Knowledge Compression

How to teach skills effectively at scale?

Taking Angela Jiang's thoughts on knowledge compression and adding my perspective. Thanks to Manas Saloi for helping me discover this article.

To efficiently share knowledge at scale, it must be compressed. Information loss is a side effect of compression. Let’s take an example of an expert writer teaching the art of writing through a course. Information loss could happen when the signal is complex, an encoding or decoding error.

Loss happens when:

• The subject matter is tough. It isn't easy to convey the complete nuance of writing. (Tricky signal)

• The expert uses a lot of jargon and unable to express the thoughts clearly. (Faulty encoding)

• Learner did not understand the subject matter clearly. (Faulty decoding)

When is the teacher successful? The learner gets better at writing after the course than before it. It is a relative metric, and therefore, a pre-course assessment and post-course assessment could help in understanding the success/failure of a course. The completion rate is good, but if we can quantify a learner's success, that is the true metric.

The right form factor of the course could also be important. Is it a book, a self-paced video or a live course? The form factor, too, could bring about varying degrees of information loss.

Different learners can have different takeaways from the same course. Identifying the right learner persona who can benefit from the course most.

The greatest loss can be seen in our traditional education system. The writing instructor may not be a professional writer. So they might be unable to convey the practical insights on being in the industry.

With a good feedback mechanism, we can control loss to a great extent. With regular feedback on writing, the craft improves consistently. Learn, Apply, Feedback - designing these loops well will lead to lossless knowledge compression, allowing scale dissemination.