Working with Imperfect Models (Part 1)

Lego HouseWhat does your prediction of an acquaintance’s reaction to a piece of news have in common with the weather forecast? They are both based on models of incomprehensibly complex systems; models that approximate, but do not fully capture the underlying reality, and can therefore prove very wrong.

I use the term “model” here to mean any of the condensed mental constructs we use to represent a far more complex reality. These include concepts, principles, visual models such as maps, mathematical representations, and representations of change such as workflows. More often than not, our models use several of these put together, to create a more complete representation of a particular system.

Human beings need models of reality to survive, of course. They allow us to explain current and past behavior, predict the future, and imagine alternative routes of action. And for the most part they work remarkably well: When an architect designs a building, for example, he can calculate the forces and tensions his building must withstand and can counteract them using his model of sound engineering.

But not all of our models are so reliable. Follow most nutritional or exercise theories with 100% dogmatic adherence in the same way that an architect must adhere to the science of building, and you are likely to harm your body. Trust the weather report too much and you’ll end up wet. Believe in your own self-image too much and you end up stunting growth and afraid to take risks.

It’s easy to conclude that the particular model is at fault: simply find another diet, find a better weather prediction algorithm, or change your conception of who you are to a more “correct” one, and you’ll be happy ever after. But many if not most models are inherently imperfect. Specifically, when dealing with systems of extreme complexity, any condensation of information is a net loss in accuracy and predictive ability. Yet, this condensation has to happen for us to be able to hold the model in our limited brain.

So what do we do? I find it useful to try to keep in mind that I am dealing with a model, not with reality itself. Once you introduce the concept of a “model” into your thinking, you can begin to interact with your models as models, and have a healthier relationship with them. Specifically, you can:

  • Keep Your Eyes on the Road – While the map is a useful model, the road is what really matters. Do not get so involved with the map that the road becomes invisible to you. If you are on a new diet, the “road” is what your body is telling you. If you are getting to know someone new, the “road” is what they actually say and do.
  • Identify Limits – Always keep in mind the limits of your models. Try to find the blind spots, the inaccuracies, the oversimplifications, and the edge cases the model doesn’t account for. If you’re an architect, for example, it’s good to ask yourself whether you’ve accounted for unusual strains such as an earthquake or a hurricane; or whether the forces in play change after reaching a certain height.
  • Expect to Adapt – Models are mental constructs, and as such easier to revise than the underlying reality. Yet most people spend their lives trying to make reality conform to their impossible models. Life is a process of revision and rewriting of your models based on new experiences and new skills. So nurture your models, learn from them, but never worship them!
  • Use the Baseline Rule – When you must rely on a model you know is not 100% accurate, such as the model for a healthy diet, or a training regime, try to determine a baseline that you would follow in absence of this model. This can be based on current or common behavior that has a known outcome. Once you’ve established the baseline, experiment in the space between that baseline and the new model. This will help protect you from the cases where the new model is imperfect and potentially harmful.

Remembering the inevitable gap between reality and our imperfect models of it is tough, especially when it comes to models that are emotionally charged. An inaccurate model, though, can yield frustration and suffering, often in direct relation to how blindly we follow it.

I’ve been fascinated with the idea of models, and the ways they change over time, and interact with our emotions in interesting ways. I’ll write more on that in Part 2 of this post. In the meantime, I hope you found these musings helpful! Would love to hear your thoughts, additions, or corrections.