Innovation and evolution

Innovation is a buzzword en vogue as of lately in the business and technological world: much material has been written on how to promote, foster and drive innovation. I would like to suggest some analogies between the innovation process and evolution as modeled by computational techniques such as genetic algorithms.

Innovation consists in the introduction of new methods, processes or technologies with the goal of improving some business metric (revenue, efficiency, etc.) Thus, the innovation process can be regarded as the optimization of a metric function across a space of possible solutions. Innovation can be incremental if the improvements are based on an already established solution, and disruptive when the solution tried completely departs from preexisting practices: it is generally assumed that incremental innovation ultimately brings marginally decreasing benefits, and that it is only through disruptive innovation that real quantum leaps can be made, even if it can take some adjusting time to catch up with established solutions. Disruptive innovation is usually associated with a radical change in the underlying technology (for instance, digital vs. chemical photography).

On the other hand, an evolutionary system implicitly optimizes some metric function (individual fitness in a given ecosystem) through extensive search across the genetic space. Genetic algorithms mimic this model for the more mundane task of looking for global maxima of numerical functions. The solution space is sampled by maintaining a pool of candidate solutions that are made to evolve according to:

• Selection of fitter solutions.
• Mating between solutions, using crossover (mixing of parental genes) and mutation (random alteration of the genetic material).
  

Allegedly, crossover is efficient at fine-tuning solutions when the population has reached an evolutionary niche, while mutation allows the population to escape from local maxima and explore new areas of the solution space: the combination of these two aspects yield a mechanism able to efficiently explore vast fitness landscapes of complex shape.

The similarities between these two worlds are obvious. Innovation also takes place in an a priori complex solution space, and different solutions thrive or perish according to its market fitness. Incremental innovation can be compared to local evolution as performed by crossover, while disruptive innovation is akin to the kind of far reaching displacements achieved primarily by mutation. If we are to take this analogy seriously, genetic algorithm techniques can provide us with some advices for improving the innovation process:

• Maintain a large pool of "beta" solutions to test for fitness.
• Quickly discard less promising solutions and replace with new ones, keeping a fast turn-around cycle. On the other hand, you must allow for a sufficiently large survival rate so that solutions can mature and some diversity is maintained.
  
• Incremental innovation is achieved by combining aspects of similar preexisting solutions (crossover).
• Disruptive innovation results from combining distant solutions and/or radically altering some of the aspects that define a candidate solution. Diversity is key to this process, which could explain why monolithic markets tend to hamper disruptive innovation.