Analysis of dynamics, structures and agent relationships in Regional Innovation Systems/Analisis de las dinamicas, estructuras y relaciones de los agentes en Sistemas Regionales de Innovacion/Analises das dinamicas, estruturas e relacoes dos agentes nos Sistemas Regionais de Inovacao.

Introduction

Antioquia is a Colombian department and one of the nation's principal economic drivers. With a population of over 6.3 million (2013), Antioquia's gross domestic product (GDP) and economic growth rate of 3% (Londono, Restrepo, Alvarez, & Ospina, 2018) are above the national average. Antioquia's economy is based on the manufacturing industry, service sector and natural resource exploitation (Llisterri & Pietrobelli, 2011). Income, industry, population, post-secondary education and investment in research and development are concentrated in the metropolitan area of Medellin. Along with Bogota and Cali, the three cities form the "Golden Triangle" (Secretaria de educacion para la cultura de Antioquia, 2011).

Development theory from an innovation systems perspective and in particular regional innovation systems (henceforth RIS) has been influenced by different schools of thought including the school of evolutionary economics, institutional economics, new regional economies, the learning economy, the innovation economy and network theory (Quintero & Robledo, 2013).

For over two decades the development of Antioquia's RIS has been based on local initiatives with a bottom-up approach, with key agents in the process considered to be the foundation (Llisterri & Pietrobelli, 2011). As early as the 1980s science and technology in Antioquia had strengths and a certain structure in the academic, productive and public sectors. During this period, the challenge was to develop a policy for science, technology and innovation (STI) that would evolve around agent interactions (Llisterri & Pietrobelli, 2011).

With changes to Colombia's political constitution in the 1990s, certain powers and functions were granted to the regions which as a result had the autonomy to make decisions to promote capacity building and institutional strengthening, as well as to create the basic infrastructure for a science and innovation system. Nevertheless, policies developed in the regions were not sufficient for shaping the dynamics of innovation (Llisterri & Pietrobelli, 2011).

In the last decade the university-business-state committee was created, and ties were built with regional competitiveness councils and the departmental Science, Technology and Innovation council. This allowed Antioquia to advance in the field of innovation among RIS agents (Llisterri & Pietrobelli, 2011).

The objective of this paper is to evaluate the structures, interactions and dynamics of agents in Antioquian RIS using an agent-based simulation model. To this end, part 2 describes the context within which RIS are developed and presents the structures and interactions of regional agents. Part 3 looks at RIS from a complex adaptive systems (CAS) perspective, which enables a better understanding of how systems operate as a support for the cognitive process of decision making for policymakers. Part 4 highlights the importance of addressing RIS dynamics and the RIS of Antioquia in particular. An agent-based model (ABM) is proposed and the results of its simulation are presented. Finally, conclusions and suggestions for future research are provided, with the aim of achieving a better understanding of the RIS of Antioquia as a product of interactions between its agents.

Theoretical framework

The concept of the RIS has been widely discussed by various authors. Since the 1980s and the beginning of the 1990s, pioneering works by Freeman (1987), Lundvall (1992) and Nelson (1992), provided the basis for the definition and characterization of the national innovation system. Asheim and Gertler (2005) define it as the institutional infrastructure that supports innovation in a region's productive structure. In 1992, Cooke coined the term "regional innovation system" and attempted to explain it as the best practices in the context of regional innovation by engaging in a process of learning through interactions (Su & Chen, 2015). In accordance with this, RIS have two central elements: 1) a strong and dense network of relationships between autonomous and heterogenous agents (subsystems of "explorers, exploiters, catalysts and regulators of knowledge") and 2) a level of competitiveness attributed to the co-evolution of the productive organization and to the location of formal and informal institutions within the system (Quintero & Robledo, 2013).

These definitions are complex and require further explanation, as they refer to interactions between different actors with the purpose of producing innovation collectively through the interaction of knowledge-based assets. This interaction is studied by looking at a set of networks made up of agents or actors where capacities, knowledge and feedback are shared. This allows them to benefit from their own infrastructure in order to modify, achive and increment knowledge and innovation (Lau & Lo, 2015). These actors operate in an area that is geographically defined and they are governed by common policies at the sectoral, regional and national levels (Doloreux, 2002).

While still recognizing the existence of sectoral and national innovation systems, this definition suggests that demarcating the RIS is key (Todtling & Trippl, 2005). If the implementation of policies at the regional level is desired, the following three parameters must be defined: firstly, the existence of industrial specialization patterns, as this generates localized economies that exist between companies with similar productive contexts and where knowledge is generated within them (Li, 2015); secondly, the influence of knowledge on the process of innovation and tacit knowledge; and thirdly, issues that are demarcated spatially. This confirms that all regions behave differently and cannot be measured in the same way.

Each RIS requires a particular type of analysis and approach in order to define and describe it, given that all RIS are made up of various subsystems of actors or agents. The first subsystem is that of the exploiters who apply and take advantage of knowledge. All companies with their respective clients, suppliers and competitors belong to this category. The second subsystem is comprised of actors or agents that explore new and existing knowledge and are defined as explorers. This group of actors produces and diffuses knowledge and skills. Research institutes including laboratories and educational institutions such as universities are grouped within this category (Todtling & Trippl, 2005). This means that regional development in innovation requires the dedication and commitment of these agents. The complex interrelationships that emerge between these actors are referred to as the innovation habitat (Oliveira, Echeveste, Cortimiglia, & Goncalves, 2017).

Actors in these systems have become more specialized in recent years in terms of their functions (Zollo, De Crescenzo, & Ponsiglione, 2011) and have emerged as four types of agents: explorers, defined as those who produce new ideas, methods and techniques from the fringes of knowledge; intermediaries (Watkins, & Horley, 1986) or catalysts (Smits & Kuhlmann, 2004), defined as those who facilitate the transfer, acceptance and use of knowledge (Ruiz, Quintero, & Robledo, 2016); exploiters, defined as actors who transform knowledge into market value and include companies (especially small and medium ones); and the fourth actor is the national innovation agency, which is the government player charged with establishing guidelines and coordinating the innovation network. An additional element (which is equally important to policy creation and implementation in a given region) is the environment or setting in which the RIS develops and evolves. This element is not identified as a key agent or actor in the current literature (Quintero, Ruiz, & Robledo, 2017).

2.1. RIS Proposed model for Antioquia

The RIS to be studied here spans the territory of the department of Antioquia, located in northwestern Colombia. Figure 1 presents the proposed model for the RIS, which is comprised of three actors with the following characteristics (Carlsson, Jacobsson, Holmen, &amp...