The purpose of Recipes is to document the learning process itself as opposed to the outcome of the learning process. Oftentimes, when we learn a new topic, we do so in the context of some task, carry out that task, and very quickly forget how we learned it. This causes many issues, one of which is inefficiency in making better decisions about how to use that knowledge in other contexts (aka generalization).
In education psychology there is a concept called meta-cognition (we will spend some time on this later in this list). It essentially means the ability of humans to observe, analyze, and optimize their learning strategy as it comes to using it for carrying out tasks. In other words, meta-cognition is our ability to "learn how to learn".
We spend a lot of time in all learning settings on "what to learn", but rarely on "how to learn". One of the most common questions that i get from aspiring data scientists is "what do i need to learn to enter the field" while my answer is often that "the field is moving so fast that anything you learn would be probably old news by the time you finish learning it; therefore, the best thing you can do for yourself is to learn how to learn because then you can always keep up".
That's why we invented and introduced Recipes as a new document format!!!
Let's go deeper in how Recipes achieve this:
Given the explosion of information and resources on the internet, creating lists as a way of communicating the importance of some of the resources is a very commonplace thing to do. Some people take the approach of creating the most comprehensive lists which is awesome but in my opinion not really practically useful. A lot of us have lists like that in our bookmarks and we never really use them in any meaningful way. Another approach is to write blog posts and have a bunch of resources laid out here and there in it. I have done that, you have done that, many of us have. But the challenge is that it is very hard to generalize something like that because information is laid out in an unstructured way.
We believe that having a format that is explicitly a list, and is modular so that it can be broken up as blocks and crafted into other lists is the right way to tackle a lot of the issues that we are facing in learning new topics collaboratively.
Based on our experience of interacting with all the awesome lists people had created on the internet, and impracticality of them, we decided to enforce a strict length limit of 7 items for the lists. This ensures that the creator focuses on finding the most important resources for the topic at hand, and provides a much needed prioritization and importantly carefully set up order of the resources. In other words, we wanted the lists to tell a story through a step by step walkthrough of a set of resources.
The crafting process asks for a lot of information to ensure the user knows exactly what they're getting and how. The info are things like "what questions does this resource answer" or "which part of this resource was most useful to you" aka "what's your recommended way of using this resource?". Although these are pretty high level information, our experiments show that they already save a lot of time.
Another important aspect of Recipes is to ensure and remind the creator that the list has to be most suitable for someone who is just encountering the concept covered in the Recipes. This person can be the creator's future self encountering the topic again after a long time, or anyone else. Therefore, ensuring that the information provided follows a good and concrete narrative, and that it's verbose enough to actually save people's time is crucial.
One of the most important things that should be clear by now is that the haphazard ways we learn things is not going to cut it. This is especially true for advanced and fast moving topics like what we deal with in AI (or emerging tech in general). Therefore introducing a structure that makes sense and is easy to follow is crucial. With SL's we first capture the main concept at hand, and top 3 prerequisite concepts necessary to learn the main one. Note that we focus on up to 3 because, let's be honest, there are a few concepts that are the most important ones to learn for any main concept. Then we attach / associate the best possible resources for each concept. This way, we have a clear structure inspired by how knowledge works (an interdependent graph of concepts and their relationships), and quickly provide what we, the creators of Recipes, think are the best resources for each concept. The resources can range from theory, to implementation, to application.Or they can be in any format that is most suitable from the subjective viewpoint of the creator of the Recipe.