such an approach is particularly relevant given that it allows both cooperation and competition behaviours to be contemporarily modeled (Durfee, 1988). In fact, MAS consist of a set of autonomous agents (single organizations), which share their information and cooperate each other to achieve a global goal while optimizing their individual objectives.

1.6

Recently, a small number of studies have adopted this approach to study IDs. For example, Boero and Squazzoni (2002) suggest an agent-based computational approach to describe the adaptation of IDs to the evolution of market and technology environment. Fioretti (2001) presents an agent-based model of the structure of the information flows within an ID.

1.7

In the paper, first, by analyzing the literature on IDs and empirical evidences, a conceptual framework of the ID coordination mechanisms based on MAS is developed. Such a conceptual framework is defined on the basis of the available MAS architectures and identifies three types of agents that interact within IDs, the interaction infrastructure, and the organizational structure. The three types of agents are: the firm agent, the stage super-agent and the ID super-agent. The latter represent specific coordination mechanisms that are spontaneously activated within IDs. These super-agents force the system to work in the way it behaves in reality. Therefore, the super-agents are basically fictitious agents, that make firms behave "as if" the spontaneous coordination mechanism would be there. However, they also might represent real agents of the ID (for example, consortium, trade associations, institutions, etc.). The conceptual framework can be used to analyze several forms of cooperation and competition within IDs. Then, on the basis of this framework, we develop a computational model by using a specific MAS software, i.e. AgentBuilder (Reticular System, 2000). In particular, the computational model is referred to a specific form of cooperation involving the organization of production, namely that aimed at both balancing the utilization of supplier production capacity and minimizing the customer unsatisfied demand. The computational model is used to evaluate the effects of the cooperation in the organization of production simulating the performances of both the single ID firms and the ID as a whole in different competitive scenarios.

1.8

The paper is organized as follows. In Section 2 we present the theoretical background. In particular, in Subsection 2.1 we briefly present describe the ID production model, emphasizing the aspects related to the contemporary presence of competition and cooperation. In Subsection 2.15 the different kinds of cooperation within IDs are presented and characterized by using three variables, namely the objective pursued, the actors involved, and the form of cooperation. In Section 3 we propose a conceptual framework for IDs based on the MAS. In Section 4, we introduce the main features of the MAS software, AgentBuilder, used to develop the computational model of the proposed MAS conceptual framework. In Section 5, the simulation analysis is carried out to study a specific kind of cooperation within IDs involving the organization of production and the attendant results are discussed.

Theoretical background

Industrial districts

2.1

IDs are specific production systems characterized by a high level of fragmentation of the production process in several production phases. Each production phases is performed by autonomous firms localized inside the ID, according to the flexible specialization model conceptualized by Piore and Sabel (1984). Thus, within IDs a spatial labor division is accomplished. In particular, two kinds of labor division characterize the ID production model: 1) vertical labor division and 2) horizontal labor division.

2.2

The vertical labor division among the ID firms results from the specialization of the single firms on one or few phases of the production process. Each phase represents a stage of the supply-chain. Firms operating in one stage of the supply chain (suppliers) provide multiple buyers with components and sub-systems of the final product, which require specific competencies and technologies not detained by the buyers (Albino et al, 2000).

2.3

The horizontal labor division is the consequence of the great number of small and medium sized firms operating in each production phase. These firms may act with different degree of specialization in doing the same production phase, for example being specialized on singular products, models, or working. In this case, a firm establishes inter-firm relationships with other firms operating in the same production phase in order to acquire an additional labor capacity (capacity subcontracting). In this way the firm achieves a desired production level or reduces production costs and/or increases its manufacturing flexibility (Albino et al., 2000).

2.4

The specific production model characterizing IDs enables ID firms to capitalize on the benefits of the economies of scale and production efficiency, characteristic of the large firm, and to gain the typical advantages of small sized firms, such as the adaptability and flexibility (Garofoli, 1981; Rabelotti, 1995). The economies of scale and the production efficiency are due to: (i) the high specialization of firms, which allows them to optimize the use of boththe equipment and workforce so thatlabor; and (ii) the great dimension of the buyer-supplier exchanges, which allow the suppliers to increase their production volume maximizing the use of their equipment (Silvestrelli, 1984). The ID adaptability and flexibility are due to the vertical division of labor, the specialization of firms, and the dense network of relationships, because each task can be re-organized with a different mix of specialized producers (Gandolfi, 1988; Paniccia, 1999).

2.5

Both types of labor division favor the formation of a dense network of inter-firm relationships in which cooperation and competition coexist. In particular, inter-firm cooperation within IDs has been traditionally developed to a greater extent vertically rather than horizontally. With this regard, Brusco (1990) argues that cooperation and collaboration, especially over technical innovation and design, take place between firms doing different things, i.e. at different stages of the supply chain. A similar concept is stressed by Sabel (1982). He points out that between final firms (i.e. those firms which have access to the final market) and stage firms (i.e. those firms involved in a phase of the production process), there are mainly cooperative interactions.

2.6

Competition characterizes the relationships among similar firms (Brusco, 1990). These compete each other in order to gain a demand volume that fills up their manufacturing capacity. To a greater extent, competition is not centered on cost, but on the product quality and on the additional services provided to buyers (Digiovanna, 1996). Many authors argue that this form of competition has a critical role for the ID competitiveness. In fact, competition stimulates the firms to continuously improve their production processes, increase their standard quality, and innovate their products and services (Becattini, 1997; Piore and Sabel, 1984; Porter, 1998).

2.7

Cooperation and competition take many forms within IDs. Regarding to the relations between different stages of the supply chain, it is possible to recognize within IDs different degrees of vertical cooperation. For example, vertical cooperation is weaker if the bargaining power in the subcontracting relations is in the hands of buyer firms. In fact, in this case the buyers often take advantage of the high level of horizontal competition that exists in the various stages of the value chain by threatening to substitute their subcontractors (Corò and Grandinetti, 1999). In particular, buyers can accomplish the following strategies:

• multiple sourcing policies, which allow the buyer to compare the performances of the suppliers and have a bidding competition between them;
• keeping part of the product in house, which restores credibility to the possibility of completing backward integration on the part of the buyer (De Toni and Nassimbeni, 1995).

