Should we take small steps when innovating?
Although the problems of the world today seem to stem from an inability to build, the 20th century is also littered with the results of grand, utopian high-modernist experiments gone wrong. Therefore, it is often argued that innovation and changes should be small, slow and reversible. In his book ‘Seeing Like A State’ which documents the failures of high-modernism, James Scott makes this argument:
Take small steps. In an experimental approach to social change, presume that we cannot know the consequences of our interventions in advance. Given this postulate of ignorance, prefer wherever possible to take a small step, stand back, observe, and then plan the next small move.
Conservatives too often make this argument. For example, Michael Oakeshott argues that a conservative prefers “small and limited innovations to large and indefinite” and “a slow rather than a rapid pace”1 of change. Oakeshott also argues that innovation that is organic and “resembles growth” is preferable to that which is externally imposed.
Scott also argues that interventions should be reversible and that we should:
Prefer interventions that can easily be undone if they turn out to be mistakes. Irreversible interventions have irreversible consequences. Interventions into ecosystems require particular care in this respect, given our great ignorance about how they interact.
In my opinion, this preference for small, incremental innovation is misguided. This approach works for exploitation, climbing up hills and reaching optima in the known domain, but it does not enable transformative change. This is also why the intuitive, conservative expertise of ‘metis’ is not enough. Byrne Hobart makes the same point that “metis is a hill-climbing algorithm. If it’s based on experience rather than theory, it’s limited by experience.” Henry Farrel has also argued the same in a review of Scott’s book:
Italian firms in small-firm districts are excellent at gradual innovation and refinement of knowledge – in part because of their reliance on metis. They are not so good at producing profound, industry-changing forms of innovation.
Like metis, Oakeshott’s conservatism also restricts us to incremental progress. However, for any complex system to be able to generate transformative innovation, it must be able to leap into the dark, even at the risk of failure. But this does not mean that the system as a whole needs to take on such risks. It is the components of the system that must be allowed to explore and implement risky, transformative changes. Macro-resilience and macro-evolvability arise not from the system as a whole taking small steps but from the diverse micro-components taking big steps and being allowed to fail, i.e. macro-resilience requires micro-fragility.
What the great high-modernist experiments of the twentieth century (e.g. the Soviet Union, had in common was not so much their planned and artificial nature but their wholly unnecessary scale that risked systemic failure rather than component-level failure.
Even Oakeshott’s preference for “natural” growth is misguided. High-modernism did not fail because it was legible and codified or because it simplified the domain - almost all human progress since the birth of agriculture has involved the simplification of the world to better control it. Human civilisation is artificial, and this artificiality is a feature, not a bug. High-modernism failed because it codified and simplified the entire macro-system in one fell swoop rather than doing the same at the component level (e.g. at the level of the firm in economies, at the level of the genotype in biological systems). The goal is not to avoid simplification and legibility but to ensure that legible, fragile micro-components are allowed to fail and can be outcompeted and made redundant when they are past their sell-by date.
Even the preference for reversible steps is not realistic. Nothing is ever truly reversible in complex systems. Complex systems need to be adaptable, not reversible. The system needs to maintain the ability to move forward past the inevitable errors and failures rather than pretending to be able to move back to an imagined utopian past. In fact, systems that only possess the ability to move in small steps rarely survive a period of severe stress. On the other hand, resilient systems in the real world rely on system features like degeneracy to enable systemic reconfiguration and adaptation whilst maintaining minimal levels of excess resources.
Michael Oakeshott in his essay ‘On Being Conservative’