Science, technology and innovation for economic growth: Linking policy research and practice in ‘STIG Systems’
GAICIL Insights
Everyone loves innovation policy until it picks a winner.1 Science, technology and innovation for economic growth: Linking policy research and practice in 'STIG Systems', by Philippe Aghion, Paul A. David and Dominique Foray, examines that awkward moment. Policy for science, technology and innovation is widely embraced in principle and contested in practice. The authors reflect on systems-theoretic approaches to the interdependent policy issues around the dynamics of science, technology and innovation and their relationship to growth.
AI ministries might start here: build an evaluation and revision clause into every AI program from day one. The controversy over picking winners is easier to survive when the policy can admit, in public and on schedule, which of its own bets failed.
The critical question: what kind of systems paradigm is likely to prove particularly fruitful here? The authors examine evolutionary approaches, approaches built on innovation and creative destruction, and complex system dynamics. Each is conceptually attractive, each has virtues and limitations, and both become visible the moment one tries to connect systems-relevant research with practical policymaking. Theory looks cleanest before it meets a ministry.
The conclusion captures innovation policy's political economy. Technology and innovation policy for growth is widely accepted, yet becomes politically controversial once implementation goes beyond support for exploratory, far-from-commercialization research and enters specific details perceived to have differential effects on particular markets, institutions and industries. There are good reasons for caution in those realms. Yet the growth potential of research and innovation is too clear to abandon policy efforts simply because they are difficult or politically charged. Difficulty is a reason for care, and a poor excuse for retreat.
Experimentation is the authors' prescription. It is critical to experiment with different ways of structuring policy to overcome conceptual and practical challenges. Where creating new scientific and technological knowledge and finding effective uses for it are central, advances are mostly incremental and cumulative over long periods. Policy, too, must learn and adapt over time. Innovation policy should itself be innovated.
The application to AI is direct. AI policy is a systems problem par excellence. It spans science, technology and innovation, and its effects run through complex systems of research institutions, firms, markets and regulations. A systems view shows how interventions in one part affect others, such as how research funding shapes talent flows or how regulation shapes investment. A visa rule can be an AI research policy; a procurement rule can be an AI market policy.
The political controversy is already visible. Support for basic AI research is broadly accepted. Policies favoring particular firms, applications or sectors are contested because they create winners and losers. The authors' counsel: proceed with caution without abandoning such efforts, given the growth potential at stake.
Experimentation suits AI's pace. AI evolves rapidly and policy effects are uncertain. Experimenting with designs, evaluating results and adapting is more promising than committing to a single approach. Pilot programs, regulatory sandboxes and staged implementation are examples. A policy built to be revised beats a policy built to be permanent.
Cumulative advance applies to technology and policy alike. AI progress builds on prior research, and effective AI policy builds on accumulated experience. Patience and sustained commitment are required, as payoffs from research and policy learning may take years. Election cycles are shorter than learning cycles, which is part of the problem.
The paradigms complement each other. Evolutionary approaches emphasize variety and selection; creative destruction approaches emphasize competition and entry; complex systems approaches emphasize interdependence and feedback. AI policy can draw on all three, each where it offers most insight.
The message is pragmatic. Innovation policy is necessary because the growth potential of research and innovation is large. It is difficult because it involves complex systems and political controversy. The response is careful, experimental policymaking informed by research and attentive to systemic effects.
Introduction
This paper reflects on the relevance of "systems-theoretic" approaches to the interdependent policy issues relating to the dynamics of science, technology and innovation and their relationship to economic growth. Considering the approach that characterizes much of the current economics literature's treatment of technology and growth policies, we pose the critical question: what kind of systems paradigm is likely to prove particularly fruitful in that particular problem-domain? Evolutionary, neo-Schumpeterian, and complex system dynamics approaches are conceptually attractive, and we examine their respective virtues and limitations. Both qualities are readily visible when one tries to connect systems-relevant research with practical policy-making in this field.
Takeaway
Technology and innovation policy for growth is widely accepted, but it immediately becomes politically controversial when its implementation goes beyond the support of “exploratory” and “far-from-commercialization” research, and enters into specific details that are perceived to have differential effects on particular markets, institutions and industries. There are good reasons for caution in entering those realms, but the growth potential of R&D and innovation is too clear to abandon policy efforts simply because they are difficult to implement or politically charged. It is therefore critical to experiment with different ways of structuring policy in this area so as to overcome the various conceptual and practical policy challenges. This essay has sought to confront these challenges by addressing the issue of the practical correction of market failures and policy coordination failures, by finding an appropriate systems paradigm and set of (simulation) tools to work within it in order to assess the dynamics of interactions among policy initiatives, and, finally, by addressing the problems of practical policy evaluation. Closing words of caution are in order on at least two points, both having to do with “ambition”. The first relates to the “scientific” ambitions of those who, through research, aim to improve the quality of STIG (and related) policy designs and their implementation. Complex systems give rise to “outcomes” that are driven by processes beyond the control of individual agencies or their policy advisors.1
One may experiment in a virtual environment using a simulation model to learn about certain qualitative dynamic properties of a complex system. However, simulation models often provide little information about critical determinants of the dynamics of systems of human actors, some of whom pursue adaptive strategies but not necessarily in all their spheres of activity. A further complicating factor is that policy-decision makers and implementation agents are themselves part of the interdependent processes and may contribute to the creation of destabilizing positive feedback dynamics. Empirical detail will best be absorbed into the structure of the model and the specification of its parameters only to specify some among the myriad features of the world that could be studied, and in order to quantify some dynamical relationships that are believed on analytical and experiential grounds to be critical in rendering the simulations able to provide robust insights that could be informative in setting policy strategies. The goal in such endeavors is, after all, not painstakingly realistic detail, but a simplified model or map with just enough detail to enable effective decisions to be made. The task of navigation in the terrain of “political economics” will not be advanced by furnishing either researchers or policy-makers with “a map that is as big as the territory”. Our last words are saved for those who aspire to “direct” the processes of scientific advance, technological change, and innovation along certain trajectories so as to improve the economic welfare and material well-being of societies and nations.
Public agency interventions in STIG processes are unlikely to yield political credits in the time frame within which most politicians and public servants in representative democracies have to function, unless their objectives are confined to redistributing resources gathered by taxation among their respective constituents. In the realms where creating new scientific and technological knowledge and finding the most effective ways to use it are central, the advances are mostly incremental and cumulative over long periods. Hence, the assignment of responsibilities for significant successes can only be retrospective rather than contemporaneous. Moreover, in complex, contingent, and at best only partially understood dynamical processes, individuals who seek to claim responsibility for changing the system’s “performance” for the better are all too likely to find that they are the recipients of blame (albeit in many instances equally unjustified) for outcomes that were unanticipated and unwanted.
Notes
- Philippe Aghion, Paul A. David & Dominique Foray, Science, technology and innovation for economic growth: Linking policy research and practice in ‘STIG Systems’, 38 Rsch. Pol'y 681 (2009), https://doi.org/10.1016/j.respol.2009.01.016. Takeaway source: 6. Concluding cautions about the ambitions of STIG policy research and practice, reproduced from a free author or repository copy. ↩
Cite as: Philippe Aghion, Paul A. David & Dominique Foray, Science, technology and innovation for economic growth: Linking policy research and practice in ‘STIG Systems’, 38 Rsch. Pol'y 681 (2009), https://doi.org/10.1016/j.respol.2009.01.016
Header image: BU Outdoor Research Facility inside.jpg by strangebiology, CC BY 2.0, via Wikimedia Commons.