2.8

Regarding to the relations between similar firms, in some IDs it is possible to find a large degree of horizontal cooperation. This form of cooperation is generally addressed to market research programs, technology acquisition, training or technology transfer, permanent exhibition, etc. (Brusco, 1990; Corò and Grandinetti, 1999; Digiovanna, 1996; Saxenian, 1996). The equilibrium between cooperation and competition represents one of the most important characteristics of IDs; in fact, as stressed by several authors (Becattini, 1997; Porter, 1998), it constitutes one of the preliminary requirements for the development of an ID, and one of the essential conditions of its reproduction. Such equilibrium can be spontaneous or it can be guided by some ID actors. The spontaneous balancing of cooperation and competition results from the special way in which the common system of ethical values and beliefs shared among the ID firms permeates and structures the market in the district (Becattini, 1990). In fact, this intangible (socio-cultural) infrastructure defines a set of unwritten behavioral rules that everyone knows and almost everyone respects, which reduce opportunistic behaviors of self-interested subjects in favor of cooperative behaviors (Paniccia, 1999). In some cases, the equilibrium between cooperation and competition does not occur spontaneously but requires the positive action of some institutional actors, from local government to trade associations, unions and local banks (Dei Ottati, 1995; Pilotti, 1999; Piore and Sabel, 1984).

The different kinds of cooperation within industrial districts

2.9

By analyzing the literature on IDs and empirical evidences (Carbonara et al., 2002; De Toni and Nassimbeni, 1995; Dei Ottati, 1994; Markusen, 1996; Piore and Sabel, 1984; Saxenian, 1996), different kinds of cooperation can be identified.

2.10

With regard to horizontal cooperation, we can refer to:

• formal or informal agreements set up among competitor firms to share risk connected to R&D activities or to gain access to specific market segments;
• informal agreements set up among competitors to keep prices fixed;
• cross-licensing patents to competitors;
• permanent exhibition areas shared by final firms to reduce trade costs;
• purchasing and offering groups, which organize supply or demand together in order to achieve economy of scale (by centralizing the activities) and to increase the contractual power.

2.11

With regard to vertical cooperation, we can refer to:

• long term contracts and commitments between buyer and supplier;
• adoption of interacting CAD-CAM tools enabling buyer and suppliers to co-develop new products;
• adoption of JIT methodologies enabling buyer and suppliers to minimize the inventory level;
• technical collaboration arrangements between manufacturing firms and infrastructure suppliers, aimed to develop new manufacturing infrastructures.

2.12

Finally, there are forms of hybrid cooperation that have received institutional planning. These are generally organized as consortiums, set up by local public institutions or trade associations. Different types of activities are carried out by these consortiums and associations in order to improve the ID performances, namely sponsoring seminars and educational activities, organizing trade shows, providing market forecasts for various segments of the industry, offering a range of financial services.

2.13

The above kinds of cooperation involve different actors operating inside the ID (namely institutions, single firm, group of firms, trade associations, etc), which pursuing a common goal. Regarding to this, two main types of cooperation can be identified:

• cooperation aimed at solving potential or actual conflicts;
• cooperation aimed at increasing the performance levels of groups of firms (such an object has the underlying primordial object of survival).

The two types of cooperation require the adoption of different coordination mechanisms. For example, resolving a conflict may require the establishment of dominance or the definition of rules of priority, whereas improving the performance level may require different forms of agreements (Ferber, 1999). Furthermore, specific communicative interactions among the actors are activated in order to guarantee effective and efficient coordination.

2.14

Thus, the different kinds of cooperation within IDs can be described by using three main dimensions: 1) the objective pursued, 2) the actors involved, and 3) the form of the cooperation.

2.15

In Table 1 some examples of forms of cooperation within IDs are reported, defining the three main dimensions.

Table 1: Examples of cooperation within IDs
Objective	Actors	Form of cooperation	Example
Inventory reduction	Buyer and her suppliers	Vertical	Just-in-Time practices adopted by leader final firms with their preferential suppliers in the glasses district of Belluno (Nassimbeni and De Toni, 1995)
Balancing the utilization of production capacity	Similar firms operating in the same production phase (e.g. buyers or suppliers)	Horizontal	Leader firms in the leather sofa district of Bari-Matera distribute the order among subcontracting firms balancing the utilization of their production capacity (Carbonara et al., 2002)
Minimizing the ID unsatisfied demand	Final firms	Horizontal	In Prato textile district, production is organized by a special class of agents, called middlemen, which distribute the customer order among final firms (Gandolfi, 1990). Final firms collaborate to satisfy the maximum value of the market demand within the district of Lumezzane (Corò and Grandinetti, 1999)
Reduction of flow traffic overload	Firms within the whole ID	Hybrid	Transportation regional plan of Emilia Romagna IDs
Co-design in new product development process	Suppliers and buyer	Vertical	Collaboration between local infrastructure suppliers and manufacturing firms in the textile districts of Como and Prato and in the ceramic district of Sassuolo (Visconti, 1996).
Minimization of purchasing, sales and distribution costs	Final firms or Buyers	Horizontal	Permanent exhibition area shared by small final firms used as trade channel in the furniture district of Brianza (Corò e Grandinetti, 1999). Joint policies coordinated by consortium or trade associations to achieve economy of scale in the purchasing and selling in the wooden chairs district of Manzano (Carbonara, 2002), jewellery district of Valenza Po (Corò and Grandinetti, 1999).
Minimization of sales and distribution costs	Firms within the whole ID	Hybrid	Joint policies coordinated by public institutions, consortium, or trade associations to promote, sale and distribute the ID products in textile district of Carpi (Gandolfi, 1990), wooden chairs district of Manzano (Carbonara, 2002)
Human resource training and qualification	Firms within the whole ID	Hybrid	Policies coordinated by public institutions, trade associations to promote and sustain the development of managerial and technical competencies in the furniture district of Brianza (CeRTET, 1995).
Development of an innovative environment sustaining the adoption of the latest technology	Similar firms operating in the same production phase	Horizontal	Semiconductor firms within Silicon Valley liberally cross-licensed their patents to competitors to ensure the quickly diffusion of technical advances (Saxenian, 1996).
Development of a competitive environment sustaining the growth	Firms within the whole ID	Hybrid	MassMEDIC within the Biomedical district in Massachusetts is a trade association consisting of the ID firms and government organizations (Porter, 1998)

A conceptual framework for industrial districts based on multi-agent systems

3.1

From the above discussion about the cooperation within IDs, some interesting features can be pointed out. Cooperation involves a group of independent firms that, while pursuing their own autonomous objectives, decide to collaborate to achieve a common goal, so as to improve their performance or to resolve a conflict. As a consequence, these actors adopt different types of coordination forms, which depend on the organizational structure in which the actors are arranged, and which require a communication infrastructure among the actors to be implemented. These features well fit the characteristics of MAS, which involves multiple autonomous optimizing agents able to learn and to communicate each other through messages. Therefore a multi-agent based approach is used to develop the following conceptual framework.

