Data beats intuition

…or, this is why we test

Some things don’t require testing, correct?

The answers are obvious. For example:

  • The larger a person’s bonus, the better the person’s performance will be.
  • A delicious cake-mix requiring only water to be added will sell better than one that requires additional steps and ingredients.
  • Giving customers 24 choices of jams will lead to higher sales than providing only 6 choices.

If you have read Dan Ariely’s “The Upside of Irrationality” you know better. None of these seemingly obvious statements are true. How do we know?

Experiments and data.

Humans are quirky and unpredictable. How else can we explain Donald Trump’s presidential bid?

Domain experts fail to accurately predict reactions of markets, masses and contests. Why are we so stubbornly attached to our preconceived notions? Haven’t each of us been proven wrong enough times to realize that “the older we get, the less we know?”

This is not an argument for blissful ignorance, but rather a call for testing. We have the tools to capture data and measure the results. Adherence to untested conventional wisdom is a barrier to innovation.

“It can’t be done, because it never has been done before.”

The above is a flawed statement, but one that we constantly fail to challenge. Because we accept the “obvious”, we miss opportunities to understand the hidden truths of our world and humanity. Copernicus and Darwin dared question conventional wisdom and used scientific methods and data to support their controversial theories. Newton and Einstein observed nature and matter and built mathematical models that could explain the logic of how these invisible forces work. These geniuses needed enabling instrument such as microscopes and telescopes to run their experiments and capture the necessary data. The results are the foundations for calculus, biology and physics. These scientific methods have provided the building blocks needed for innovations in architecture, civil engineering, chemistry, technology and medicine.

Darwin made his discoveries mostly with observations of the naked eye, calipers and magnified glasses for measurements, and journals to record his results. We now have incredibly efficient and inexpensive computing devices to store and analyze data. We also have ways to convert analog signals like text, photos and sound into digital signals. Innovation is no longer limited to those who have the stamina and meticulousness to capture the necessary data. Now it is easy for a researcher to conceive an experiment and review the corresponding data sets.

From recording to modeling

In order to validate a hypothesis, a large enough data set is required. What to do when the events being tested do not have enough real-world results? Back to Darwin. He had a literal globe of animals, insects and plant data, as well as historical artifacts (fossils), to pull from. Enter computer simulations. Medical research is notoriously expensive, error-prone and time consuming. The hack that medical R&D labs have used to combat the cost of human trials is to experiment on animals first.

What if there was a computer model to simulate human trials? Southern Methodist University recently announced a successful protein model simulation. The research team virtually tested millions of available chemical compounds with the model to see which ones bonded with the protein. They identified the successful compounds and even were able to see where on the protein the compounds were likely to bond. Amazing! Using this brute force methodology, the research did not depend on human intuition. The promising compounds identified from the models may help defeat cancer.

Frailty, thy name is human intuition

To paraphrase from “Colors”:

There’s two scientists working in a lab. The younger one says to the older one: “Hey pop, let’s say we set up the lab, and test these promising compounds”. The older one says: “No son. Lets setup a computer simulation, and test them all”.

To be sure, building reliable simulations is not trivial. A weak simulator can create inaccurate data, in turn leading to dangerous conclusions. Even with rudimentary simulators for whether forecasts, space travel, self-driving cars, and medical research, the results are encouraging and improving rapidly. As we enhance simulators to better replicate real-world conditions, the speed and breadth, not to mention the precision of innovation will continue to accelerate. There is logic in our world, we just need more computational power to decipher the code.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.