Problem: Bad Explanations

Giving a good explanation is deceptively difficult. When we give an explanation, the idea we are communicating is clear in our mind. Everything fits into place. It’s frustrating when the person we’re explaining to doesn’t understand. Why does this happen and how can we prevent it?

Cause: A Proliferation of Details

When teaching a concept, it’s difficult to know where to start. We explain lots of details but forget that, while we have a framework for combining all of the pieces, the listener does not. The listener is left floundering, unsure of how the details are related. Enter our savior, hierarchical abstractions.

Solution: Hierarchical Abstractions

We naturally decompose the world into hierarchical abstractions. This idea will be the basis for guidelines on giving an excellent explanation. First, a theoretical definition. This definition is too academic to be clear. Don’t feel frustrated if you can’t visualize the abstraction as we’ll go over an example next. The key piece is that this is already how you think.

In a hierarchical abstraction:

1. A concept is decomposed into layers of abstraction.
2. Each layer is composed of elements from the layer below.
3. The topmost abstraction is the concept itself.

A hierarchical abstraction is “good” when each layer can be understood without knowing the details of the layers below it.

An Example

Let’s use an apple as a clarifying example. 

Top Layer – The Concept

An apple is a solid, mostly round fruit, that is often red. The top layer of the abstraction is just the concept itself (an apple) along with a description of it’s properties.

Layer Two

Apples are made up of several components including skin, flesh, seeds, stem, and core. We can describe the properties of teach of these elements in a way that is parallel to the description of properties in the top layer. For example, the skin is smooth.

Layers Three + Four

Each item in layer two is made of molecules (layer 3) such as fiber and carbohydrates. Those are made of various atoms (layer 4) such as Carbon, Oxygen, and Hydrogen. Again, each of these components has a set of properties we can describe. We could go into deeper layers but I’m sure you get the idea.

Key Points

The example above illustrates a few key points that teach us something about giving excellent explanations.

1. We do not need to know the details of a lower layer (apples are made of atoms) in order to understand many things about an apple (an apple will roll because it is a round solid). Takeaway: It is possible to explain one layer without explaining every layer below it.
2. The choice of abstractions is somewhat arbitrary. Where does the skin of an apple end and the flesh begin?  Takeaway: Others may have different abstractions in mind when we are explaining something.
3. We can move fluidly between abstractions we have a fully understand. It is easy to move from apple to skin and back again. Takeaway: We can reference up or down the abstraction hierarchy so long as the abstraction we are referencing has already been explained.
4. Knowledge about each layer informs knowledge about other layers. We know that an apple is made up of oxidizing molecules which explains why cut apple’s turn brown. Takeaway: We can use knowledge from one layer to explain the properties of another.

Explanation Guidelines

A few guidelines are elicited from treating explanations as transferring our hierarchical abstraction to another person.

1. Start with a simple, shared, high-level abstraction. If you’re explaining an apple, talk about its properties, not its atoms.
2. Ensure your abstractions are shared. If you say apple skin and the other person thinks apple flesh you will struggle to make progress.
3. Use lower layers of abstraction sparingly. If you reference apple skin while explaining an apple and the listener doesn’t know what that is you have another concept to explain.
4. Contextualize lower layer abstractions by fitting them into the layers above. When you start explaining what apple flesh is, start by explaining where it is in the apple and the properties it lends to the apple.
5. Once multiple layers of the abstraction are shared, use lower layers to inform properties of higher layers. An apple browns because of oxidizing molecules.

Limitations

Despite the guidelines above, giving an excellent explanation may still prove difficult.

First, creating a good, non-leaky abstraction is hard. The example above was easy. We all know what an apple is and the molecule/atom abstractions were taught in school. Most explanations you give will be for concepts you don’t have a strong grasp on. You’ll need to spend some time thinking about how you have mentally modeled the system before you can decompose it for the listener.

Second, it’s tempting to dive into the details. Details are often the interesting part. A specific detail might be the reason you’re giving the explanation.  You might have just remembered a detail you don’t want to forget to explain. Whatever the case, make it easier for your listener, ensure higher levels of the abstraction are understood before jumping into the lower layers.

Finally, giving too many details is not the only reason for poor explanations. Maybe you don’t understand the topic that well or giving an overview of each layer will take more time than you have. There are other techniques, such as using metaphors and concrete examples, which could assist in your explanation.

Bonus

When someone is giving you a bad explanation, you can use the idea of hierarchical abstractions above to help them clarify their explanation. Ask questions until you understand the top level abstraction before letting them move on to explain details. Have the explainer show how each detail fits into the overarching system. Ask what properties the system inherits from the piece they are explaining.