Design vs Selection
How much randomness will you accept when solving a problem?
Last updated
How much randomness will you accept when solving a problem?
Last updated
When facing a problem, you must decide on the levels of randomness and uncertainty that you're willing to accept in the solution process. In some cases, there's very little randomness that can be tolerated because the consequences are simply too great. In others, higher levels of randomness are either tolerated or encouraged. This can be framed in terms of design versus evolutionary selection, two distinct approaches that are appropriate in specific types of domains.
Design, often the default human choice, represents a methodical, planned approach. It's similar to setting out on a journey with a detailed map and a clear destination. Algorithmic steps are pre-planned to address the problem at hand and reduce uncertainty before any effort is expended.
This top-down process, deeply ingrained in various aspects of our lives like education, work, and even daily activities like cooking, relies on predictability and control. It assumes a comprehensive understanding of the environment, where variables are known and outcomes can be anticipated. In design, the path is charted beforehand, and deviations are often seen as setbacks rather than opportunities.
In contrast, selection embodies the acceptance of uncertainty and flexibility. Deriving its name from the evolutionary concept of natural selection, this approach is about adapting and evolving in response to environmental pressures. It’s akin to navigating a wild, uncharted territory where the landscape itself is in constant flux.
With selection, there is no predetermined path, but rather a continuous process of adaptation, where what works is kept and what doesn't is discarded. This method aligns with the principles of genetic replication, where heredity, variation, and environmental pressures interact to select the traits that are most suited for survival and reproduction.
The selection approach is dynamic and inherently uncertain. It recognizes that the environment is not fully knowable, which is often the case when dealing with complex adaptive systems. In this context, solutions and strategies evolve organically.
They are not designed from the outset but emerge through a process of trial and error, experimentation, and adaptation. This method can be more resilient and robust in the face of unknown futures, as it does not rely on rigid plans but on the ability to adapt and change course as needed.
This approach is particularly effective in environments where variables are known and stable, allowing for algorithmic solutions. However, the major trade-off here is inflexibility. Rigid adherence to a pre-defined plan leaves little room for adaptation to unexpected changes or new information. In dynamic environments, this inflexibility can be a significant drawback, as it may lead to inefficiency or failure when conditions deviate from the initial assumptions.
Selection thrives on adaptability and resilience. It's suited for environments characterized by uncertainty and change, where the ability to evolve and respond to new challenges is key. This approach is less about following a predetermined path and more about navigating an evolving landscape through trial and error. The trade-off, however, is that you have to let go of your deterministic beliefs about outcomes. If nothing else, selection processes will remind you of how little you actually know and how impossible it is to predict the future.
Selection is more appropriate for businesses in fast-moving industries, such as those that thrive on the internet. You can design all you want, but it's impossible to know ahead of time what will stick and what won't. Even with deep industry knowledge, all you can do is make educated guesses, so your problem-solving process should be heavily focused on selection.
For example, I used to have a job running Facebook ads that involved creating dozens of different ad variations every day. Nobody knew exactly what would work ahead of time, although I often assumed I could spot a winning ad from a mile away. On some occasions I was correct, but it shocked me how often ads that I thought would be total duds ended up bringing in enormous revenues. This is selection in action: throw out tons of variations into the environment and see what you get back.
Another example you're hopefully familiar with is building your social network. First you go out into the world and interact with people who happen to be in your environment. You get to know them, and some of these people you develop relationships with. Those who you don't want to be friends with (or don't want to be friends with you) are removed from your pool of contacts, and those who you get along with stay.
Although my focus tends to be on problems that are complex and more geared towards selection-style processes, there are times when design is the better choice. For example, if your job is to build bridges, it makes sense to utilize a design process. Bridges operate on the laws physics, which are well-understood at that scale, and computers can now do all the hard calculations needed to ensure a bridge is safe to cross.
I personally would not want to fly in a plane, drive in a car, or walk over a bridge that wasn't designed by engineers who reduced the randomness of their systems as much as possible. That being said, chances are high you're using design to deal with problems that should be handled via selection.