I just came across MIT‘s Strategic Engineering website, and flicked through a very interesting conference paper entitled Time-Expanded Decision Network: A New Framework for Designing Evolvable Complex Systems. This piece, by Matthew Silver and Olivier de Weck, describes a method for helping choose between different designs of high-cost, complex engineering system in the context of uncertainty as to the future operating environment.
The idea behind it all is that when you design a complex system such as a spacecraft (the paper was written by members of MIT’s Department of Aeronautics & Astronautics), there is considerable uncertainty as to the operating demands on the system (in this case, how many moon or mars missions will be required, thus influencing the suitability of different designs in terms of load efficiency). Such uncertainty means that adaptability of the system is important, and yet such complex systems often “exhibit high degrees of architectural lock-in”, hindering overall adaptability. In such circumstances (presumably where you can’t have a single system that does everything equally well) it is useful to consider the switching costs of moving from one design to another – if these switching costs are low enough, a change in the operating environment might favour a switch to a more efficient vehicle. The question becomes, though, given a pre-determined set of initial designs, which designs are most robust across a range of operating scenarios, and where are the most effective cost-savings to be found in terms of lowering switching costs? The model employed, which they call “Time-Expanded Decision Networks” (TDN) is really well done, employing scenarios, decision-trees and network theory to produce a useful “matlabbable” tool.
Now, while I’m interested in space travel and in awe of complex engineering projects, what really got me thinking in this paper was a potential analogy to another very complex design problem currently being undertaken by governments around the world – the “New Financial Architecture” that is being designed by the G20 as well as every man and his dog.
Like Heavy Lift Launch Vehicles, financial regulation is a complex system which involves a relatively high degree of lock-in, particularly when the imperatives of coordination and international cooperation are as high as they are today. And yet, as we have too-well learned recently, the future remains uncertain. When we even begin to apply suitable regulation to a sector, we too often regulate for crises past, rather than the crisis to come.
While we might not be able to design a regulatory system that can itself adapt to every possible circumstance in terms of economic growth, investor behaviour, geopolitical power or ideological shifts, capital flows or market stability (etc etc), perhaps the G20 negotiators should draw a few lessons from the MIT aeronautical engineers in terms of searching for sites where adaptability and robustness can be increased.
(1) What is the current range of possible designs being considered for global financial regulation?
(2) What are the costs of switching between possible regulatory designs should the external environment place new or different demands on the system?
(3) What relevant decision points and chance nodes can be used to develop a rough model of how business cycles, external shocks and behaviour shifts might impact relevant aspects of the global and regional economies?
(4) When considering a broad range of plausible, dynamic scenarios for other exogenous forces the affecting global financial system (and assuming estimated development, implementation, enforcement, operating costs and switching costs for policy designs), what are the most efficient paths through the model?
(5) How does path efficiency change when switching cost assumptions are lowered (i.e. adaptability is increased), and therefore where are the highest points of leverage for increasing adaptability?
(6) Consider returning to (1) to redesign options with increased flexibility and lower switching costs for appropriate modules.
I think this is an area worth investigating more – both in terms of considering the general costs of policy lock-in with regard to financial regulation, and whether models such as TDN might be useful in designing complex policy instruments, rather than “simply” space-craft.
As a final note, I wanted to highlight a couple of imporant points in Silver and de Weck’s analysis. First, the need for a feedback loop to the designs themselves. One aspect that decision-making tends to leave behind is the creative production of the options between which a decision-maker needs to choose. Option-creation (also known as problem-solving) is a critical part of decision-making, and we should spend more time on this. Second, the creation of relevant operating scenarios is no easy task, as it is in this step that the uncertainty inherent in the future is partially captured by the model. These need to be challenging, plausible and anticipate forces that might not seem important today. From a policy perspective, I would charge scenario creation to a different group than the initial designers (although the latter presumably need to anticipate a range of certain base scenarios in order to develop relevant options), and assign a different group again to conduct the analysis and modelling (to reduce the tendency to favour a certain design, or to adapt to the scenarios influence the outcome).
Oh, and if you are interested by any chance in scenarios on the future of the global financial system, here are four I prepared earlier.