3.2

Based on various designs for MAS in the literature (Bond and Gasser, 1988; Brazier et al., 1995; Carley and Prietula, 1998; Ferber, 1999; Fisher, 1994; Lin and Shaw, 1998; Kang et al., 1998), we propose a conceptual framework made up of the following parts: the agents, the interaction infrastructure, the organizational structure.

3.3

Agents are active goal-oriented objects that possess certain capabilities to perform tasks and communicate with other agents. Each agent is provided with the following characteristics:

• a set of goals that have to be accomplished;
• a set of tasks that have to be performed in order to accomplish the goals;
• a social memory that keeps the knowledge about the other agents social behavior;
• a working memory that keeps the knowledge about the agent state. The content of the working memory is about the agent capabilities.
• a set of rules of social engagement, which define the agent social behavior. The rules concern with the honesty of the agent in working with other agents.

3.4

The interaction infrastructure defines how agents interact each other. The interaction infrastructure includes the communication protocols, which describe mechanisms for agents to communicate single messages (Huhns and Stephens, 2000). These specify the message types that can be exchanged among agents (query or assertion) and the communication methods (one-to-one, broadcasting).

3.5

The organizational structure represents the system of relationships among agents and the hierarchical structure in which agents are arranged. According to the distribution of decision-making control, it is possible to define classes of agents at different hierarchical levels. The organizational structure defines the architecture of the coordination mechanisms within the MAS.

3.6

The proposed conceptual framework consists in three typologies of agents, which communicate each other and are organized within a three-level organizational hierarchy.

Figure 1. The multiagent-based conceptual framework of the ID.

3.7

The three typologies of agents are: (i) the firm-agent, (ii) the stage-super-agent, and (iii) the IDsuper-agent. A firm-agent corresponds to a single organization that is involved in a phase of the production process. This phase identifies the stage of the supply chain within the ID, which the firm belongs to. A stage can include different firms, which carry out the same production phase. They can compete for the same customer, but also cooperate for common objectives (horizontal cooperation). The cooperation and competition within the stage is managed by the stage-super-agent. When the cooperation involves firms belonging to subsequent stages along the SC (e.g. buyer-supplier), the respective stage-super-agents negotiate adopting a given negotiation mechanism (vertical cooperation). Finally, the IDsuper-agent is responsible for the cooperation among the ID firms, aimed at both increasing the performance levels of all the firms and guaranteeing their survival (hybrid cooperation) and for resolving conflicts between the stage-super-agents (vertical cooperation).

3.8

The described agents negotiate at three hierarchical levels, namely the firm level, the stage level, and the ID level, pursuing different goals.

The agents

3.9

As already mentioned the proposed conceptual framework consists in three typologies of agents. The firm-agent can assume one of the two roles, buyer agent or supplier agent, when interacts with a firm agent of an adjacent stage. They correspond to real firms localized in the ID.

3.10

The buyer agent is characterized by:

• a set of goals, e.g. minimizing the unsatisfied demand, minimizing the distribution and sales costs;
• a set of tasks, e.g. purchasing from suppliers, sales to the customer;
• a social memory, knowledge on the trustworthy behavior of the suppliers, which is updated by using a scoring scheme;
• a working memory, including the agent capabilities to perform its tasks, e.g. maximum production capacity;
• a set of rules of social engagement, e.g. long-term relationship with a preferential supplier.

3.11

The supplier agent is characterized by:

• a set of goals, e.g. maximizing the utilization of production capacity, keeping multiple buyers;
• a set of tasks, e.g. providing the buyers with components ;
• a social memory, knowledge on the trustworthy behavior of the buyers, which is updated by using a scoring scheme;
• a working memory, including the agent capabilities to perform its tasks, e.g. maximum supply capacity;
• a set of rules of social engagement, e.g. satisfying the demand of a preferential buyer.

3.12

The stage-super-agent and ID-super-agent are basically fictitious agents. They play roles that correspond to coordination mechanisms that are spontaneously activated within IDs (for example, local and social mechanisms of sanctions and rewards). Their characteristics such as goals, tasks, and rules, are based on the ID literature and the empirical cases. In particular, the stage-super-agents aim at optimizing the stage performance. With this aim, they manage the cooperation and resolve the conflicts within the stage. The stage super-agent is also distinguished into: the buyer super-agent and the supplier super-agent.

3.13

The buyer super-agent is characterized by:

• a set of goals, e.g. balancing the unsaturated production capacity within the stage buyer, minimizing the total unsatisfied demand, minimizing the distribution and sales total costs; optimizing the customer satisfaction;
• a set of tasks, e.g. allocation of the unsatisfied demand within the stage, coordination of the cooperation among buyer agents;
• a social memory, including the meta-knowledge about the buyer agents (their performances, states, and social behavior);
• a working memory, including the super-agent capabilities to perform its tasks, e.g. mechanisms of allocation, decision scheme;
• a set of rules of social engagement, e.g. guaranteeing the survival of the buyer agents.

3.14

The supplier super-agent is characterized by:

• a set of goals, e.g. balancing the unsaturated production capacity within the stage supplier, minimizing the total unsatisfied buyer demand, reinforcing the contractual power of the stage supplier;
• a set of tasks, e.g. allocation of the unsatisfied demand within the stage, coordination of the cooperation among supplier agents;
• a social memory, including the meta-knowledge about the supplier agents (their performances, states, and social behavior);
• a working memory, including the super-agent capabilities to perform its tasks, e.g. mechanisms of allocation, decision scheme;
• a set of rules of social engagement, e.g. guaranteeing the survival of the supplier agents.

3.15

The ID super-agent manages the cooperation among the ID firms and resolve conflicts between the stage-super-agents. The ID super-agent is characterized by:

• a set of goals, e.g. knowledge transfer, firm knowledge improvement, assuring the persistence of inter-firm relations within the ID, reducing the exit of the firms from the ID, minimizing the death rate of the firms within the ID;
• a set of tasks, e.g. allocation of the customer demand among the supply chains within the ID, supporting increase agent performance due to the knowledge transfer process within the ID;constraint on the number of average connections for every agent
• a social memory, including the meta-knowledge about all the agents and stage super-agents (their performances, states, and social behavior);
• a working memory, including the super-agent capabilities to perform its tasks, e.g. mechanisms of allocation, decision scheme;
• a set of rules of social engagement, e.g. guaranteeing the reproduction of the ID, balancing cooperation and competition within the ID.

