All models are wrong. Some are useful.
When I first came across that quote, I will admit I think I missed a large part of the lesson. All models are wrong. Even the useful ones.
When choosing a model, the right fit has nothing to do with true and false – but simply if it is useful or not. A model exists to simplify. To illustrate relationships and allow predictions. This is done, primarily, by editing. By labeling a few key points as important and ignoring the rest. Take a map, for example: A very simply model that we have all interacted with. A perfectly detailed map that preserved all information at a 1:1 scale offers no advantage. It is death by fidelity. An absurd example, no doubt – but illustrative. The utility of a map is dictated by what few pieces of information are represented and at what scale. A model exists for a purpose, for an audience and a moment. A road map of your country would differ from a bike map of your city would differ from a child’s drawing of where treasure is buried in your yard. Each iteration has a person, a place, and a goal in mind. Each is wrong in its own ways, and through those omissions – the information is given shape and purpose.
If the previous idea is accurate, then each model has a sort of world-view. Crafted by a person, as a means towards an objective. Each model contains the bias of its creator. The simple judgement of what is relevant and irrelevant to the goal, what should and shouldn’t be included in the model shapes all the information that can be extrapolated. What I missed in my first reading of that quote is that a useful model is just as wrong as a useless one. As we progress and the scope of our endeavor matures, the information that meaningful towards our goals change – and so must our models. The bias inherent in the model is only really a problem if those assumptions go unaccounted for. Like the perfect map, we cannot afford to pay attention to everything. We need to make judgements, and absent experience: we borrow the judgment of others. The trick is to not do this blindly. Those biases function almost as an organizing principle. Each model tagged in a way with its objective and starting parameters. Because nothing is useful in a vacuum. Current goals and past experience create an intersection of language and scope. The model is a tool – shaped with purpose, to be wielded with both skill and the knowledge of its limitations.
And to be put down when its job is done.
As interesting as this all may be, we started on the topic of utility. If we put in the effort to understand the issues with models, with all models, we can begin to use them more effectively to make progress in our own lives. When learning a new skill, we lack the knowledge required to edit out what is important, to simplify the task and create a model – so we borrow. When I started jiujitsu, I didn’t have the experience to understand the motivations behind my opponents movements. I didn’t know much, but I figured my opponent didn’t want anything good for me, so I would resist every movement they would make with equal intensity. Fear and ignorance made me predictable and easy to manipulate. The assumption about my opponents intentions was correct, but I needed a filter. I needed to find a better model.
In a parallel to the Dunning–Kruger effect, there seems to be a gap between being able to recognize that you need a teacher to help, and the ability to gauge if someone can be a useful teacher for you. We are often in such a rush to make progress that we forget to examine what progress even looks like. When I am learning a new skill, before I think about trying to “do well” my focus is now to learn what resources are even required to do well. What traits seem to be rewarded the most? What are all the “good” people consistently good at? If I am lucky, I can recognize greatness when I see it – but I often can’t pinpoint why it is great. So I try and understand the environment. The prerequisites that greatness arises from. To understand how someone sees and interacts with the task, as opposed to just how they ultimately deal with it. When I realize I need a teacher, I try and engage in the same process. The goal is to first figure out how to identify what makes a useful teacher for me at this moment. The best teacher isn’t always the best performer, and the best teacher for me right now isn’t always the best teacher for me next year. I am looking for someone to guide me, to listen and weigh in. To filter. More than that: I am looking for someone to help me build my own models. Someone who understands that their skillset is also a world view, someone who won’t just show me a map but teach me how and why it was created.
One of the first hurdles we encounter when trying to learn something new is finding out what we can ignore. One of the most difficult hurdles we encounter when trying to refine a skill we have practiced for years is seeing the impact of something we once saw as trivial. Reminding myself that these models are inherently “wrong” as a design parameter has helped me switch between them more readily. To keep the emphasis on utility as opposed to using them to define myself. It has also helped me identify why I value teaching so much. A new student, one who is eager and interested will make me examine and reexamine my own models. Will allow me to see the problem from a new light and interrogate things I have taken as gospel. At this point in my growth – a good student is a valuable teacher, and often harder to find.
Ultimately, we are all teachers.
We are all students.
We are all wrong.
And if we all try very, very hard…
We can all be useful.
