Craving Connection: Finding Meaning in Models, Frameworks and People

Can we connect?

We crave connection. Everywhere we turn, we want to be tuned in, turned on, plugged into the zeitgeist of the moment. We especially crave connections with people. But people can be hard work. As you travel through your professional life you begin to see common characteristics. Some (not all) of them are:

All tip no iceberg: They see the forest, but can't see the trees. They have knowledge a mile wide but an inch deep. They know a little about a lot of things. Fluff without the stuff. The conversational equivalent playing wack-a-mole. You just can't pin them down to anything.

All iceberg no tip: You've known them too. They drown you in detail. They explain at great length every intricacy until you're drowning in sea of information unable to make sense of it. You wanted the cake - they gave you the recipe.

Which should I be?

The answer is - somewhere in the middle:

Aim to be both but with a difference. Be a mile deep and plumb the depths of your area of expertise. In adjacent knowledge domains be a mile wide and go a mile deep but with shortcuts. What shortcuts?

Frameworks and Heuristics

Use frameworks and heuristics so you have a working frame to navigate new territory when required. These abstractions are the shortcuts. They're not 100% accurate but they're useful enough.

Who were the best leader or manager you worked with? I bet a common trait was they were across what you did without micro-managing. When you had a problem, they either had a solution or pointed you in the right direction. Chances are they had lots of frameworks to work from.

Their frameworks allowed them to illuminate a way forward. Intellectually it felt like they were holding a flashlight helping you navigate new terrain. They were also honest enough to tell you when they didn't know so you both didn't fall of the knowledge cliff.

How does this relate to modelling?

Modelling is also a form of abstraction. The world is complex place filled with countless variable and many unpredictable elements. Modelling requires a distillation of the many down to the smallest number predictive independent variables.

Algorithm , models, frameworks blah, blah..

Stay with me here. Let's use a simple equation we all know from high school maths. The generalised form (algorithm) of a straight line, y=mx+c. The algorithm is made flesh with data. y=mx+c becomes y=3x+1 with data points. We have taken the data and fit it into the frame (algorithm) that we already know. The co-efficient (m changed to 3) and the y-intercept (c changed to 1) are different but we understand what it looks like. We've gone from something representing all straight lines to a representation of this particular straight line with the addition of data. The addition of a sample of reality. It's a simple but powerful concept. We have made a connection. And this is just one algorithm.

From the descriptive to the predictive

So as modeller we perform the appropriate descriptive analytical work, to get from algorithm to model. A frame to aid in sense-making.

But only for a time. The world changes, the environment changes. Even the decisions we make using the model can change the very environment we are trying to predict. Think stock market prices after a major announcement. High frequency algorithms jump in and trade the price arbitrage. The trade is profitable until it's not. Back to new training data…

How are we to make sense of this?

Ultimately we are trying to make sense of the world around us. We are trying to connect or connect the dots.

We connect with people because, they have certain value systems, history and characters that are analogous to how we see the world.

Making a connection is recognizing something. Something in the underlying algorithms and models. Recognizing our model of the world is shared by others.