Learn from 3.8 billion Years of Experimentation Experience

TL;DR: What nature can teach us about being good experimenters

Who designed your eyes to dynamically adjust to ambient lighting? Who decided how the folds of your ears should bend to optimally capture sound? Who architected the size, shape, and number of teeth to allow for ripping, tearing, chewing, and grinding food?

If you believe in Darwin’s theory of evolution, then your answer to all of these questions would be “No one did.” Instead, you’d believe the results above are the product of around 3.8 billion years of experimentation through evolution and natural selection.

In this blog post, I’ll show examples of how nature may be seen as the oldest experimenter and what we can learn from the billion years of experimentation experience.

Nature: The world’s oldest experimenter?

How does nature exemplify experimentation?

Check out this simple A/B test happening right now in people around you. Some people have lobed ears (variation A) and some people have lobeless ears (variation B) dictated by genetic variation [1]. Do you think there’s a winning variation?

But A/B tests aren’t all that’s happening in nature. In genetics classes, students often study the story of the Punnett square, which is a way of describing and predicting all combinations of traits [2]. One historical example is the combinations of traits that Gregor Mendel found in his pea plants [3]. This is nature’s multivariate test, the way of generating and testing all combinations and interaction effects:

With so more variations, the single winning trait may be less clear. What if the winner depends on the environment? Look to Darwin’s finches, which show that certain beaks are ideal for different food sources provided by the environment. Shorter sturdier beaks are better for seeds and nuts, whereas sharper beaks are better for insects [4].

Nature’s process of creating different species which are tuned to their environments is analogous to the concept of personalization. Some environments will have different better variations:

What happens when you combine all of the above experiment types and continuously iterate? You get continuous optimization, a process of using experimentation to constantly iterate towards favorable improvement:

What can we learn from nature’s experimentation?

Although you may be excited about mimicking natural selection’s methods of experimentation, you should learn from both the benefits and pitfalls before following Darwin’s theories blindly. Here are 6 tips:

1. Choose variations inspired by an educated hypothesis

One learning is that if you use random variations like nature does, then the process of optimization can take an extremely long time. Much of nature’s major milestones towards meaningful results are measured on the order of millions of years:

Instead of choosing random variations, you’re more likely to achieve a successful result faster if your variations are inspired by an educated hypothesis rather than trying everything under the sun.

2. The environment strongly influences what is optimal

Similar to Darwin’s finches, it’s really important to understand the environment in which your experiment is running to get to an optimal variation.

The below three animals are shockingly similar in size, shape, and features even though they all originated from very different paths. Sharks came from bony fish which lived in the sea. However, Ichthyosaurs and dolphins evolved from land reptiles and land mammals respectively [5].

What drove these very different animals into such close similarity? Their environment. The winning variation will strongly be influenced by the environment and that is why personalization can be so important and so powerful.

Catering variations to the surrounding environment is key to finding the true ideal variation for a given environment.

3. Some technical debt is okay!

Have you ever met someone who had to have their appendix or wisdom teeth removed, but were fine living without them? Have you ever wondered why we have a tailbone but no tail? These vestigial organs are the remnants of previous variations that are no longer necessary [6].

After doing a lot of experimentation, you may end up with remnants of old winners or outdated experiments in the form of technical debt. However, if things are working, it’s not that detrimental to have some technical debt lying around. Some technical debt is okay, which is especially true if it’s complicated to take that technical debt out.

4. A few simple metrics make declaring winners easy

What are the metrics that nature uses for experimentation? Animals either survive and reproduce or they die and they don’t.

If you can reduce your metrics down to something as simple and as meaningful as life or death, then it’s easy to optimize for winners.

5. Use incremental as well as dramatic changes

Nature often incrementally changes variations from generation to generation. But that only gets nature so far. Beavers cut down trees with massive front teeth evolved from rodents. Although the task of cutting down trees may have a more efficient tool, nature may not arrive at that conclusion from taking incremental steps.

Incremental changes are useful for optimizing towards a local maximum, but finding a global maximum sometimes requires a large shift or major change.

6. Experimentation is most powerful as a continuous, always on, optimization process

It can be hard to believe that all the incredible species around us are the result of a simple ongoing process of experimentation. However, it’s important to realize that at no point has natural life stopped experimenting.

Competing variations are constantly being created to continuously adapt to ever changing environments and conditions. Continuous optimization is what has lead the natural world to where it is today.

One of the most important takeaways from the natural world is that even 3.8 billion years of experimenting is not enough.

Always keep experimenting!