The interaction infrastructure

3.16

The interaction infrastructure characterizing the conceptual framework includes:

1. Interactions among firm agents (buyers and suppliers). Each buyer agent is able to communicate with the supplier agent in order to achieve its goals. Communication protocols characterizing these interactions consist of two communication methods, namely one-to-one and broadcasting. One-to-one communication occurs when one buyer (supplier) sends a message to a supplier (buyer). Broadcasting occurs when one buyer (supplier) sends a message that goes simultaneously to all suppliers (buyers). Two different message types are exchanged among buyer and supplier agents, namely query and assertion. The agent social memory and the rules of social engagement affect the interaction between buyer and supplier agents. In fact, preferential relationships can exist among them (vertical cooperation).
2. Interaction between firm agent and stage super-agent, which is devoted (i) to resolve conflicts among the firm agent that belong to a given stage and (ii) to assure horizontal cooperation. Firm agent communicates with the stage super-agent sending a query message, whereas stage super-agent communicates through assertion messages.
3. Interaction between stage super-agents, which occurs when the coordination of different cross-stage activities is required (vertical cooperation). Both query and assertion messages are possible.
4. Interaction between ID super-agent and firm agents, which occurs when the cooperation within the ID is aimed at increasing the performance levels of all ID firms or at guaranteeing their survival. Assertion messages come from ID super-agent to firm agents.
5. Interaction between ID super-agent and stage super-agents, which is devoted (i) to solve conflicts that occur when the stage pursuing conflicting goals, and (ii) to increase the performance of the supply chains within the ID. Both query and assertion messages are possible.

The organizational structure

3.17

The three typologies of agents of the framework (firm-agent, stage-super-agent, IDsuper-agent) are organized in three hierarchical levels, namely the firm level, the stage level, and the ID level (Figure 1). In particular, at the first hierarchical level (firm level) there are the firm-agents; at the second hierarchical level (stage level) there are the stage-super-agents; at the third hierarchical level (ID level) there is the ID-super-agent. Each agent at a certain hierarchical level owns the knowledge about the agents of lower hierarchical levels and can take decisions that affect the lower hierarchical levels.

3.18

Vertical and horizontal relationships among agents define the organizational structure of our model. Vertical relationships connect agents at different hierarchical level. These relationships are coordinated by hierarchy-like mechanisms. Horizontal relationships exist between agents at the same hierarchical level. These relationships are coordinated by different negotiation mechanisms.

3.19

The horizontal relationships, which involve agents at the same hierarchical level and pursuing different goals, could be stalled due to a conflict between agents. The decision to be taken to solve the conflict is reached by the agent at the higher hierarchical level. To this end, vertical relationships are activated.

The computational model

4.1

The proposed MAS conceptual framework can be easily made computational by adopting a specific MAS software. In this way, it is possible to carry out a simulation analysis to test general hypotheses and to show how the system behaves in different contexts.

4.2

We adopt the AgentBuilder software (Reticular Systems, 2000), which is characterized by:

1. Belief-Desire-Intention Agents;
2. Reticular Agent Definition Language (RADL) programming language for the agents;
3. Knowledge query and manipulation language (KQML) communication language for the interaction among agents.

4.3

A BDI agent is characterized by seven main components (Wooldridge, 2000): < type=>

A set of current beliefs, representing information the agent has about its current environment;

A belief revision function, which takes a perceptual input and the agent's current beliefs, and on the basis of these, determines a new set of beliefs;

An option generation function, which determines the options available to the agent (its desires), on the basis of its current beliefs about its environment and its current intentions;

A set of current options, representing possible courses of actions available to the agent;

A filter function, which represents the agent's deliberation process, and which determines the agent's intentions on the basis of its current beliefs, desires, and intentions;

A set of current intentions, representing the agent's current focus - those states of affairs that it has committed to trying to bring about;

And action selection function (execute), which determines an action to perform on the basis of current intentions.

4.4

RADL is an object-oriented programming language derived by the AGENT0 language. In the RADL language, an agent is specified in terms of a set of capabilities, a set of initial intentions, beliefs, and commitments, and a set of behavioral rules. Beliefs, capabilities, commitments, and intentions consist in the agent mental model. In particular, beliefs represent the current state of the agent's internal and external world and are updated as new information about the world is received. An agent can have beliefs about the world, about another agent's beliefs, about interactions with other agents, and about its own beliefs. The agent's list of capabilities defines the actions that the agent can perform. To each action a list of preconditions are associated, which must be satisfied before execution of the action (Thomas, 1993). A capability is static and holds for the life-time of an agent. However, the actions an agent can perform may change over time because changes in the agent's beliefs may alter the true value of precondition patterns in the capability. The AgentBuilder agent architecture classifies actions in two main categories, private actions and communicative actions. Private actions are actions that affect or interact with the environment of the agent. Communicative actions are defined as actions that send messages to other agents. A commitment is an agreement, usually communicated to another agent, to perform a particular action at a particular time. An intention is an agreement, usually communicated to another agent, to achieve a particular state of the world at a particular time. Intentions are similar to commitments in that one agent performs action(s) on behalf of another. However, a commitment is an agreement to perform a single action whereas an intention is an agreement to perform whatever actions are necessary to achieve a desired state of the world.

4.5

Behavioral rules determine how the agent acts. Each behavioral rule contains a message condition, a mental condition, and an action. In order to determine whether such a rule fires, the message condition is matched against the messages the agent has received; the mental condition is matched against the beliefs of the agent. If the rule fires then the agent performs the action (Wooldridge, 2000). In the RADL programming language, behavioral rules can be viewed as WHEN-IF-THEN statements. The WHEN portion of the rule addresses new events occurring in the agent's environment and includes new messages received from other agents. The IF portion compares the current mental model with the conditions that are required for the rule to be applicable. Patterns in the IF portion match against beliefs, commitments, capabilities, and intentions. The THEN portion defines the agent's actions and mental changes performed in response to the current event, mental model, and external environment. These may include: mental model update, communicative actions, and private actions.

4.6

The KQML is a protocol for exchanging information and knowledge among the agents (Finin et al., 1994). A KQML message consists of a performative, the content of the message, and a set of optional arguments. The performative specifies an assertion or a query used to examining or changing the mental model of a remote agent. In particular, KQML provides the following types of performatives:

• performatives enabling an agent to send requests to another agent (ask-if, ask-one, achieve, etc.);
• performatives that enabling an agent to reply to another agent (tell, eos, sorry, etc.);
• performatives supporting the information exchange (tell, untell, deny, etc.), the transferring of functionalities (insert, tell, etc.), and the definition of capabilities (advertise, subscribe, etc.).

