Monday, February 17, 2020

The Value of Formalization

Suppose someone hires us to solve a problem that is currently being solved by trial and error, rules of thumb, or by professional tradition. We are supposed to solve this using math, or programming, or something similar. Or as we say between you and me: using formal methods. Here is a naïve description of where the value comes in to the problematic system when using a formal approach:

First, we take a description of the problem and turn it into an equation, or whatever. Then we calculate the logical consequence of the equation and Oh My God: look at the Result! Now we know; we put this piece of information back into our process helps us save money. 

Today, the "equation" part is almost always realized as software. But the thinking still applies: the value is expected to come at the end. Here is a more realistic version of what happens:

First, we take a description of the problem and turn it into an equation, but wait... this description contains a lot of holes. It's not that the computer has a hard time crunching the numbers, it's that we can't even tell it what to do in the first place. The way the problem is currently understood makes no sense, logically. Parts of the process that are thought to be clearly prescribed by rules actually require a lot of human judgement and discretion to work, and this is a large variational factor in practice. Before we can even begin to optimize this in software, we need to gain a more exact understanding of the process than anyone has had before. 

In this version, the value starts coming in at the beginning, and it's a different kind of value. A better understanding of one's process is not always actionable. On the other hand, it can be actionable in very unexpected and profitable ways. Knowledge has that property. Currently, knowledge has to pass through a person's brain to have a chance at being used at its full value. So when given the task of formalization, take the chance to discover something new about a familiar process. It's often not an option.

No comments:

Post a Comment