Today, investments in R&D—be it through higher education institutions or science-industry networks—are expected to immediately produce high commercial returns. Science policymakers, innovation managers, and even the public are often disappointed and raise legitimacy issues, when such returns fail to materialise promptly. These situations show the limits of conventional steering, control and policy making associated with research funding.
Science policy experts often refer to the frustration of third parties, for whom the messy and complicated features of funding targets simply “do not seem to compute.” Better solutions to help improve returns from research funding are therefore needed. Unfortunately, what is referred to as the science of science policy is still in its infancy. To assist in R&D policy decisions concerning funding issues, modelling and simulation can help. Indeed, by harnessing the capabilities afforded by complex systems analysis tools, it is possible to gain unprecedented insights into the consequences of specific policy decisions. The question is how such tools can contribute to policy making in complex environments such as science, research and innovation.
Dealing with complexity
Without a doubt, socio-economic systems are confronted with a high degree of complexity. This is particularly true, when it comes to the development of new knowledge, its diffusion, and its commercial application in innovation. Furthermore, the trouble is that actors of such activities—that we can refer to as agents—are confronted with true uncertainty. This makes any forecasts and predictions on innovation success or failure impossible. Any analytical approach that tries to offer guidance and support for policy decision makers has to acknowledge this intermingling of rich complexity and uncertainty.
Enters the agent-based SKIN model, which stands for Simulating Knowledge Dynamics in Innovation Networks. It has been designed to simulate knowledge generation and diffusion in inter-organisational research and innovation networks. Since its first prototype in 2001, it has been developed further into a platform with many modules and applications. And it has since been adopted by a number of policy modelling studies, applying it for science policy.
The largest application of the SKIN model to date focuses on impact assessment and ex-ante evaluation of European funding policies in the Information and Communication Technologies (ICT) research domain. The corresponding version of the model, referred to as INFSO-SKIN, had been developed for the Directorate General Information Society and Media of the European Commission (DG INFSO) in 2011. It was intended to help to understand and manage the relationship between research funding and the goals of EU policy.
Testing future policy
The agents of the INFSO-SKIN application are research institutions such as universities, large diversified firms or small and medium-sized enterprises (SMEs). The model simulates real-world activity in which the funding calls of the European Commission specify the composition of consortia, the minimum number of partners, the length of the project, the deadline for submission and a range of capabilities, including a sufficient number of which must appear in an eligible proposal, as well as the number of projects that will be funded.
The model implemented rules of interaction replicating the actual Framework Programme (FP) decision paths. To increase the usefulness of the model to policy makers, the names of the rules within the model closely matched FP terminology. For the Calls numbered 1 to 6 in FP7, the model used empirical information incorporating the number of participants and the number of funded projects, together with data on project duration, average funding and size; the latter measured through the number of participants.
Analysis of this information produced data on the functioning of actual FP collaborative networks and their internal relationships. Using this data in the model provided a good match with the empirical data from EU-funded ICT networks in FP7. Indeed, the model accurately reflected what actually happened. And it could be used as a test bed for potential future policy choices.
Changing parameters within the model is analogous to applying different policy options in the real world. The model could thus be used to examine the likely real-world effects of different policy options before they are implemented. Thus, altering elements of the model that equate with policy interventions—such as the amount of funding, the size of consortia, or to encourage specific sections of the research community—made it possible to use INFSO-SKIN as a tool for modelling and evaluating the results of specific policy interactions; typically occurring between policy interventions, funding strategies and agents.
On its most general level, SKIN is an agent-based model. Its agents are knowledge-intensive organisations, which try to generate new knowledge by research—be it basic or applied—or by creating new products and processes through innovation. These agents are located in a changing and complex social environment, which evaluates their performance. For example, what matters are the market performances, if the agents target innovation, or those of the scientific community, if the agents target publications through their research activities.
Agents have various options to act: each agent has an individual knowledge base called its ‘kene,’ which it takes as the source and basis for its research and innovation activities. The agent’s kene is not static. The agent can learn, on its own by doing incremental or radical research. Or it can also learn from others, by exchanging and improving knowledge in partnerships and networks. The latter feature is important, because research and innovation happens in networks, both in science and in knowledge-intensive industries.
This is why SKIN agents have a variety of strategies and mechanisms for collaborative arrangements. They are able to choose partners, form partnerships, start knowledge collaborations, create collaborative outputs, and distribute rewards. In short, usually a SKIN application has agents interacting on the knowledge level and on the social level while both levels are interconnected. It is all about knowledge and networks.
This general architecture is quite flexible, which is why the SKIN model has been called a platform. It has been tested for a variety of applications ranging from small scale—such as simulating the Vienna biotech cluster—via intermediate size —through simulations of the Norwegian defense industry—to large-scale, such as INFSO-SKIN described above.
The SKIN model applications use empirical data and claim to be realistic simulations insofar as the aim is to derive conclusions by so-called inductive theorising. This means that the quality of the SKIN simulations derives from an interaction between the theory underlying the simulation and the empirical data used for calibration and validation.
These new approaches enable the modelling of science policy initiatives to take into account more parameters than previously possible. It also makes it possible to perform simulations to forecast potential impacts of proposed science policy measures. Yet, it is still early days for this field of the science of science policy.
Looking into the future of the SKIN model development, the establishment of this conceptual framework combining the application of empirical research, computational network analysis, and agent-based modelling will yield a more integrated and comprehensive understanding of science policymaking than has been achieved to date. In contrast to conventional methods of social research, this approach will be capable of dealing with that fact that research and innovation do not follow a linear path and are highly complex.
Petra Ahrweiler, director of the European academy of technology and innovation assessment in Bad Neuenahr-Ahrweiler, Germany
Featured image credit: CC BY-SA 3.0 by Nick Youngson from ImageCreator
Go back to the Special Issue: Alternative research funding