4.7

The agent execution cycle consists of the following steps:

• processing new messages;
• determining which rules are applicable to the current situation;
• executing the actions specified by these rules;
• updating the mental model in accordance with these rules;
• planning.

4.8

Processing a new message requires identifying the sender, then the message is analyzed and made part of the mental model. The next step is determining which rules match the current situation. That means comparing the message and the current mental state with the message and mental conditions of each rule. A rule is marked for execution when all of its conditions are satisfied. Rule execution consists of performing private and communicative actions and making mental changes. Next, the agent's mental model is updated by changing, adding, and removing mental elements as specified by the execution behavioral rules. The final step in the cycle requires performing a plan of tasks in order to satisfy goals specified by the agent's intentions. This step is activated when the message contains a request to perform an intention.

The study of cooperation in the organization of production

5.1

In the following we are interested to study a specific kind of cooperation within IDs involving the organization of production. This can be described by adopting the three dimensions presented in Section 3 (Table 2).

Table 2: The cooperation in the organization of production

Objectives

Balancing the utilisation of production capacity (supplier stage)

Minimising the ID unsatisfied demand (buyer stage)

Actors

firms belonging to the final assembly stage (buyer)

firms that supply parts (supplier)

Form of cooperation

Horizontal and vertical

5.2

In the following, we first characterize the MAS conceptual framework for the specific kind of cooperation (See 5.4) and then we carry out the simulation analysis by using the associated computational model developed by adopting AgentBuilder (See 5.28).

5.3

Simulation analysis is carried out for two main purposes, namely to prove the benefits of the selected kind of cooperation for the ID and to evaluate the benefits of the cooperation in different competitive scenarios and diverse ID organizational structures.

The instantiated conceptual framework

5.4

In the following we specify the characteristics of the agents, the interaction infrastructure and the organizational structure for the selected kind of cooperation. Such characteristics are expressed in terms of beliefs, capabilities, and behavioral rules in the computational model.

The agents

5.5

The agents that interact within the ID are the following five:

• two types of firm agents, i.e. the supplier agent and the buyer agent,
• two types of stage super-agents, the supplier super-agent and the buyer super-agent,
• one ID super-agent.

5.6

The buyer agent represents the ID final firm, that is the firm assembling the final products and receiving the final customer demand. The supplier agent represents the ID subcontracting firm, that is the firm receiving purchasing orders from the final firms and supplying the required components to them. The buyer super-agent is engaged in minimizing the customer unsatisfied demand, giving the demand unsatisfied from a firm to the others. The supplier super-agent is engaged in balancing the utilization of supplier production capacity. The ID super-agent is engaged in guaranteeing both the critical success factors characterizing the IDs, namely the economy of scale for each ID firm and the flexibility of the whole system.

The interaction infrastructure

5.7

Customer demand arrives at each buyer agent, who has to coordinate his production activities for satisfying it, by communicating with the other agents.

5.8

Interaction between buyer agent and supplier agent The buyer agent asks to the supplier agents the availability to satisfy an order, and in turn suppliers answer to this message on the basis of their supply capacity. The agent social memory and the rules of social engagement affect the interaction among buyer and supplier agents. In fact, given that preferential relationships exist between buyers and suppliers (vertical cooperation), each buyer will prefer to place an order to her preferential supplier (e.g. due to the fact that the latter assure the higher performance). Moreover, in the case the preferential supplier is not available, the buyer agent could select a supplier who could be untrustworthy (e.g. he could promise a supply quantity that is not able to produce).

5.9

Interaction between buyer agent and buyer super-agent The interaction occurs when the buyer agent cannot satisfy the customer demand. In this case he sends a message to the buyer super-agent, communicating the quantity of the unsatisfied customer demand . The latter will make decision on the allocation of the unsatisfied demand to the available buyer/s, communicating the assigned demand to the selected buyer/s. To select the buyer, the super-agent takes into account two criteria, namely the production capacity of the buyer and the availability of the attendant preferential supplier. Both these criteria are coherent with the goal of minimizing the unsatisfied customer demand. In fact, the first criterion is related to the assumption that concentrating the unsatisfied demand to a unique buyer assures the higher stage performance. The second criterion is related to the assumption that preferential suppliers assure the higher stage performance.

5.10

Interaction between supplier agent and supplier super-agent The interaction occurs when the supplier agent receives a message of bid from a not preferential buyer. In this case she communicates is availability to satisfy the buyer demand to the supplier super-agent. The latter is in fact responsible for the decision pertaining the allocation of the buyer demand among the available not preferential suppliers, which will be communicated to the supplier agent. The supplier super-agent will make decision coherently with the goal of balancing the utilization of the production capacity of the suppliers available to satisfy an order (fragmentation policy).

5.11

Interaction between buyer and supplier super-agents In the case of unsatisfied customer demand, the buyer super-agent asks to the supplier super-agent the availability of the preferential suppliers, so as to make decision on the allocation of the unsatisfied demand to the attendant buyer. The supplier super-agent communicates this information. Two different situations might originate a conflict between buyer and supplier super-agent . First, alternative supply chains made up of different couples of preferential buyer-supplier are available. Second, no alternative couples of preferential buyer-supplier are available. In both cases a conflict between the buyer super-agent goal (i.e. concentrating the demand) and the supplier super-agent one (i.e. distributing the buyer demand to all available suppliers) is determined. The resolution of the conflict pertains to the ID super-agent.

5.12

Interactions between supplier and ID super-agents When multiple not preferential suppliers are available to cover the demand expressed by the same buyer, originating a conflict between buyer and supplier super-agents, the latter sends a message to the ID super-agent. This will make decision, concerning the allocation of the unsatisfied demand among the suppliers, coherently with the goals of assuring the economy of scale for each ID firm and the flexibility of the whole system. Then, the ID super-agent communicates her decision to the supplier super-agent. This can be coherent with the goal of the supplier or the buyer super-agent, namely ID super-agent can decide to distribute the order among all the trustworthy suppliers (fragmentation policy) or to assign the order to a unique supplier (concentrating policy). The criterion that the ID super-agent adopts to choose between the two policies is based on the reliability of the suppliers. If there are trustworthy suppliers, these will receive a quota of the unsatisfied demand, namely the ID super-agent will communicate to the supplier super-agent to adopt the fragmentation policy. Otherwise, ID super-agent will communicate to the supplier super-agent to adopt the concentration policy, selecting the supplier with the maximum supply capacity.

