When it comes to sharing knowledge with non-experts or out-of-domain people, researchers often fail to find a common language, which would be the basis for fruitful discussions and transfer of expertise. It indeed is a difficult balancing act, trying to abstract complex concepts, without losing scientific accuracy. Therefore, for the most part, one out of two strategies is chosen when trying to introduce non-experts to a topic.
One way is to stay at a very high level, only explaining the main theory, maybe some applications. Like that, the number of concepts that need to be explained is limited, and the content can almost be understood by anyone, however, the other person will most likely not get very far with this information, apart from being able to have an overall idea of what a specific methodology is about. Also, this approach often dangerously contributes to the Dunning–Kruger effect, where people with relatively little experience in a field drastically overestimate their understanding.
On the other hand, more in-depth introductions mostly provide a button-up explanation of all essential concepts, which is excellent if you want to get into the field. However, business people, for example, do not need to understand all essential concepts, and will most likely not be very happy with an “If I could just explain to you the mathematical concepts behind it, the method will become obvious” approach. They often want to have cooking recipes, short and easy with all the essential stuff in it. But as we all know, such predefined rules are very often inaccurate, as the answer to most of the “what to do if ...” questions should actually be “it depends …”.
With these approaches, you are somehow stuck between giving out too little and the wrong information while we know how crucial it is to find a common understanding in order to enable efficient communication, which is the key to bringing all these exciting technologies into the real world.
For this reason, I am starting a blog post series called ML ED: Machine Learning, Easily Digestible, where I try to provide an overview of topics from machine learning, knowledge graphs, and NLP in a lightweight way, without lacking the essential details needed to comprehend the capability of the technology. The target group of this series isn’t necessarily someone who wants to apply or research in this direction, but those who have little to no background and want to understand the potentials and shortcomings of different technologies. And who knows, maybe I can even infect someone with the excitement for machine learning and data science :)