5.13

Interactions between buyer and ID super-agents When multiple couples of preferential buyer-supplier to which assign the unsatisfied customer demand are available, originating a conflict between supplier and buyer super-agents, the latter sends a message to the ID super-agent. She will decide the allocation of the unsatisfied demand among the alternative supply-chains (couples of preferential buyer-supplier). The ID super-agent makes decision coherently with the goal of maximizing the ID customer satisfaction, that in turn means minimizing the unsatisfied customer demand. Therefore, if it is possible to allocate all the unsatisfied demand to a unique buyer, she will choose the policy to concentrate the unsatisfied demand, selecting the buyer characterized by the maximum supply chain capacity^[2] (concentrating policy). Otherwise, he will choose to distribute the unsatisfied demand among all the available supply chains (fragmentation policy). Finally, the decision is communicated to the buyer super-agent.

The organizational structure

5.14

The five agents of the model are organized in three hierarchical levels. In particular, at the firm level there are the buyer and supplier agents; at the stage level there are the buyer and supplier super-agents; at the ID level there is the ID-super-agent. According to what described in the previous Section, the decision-making control is distributed among the three hierarchical levels. From the bottom hierarchical level (firm level) to the top level (ID level) a higher decision-making power pertains. Therefore, the decision flows are top-down.

The simulation analysis

5.15

To prove the benefits of the selected kind of cooperation, we run the described computational model (Coop_Case), in which three buyer and three supplier agents act, and measure specific performances referred to the single ID firm and the ID as a whole. Then, these results are compared with those achieved by running a different computational model in which the inter-firm relationships are characterized by the absence of cooperation (NoCoop_Case).

5.16

Furthermore, to evaluate the benefits of the cooperation in different competitive scenarios and diverse ID organizational structures we define different experimental settings. In particular, the competitive scenarios are defined by using three different patterns of final demand variability (low, medium, high). We assume that a high level of final demand variability corresponds to a high level of market instability of the competitive scenario. As regard to the ID organizational structure, we consider two different models. The first one corresponds to the canonical form of the Marshallian ID characterized by the presence of many small, locally owned firms specialized in just one phase, or a few phases, of the production processes typical of the district (Becattini, 1990; Markusen, 1996; Piore and Sabel, 1984). The second one corresponds to a new emerging model of ID, characterized by the presence of one or more leader firms with superior equipment (Carbonara et al., 2002; Carbonara, 2002; Markusen, 1996).

5.17

Therefore, we compare the above defined two cases (Coop_Case and NoCoop_Case) in six experimental settings. In Table 3 the variables characterizing each experimental setting are shown. We model the final demand by adopting a uniform distribution with mean equal to 250. The three patterns of final demand variability are modeled by using three increasing values of standard deviation. We use two patterns of buyer production capacity to represent the existence and the absence of the leader firm within the ID. In particular, a concentration of the production capacity on a single buyer, resulting in an uneven distribution of the production capacity within the buyer stage, corresponds to the existence of the leader firm. On the contrary, the equal distribution of the production capacity in the stage represents the Marshallian ID.

5.18

In each experimental setting the supplier production capacity of each supplier agent is a constant and is modeled by using a uniform distribution with mean equal to 250 and standard deviation equal to 28.87.

Table 3: Values of the variables to define the experimental settings
Buyer Demand
Low demand variability (LV)	Unif[200,300]
Medium demand variability (MV)	Unif[150,350]
High demand variability (HV)	Unif[50;450]
ID organizational structure
Marshallian ID	Production Capacity Buyer 1 = Unif[200,300] Production Capacity Buyer 2 = Unif[200,300] Production Capacity Buyer 3 = Unif[200,300]
ID with leader firm	Production Capacity Buyer 1= Unif[75 , 175] Production Capacity Buyer 2 = Unif[200,300] Production Capacity Buyer 3 = Unif[325,425]

5.19

To evaluate the benefits of the cooperation in each experimental setting, we use the following performance measures:

• total customer demand satisfied by the ID (ToT_dem%);
• production capacity utilization of the buyer stage (B_CapSat%);
• production capacity utilization of the supplier stage (S_CapSat%);
• index that measures the balance of the supplier production capacity utilization (Unbalance_I).

5.20

In the Table 4 the results of simulation for 10.000 cycles are shown. The results show that cooperation has a positive impact on the ID performances. In fact, first the total customer demand satisfied by the ID characterized by cooperative relationships is higher than that satisfied by the ID with absence of cooperation. Second, the production capacity utilization of buyer and supplier stages is higher in the case of cooperation. The cooperation effect is more complex with regard the balance of the supplier production capacity utilization. In fact, in the experimental settings associated with the Marshallian ID, the results are counterintuitive. We expected that the cooperation determined a higher balance of the supplier production capacity utilization, given that one of the aims of the cooperation is to balance the supplier production capacity utilization, but we measured an opposed result. This is due to the cooperation among buyers, which advantages only a few suppliers with a negative impact on the total supplier stage. Similar results hold for the ID with leader firm with medium/high demand variability.

5.21

When demand variability is low, the cooperation in ID with leader firm improves the balance of the supplier production capacity utilization. This result depends on the buyer stage configuration, which allows only the leader firm to actually cooperate. As a consequence the negative impact on the supplier stage, due to the cooperation among buyers, is limited.

Table 4: . Simulation results.
Experimental settings	ToT_dem%	B_CapSat%	S_CapSat%	Unbalance_I
Marshallian ID & LV
- Coop_Case	92.9	92.9	93.1	1.2
- NoCoop_Case	91.5	92.1	91.8	0.15
Marshallian ID & MV
- Coop_Case	90.4	90.5	90.4	0.89
- NoCoop_Case	87.8	88.5	81.1	0.26
Marshallian ID & HV
- Coop_Case	86.0	84.2	83.3	0.8
- NoCoop_Case	80.4	78.5	79.8	3.2
ID with leader firm & LV
- Coop_Case	83.3	89.2	85.1	21.10
- NoCoop_Case	79.0	83.6	81.1	23.95
ID with leader firm & MV
- Coop_Case	83.2	87.9	82.9	20.43
- NoCoop_Case	79.0	84.4	71.2	19.50
ID with leader firm & HV
- Coop_Case	81.0	82.2	78.4	17.8
- NoCoop_Case	74.9	78.5	74.1	14.9

5.22

In Figure 2 ToT_dem% and the Unbalance_I in the three different competitive scenarios are depicted for the Marshallian ID and the ID with leader firm, respectively. In the Marshallian ID the benefit of cooperation on the satisfied total customer demand raises when the demand variability increases. When the final demand is more stable (LV), the cooperation is not relevant given that the performance measured in the case of cooperation and absence of cooperation is similar. As regard the Unbalance_I, the results show that when the demand variability increases the balance of the supplier production capacity utilization tends to increase. In fact, when the demand variability is high, firms adopt more frequently cooperative behaviors, which in turn impact on the balance.

Figure 2. Simulation results for the Marshallian ID and the ID with leader firm.

5.23

In the ID with leader firm the cooperation is beneficial on the satisfied total customer demand in all the three competitive scenarios and is more critical when the demand uncertainty is high. Similarly to the Marshallian ID, as the demand variability increases the Unbalance_I decreases.

5.24

With specific regard to the leader firms we notice that it is worth cooperating. In fact, the production capacity utilization improves when the ID is characterized by cooperative relationships. A similar result holds for the leader firm preferential supplier (Table 5).

Table 5: Production capacity utilization of leader firm and its preferential supplier
	LV	MV	HV
Leader firm
- Coop_Case	73.6	73.2	69.9
- NoCoop_Case	65.6	64.5	62.2
Preferential supplier
- Coop_Case	96.8	94.4	88.7
- NoCoop_Case	93.9	89.6	82.3

5.25

Comparing the performances achieved for the two different ID organizational structures, the Marshallian ID performs better than the ID with leader firm. However when the demand variability increases, the ID with leader firm makes up for the disadvantage in comparison with the Marshallian ID. In fact, the satisfied total customer demand in the Marshallian ID decreases more than in the ID with leader firm when the demand variability increases.

Conclusions

6.1

In this paper a computational approach based on the MAS has been proposed to study cooperation and competition within IDs. The rationale to use the computational approach is to test the benefits of cooperation within IDs and study the effects of cooperative behaviors in different competitive scenarios. In the paper we have developed an agent-based computational model of the cooperative relationships aimed at both balancing the utilization of supplier production capacity and minimizing the customer unsatisfied demand.

6.2

Then we have run the computational model for two different level of cooperation (presence vs. absence). The performances related to the satisfied total customer demand, the production capacity utilization of the buyer and supplier stages, and the balance of the supplier production capacity utilization, have been measured and compared.

6.3

A further analysis has been carried out by comparing the ID performances when the demand variability increases and in two different ID organizational structures, namely Marshallian ID and ID with leader firm. The results show that, to a great extent, the cooperation has a positive impact on the ID performances. However, under different competitive scenarios the two ID organizational structures perform in different ways. Therefore, a further research is necessary to better understand the relations among organizational structure, competitive environment, and cooperation. This could be useful to define coordination forms that are more appropriate to different ID structures and competitive scenarios. To this aim we intend to provide the proposed computational model with a dynamic behavior, which is given by agents that have a learning capability and changing decision rules.

Program Code

Click here for a ZIP Archive containing the Program Code (1.2 MegaBytes)

Notes

¹ The industrial district is a specific type of geographical cluster. The latter is defined by Porter (1998) as a geographically proximate group of interconnected companies and associated institutions (for example universities, standards agencies, and trade associations) in particular fields, linked by commonalities and complementarities. Clusters also promote both competition and cooperation. Therefore, the cooperative and competitive relationships we investigate in the paper are not specific of the Italian industrial districts but are typical features of the geographical clusters localized in other countries.

² The supply chain capacity is equal to the minimum between the buyer and the supplier capacity.

References

AGENTBUILDER Reference Manual and AgentBuilder Users Guide, Version 1.3 Rev. 0, 2000, Reticular Systems, Inc. http://www.agentbuilder.com.

ALBINO, V., Carbonara, N., Schiuma, G., 2000 Knowledge in inter-firm relationships of an industrial district, Industry and Higher Education, December, 404-412.

AMIN, A., Thrift, N., 1992, Neo-Marshallian nodes in global networks, International Journal of Urban and Regional Research, No. 16, 571-587.

AXELROD, R., 1997, Advancing the art of simulation in the social sciences, Complexity, Vol. 3, No. 2, 16-22.

BAGNASCO, A., 1988, La Costruzione Sociale del Mercato, Il Mulino, Bologna.

BAPTISTA, R., Swann, P., 1998, Do Firms in Clusters Innovate More?, Research Policy, Vol. 27, No. 5, 525-540.

BECATTINI, G., 1987, Mercato e Forze locali: il distretto industriale, Il Mulino, Bologna.

BECATTINI, G., 1990, The Marshallian industrial district as a socio-economic notion, in Becattini, G., Pyke, F., Sengenberger, W. (eds), Industrial districts and Inter-firm co-operation in Italy, International Institute for Labour Studies, Geneva, 37-52.

BECATTINI, G., 1997, Prato nel mondo che cambia, in Becattini G. (ed.), Prato, storia di una città. Il distretto industriale (1943-1993), Le Monnier, Firenze.

BECATTINI, G., Pyke, F., Sengenberger, W. (eds), 1990, Industrial districts and Inter-firm co-operation in Italy, International Institute for Labour Studies, Geneva.

BERENDS, P., Romme, G., 1999, Simulation as a research tool in management studies, European Management Journal, Vol. 17, No. 6, pp. 576-583.

BOERO, R., Squazzoni, F., 2002, Economic performance, Inter-firm relations and local institutional engineering in a computational prototype of Industrial Districts, Journal of Artificial Societies and Social Simulation, Vol. 5, No. 1, printed from https://www.jasss.org/5/1/1.html .

BOND, A.H., Gasser, L. (eds), 1988, Readings in Distributed Artificial intelligence,Morgan Kaufmann Publishers, San Mateo, CA.

BRAZIER, F.M.T., Dunin-Keplics, B., Jennings, N.R., Treur, J., 1995, Modeling Distributed industrial processes in a Multi-agent Framework, printed from http://www.wmw.ac.uk

BRUSCO, S., 1990, The idea of the industrial district: Its genesis, in Becattini, G., Pyke, F., Sengenberger, W. (eds), Industrial districts and Inter-firm co-operation in Italy, International Institute for Labour Studies, Geneva, 10-20.

CARBONARA, N., 2002, New models of inter-firm network within industrial district, Entrepreneurship and Regional Development, (forthcoming).

CARBONARA, N., Giannoccaro, I., Pontrandolfo, P., 2002. Supply chains within industrial districts: a theoretical framework, International Journal of Production Economics, Vol. 76, No. 2, 159-176.

CARLEY, K.M., Prietula, M.J., 1998, Webbots, Trust, and Organizational Science, in Carley, K.M., Gasser, L. and Prietula, M.J. (eds), Simulating Organizations. Computational Models of Institutions and Groups, The MIT Press, Massachusetts.

CERTET,1995, Il distretto industriale brianzolo del legno-arredo, Collana OPES - Unione delle Camere di Commercio della Lombardia, Quaderno 6.

COR, G., Grandinetti, R., 1999, Evolutionary Patterns of Italian Industrial Districts, Human Systems Management, Vol. 18, No. 2, 117-130.

DE TONI, A., Nassimbeni, G., 1995, Supply network: genesis, stability and logistics implications. A comparative analysis of two districts, Omega, International Journal of Management Science, Vol. 23, N. 4, 403-418.

DEI OTTATI, G., 1994, Co-operation and Competition in the Industrial Districts as an Organizational Model, European Planning Studies, No. 4.

DEI OTTATI, G., 1995, Tra Mercato e Comunità: Aspetti concettuali e ricerche empiriche sul distretto industriale, F. Angeli, Milano.

DIGIOVANNA, S., 1996, Industrial Districts and regional economic development: a regulation approach, Regional Studies, Vol 30, No. 4, 373-386.

DURFEE, E., 1988, Coordination of distributed problem solvers,Kluwer Academic Publishers, Boston.

FERBER,J., 1999, Multi-Agent Systems. An Introduction to Distributed Artificial Intelligence, Addison-Wesley, London.

FININ, T., Fritzon, R., McKay, D., McEntire, R., 1994, KQML as an agent communication language, Proceedings of the Third International Conference on Information and Knowledge Management, ACM Press.

FIORETTI, G., 2001, Information structure and behaviour of a textile industrial district, Journal of Artificial Societies and Social Simulation, Vol. 4, No. 4, printed from https://www.jasss.org/4/4/1.html

FISHER, M.L., 1994, Making supply meet demand in an uncertain world, Harvard Business Review, May-June, 83-93.

FUÀ, G., 1983, Rural Industrialization in Later Development Countries: The case of Northeast and Central Italy, Banco Nazionale del Lavoro Quarterly Review, December, 351-377.

GANDOLFI, V., 1988, Aree sistema: internazionalizzazione e reti telematiche, F. Angeli, Milano.

GANDOLFI, V., 1990, Relazionalità e cooperazione nelle aree-sistema, Economia e Politica Industriale, No. 65, 93-117.

GAROFOLI, G., 1981, Lo sviluppo delle aree "periferiche" nell'economia italiana degli anni '70, L'industria, Vol. 3, 391-404.

GOLDSPINK, C., 2002, Methodological Implications of Complex Systems Approaches to Sociality: Simulation as a foundation for knowledge, Journal of Artificial Societies and Social Simulation, Vol. 5, No. 1, printed from https://www.jasss.org/5/1/3.html .

GRANOVATTER, M., 1985, Economic action and social structure: the problem of embeddedness, American Journal of Sociology, Vol. 91, No. 3.

HARRISON, B., 1993, "Industrial districts: old wine in new bottles?", Regional Studies, No. 26, 469-483.

HUHNS, M.N., Stephens, L.M., 2000, Multiagent Systems and Societies of Agents, in Weiss G. (ed.), Multiagent Systems. A modern approach to distributed artificial intelligence, The MIT Press, Cambridge (Massachusetts).

KANG, M., Waisel, L.B., Wallace, W.A., 1998, Team-Soar: A Model for Team Decision Making, in Carley, K.M., Gasser, L. and Prietula, M.J. (eds), Simulating Organizations. Computational Models of Institutions and Groups, The MIT Press, Massachusetts.

KRUGMAN, P.R., 1991, Geography and Trade, MIT Press, Cambridge.

LIN, F., Shaw, M.J., 1998, Reengineering the order fulfillment process in supply chain networks, The International Journal of Flexible Manufacturing Systems, Vol. 10, 197-229.

MAILLAT, D., Lecoq, B., Nemeti, F. and Pfister, M., 1995, Technology district and innovation: the case of the Swiss Jura Arc, Regional Studies, Vol. 29, No. 3, 251-263.

MARKUSEN, A., 1996, Sticky places in Slippery Space: A typology of Industrial Districts, Economic Geography, Vol. 72, No. 3, 293-313.

MARSHALL, A., 1919, Industry and Trade, Macmillan, London.

MARSHALL, A., 1920, Principles of Economics, Macmillan, London.

PANICCIA, I., 1999, The performance of IDs. Some insights from the Italian case, Human Systems Management, Vol. 18, 141-159.

PILOTTI, L., 1999. Evolutionary and adaptive local systems in North East Italy: Strategies of localized learning, open leadership and cooperation. Towards imperfect "communitarian capitalism", Human Systems Management, Vol. 18, 87-105.

PIORE, M., Sabel, C.F., 1984, The Second Industrial Divide, Basic Books, New York.

PORTER, M. 1998, Clusters and the new economics of competition, Harvard Business Review, No. 76, 77-90.

PORTER, M., 2001, Regions and the New Economics of Competition, in Scott A. J. (ed.), Global City-Regions, Oxford University Press, New York, 139-157.

PYKE, F., Sengenberger, W., 1990, Introduction, in Becattini, G., Pyke, F., Sengenberger, W. (eds), Industrial districts and Inter-firm co-operation in Italy, International Institute for Labour Studies, Geneva, 1-10.

RABELOTTI, R., 1995, Is there an Industrial District Model? Footwear districts in Italy and Mexico compared, World Development, Vol. 3, No. 1, 29-41.

SABEL, C., 1982, Work and politics, Cambridge University Press, Cambridge (Massachusetts).

SAXENIAN, A., 1996, Regional Advantage. Culture and competition in Silicon Valley and Route 128, Harvard University Press.

SFORZI, F., 1990, The Quantitative importance of Marshallian industrial districts in the Italian Economy, in Becattini, G., Pyke, F., Sengenberger, W. (eds), Industrial districts and Inter-firm co-operation in Italy, International Institute for Labour Studies, Geneva, 75-107.

SILVESTRELLI, S., 1984, L'impianto, in Rispoli M (ed), L'impresa industriale, Il Mulino, Bologna.

STORPER, M., Harrison, B., 1991, Flexibility, hierarchy and regional development: The changing structiure of industrial production systems and their forms of governance in the 1980s, Research Policy, Vol. 20, 407-422.

THOMAS, S.R., 1993, An Agent Oriented Programming Language, PhD Thesis, Stanford University.

VISCONTI, F., 1996, Le condizioni di sviluppo delle imprese operanti nei distretti industriali, EGEA, Milano.

WOOLDRIDGE, M., 2000, Intelligent Agents, in Weiss G. (ed.), Multiagent Systems. A modern approach to distributed artificial intelligence, The MIT Press, Cambridge (Massachusetts).

Return to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, [2003]