Designing for Reading Comprehension

1. Let us make the distinction between comprehending and knowing.
2. Knowing a topic refers to possessing information about it — being aware that it exists. There’s no requirement regarding the depth of understanding — its meaning or applicability.
3. Comprehending requires deep understanding of a topic — its meaning, implications, and surrounding aspects are well understood.
4. Comprehension is a much better goal when designing a reading system.
5. Just being superficially aware of a topic is not enough for most purposes. We can’t truly discuss such topics; we can only fake a discussion.
6. Systems designed for knowledge lack the ability to foster real thinking. Engineering courses, academic programs, and schools are prime examples of what happens when systems are designed for mere knowledge.
7. You end up with information, but it’s disconnected — dangling — with no roots in your existing conceptual base. Any gentle breeze can pull it away. Suddenly, we’re not even aware of it anymore.
8. I think about all the concepts I learned during school and engineering; they didn’t stick. At a foundational level, only the things I kept using and was more motivated about endured.
9. One might say that it’s not the job of universities or schools to teach comprehension of these topics. But then — whose job is it? Do we have to figure it out ourselves? We spend so much time in these places — really, a big chunk of our lives — and we’re left with blankness, or at best, a slightly less blank view of reality.
10. Comprehension is, therefore, key. All meaningful learning systems should have this as their goal.

How to foster comprehension in a learning system?

The limitations of the brain

1. We must recognize that we are designing systems for humans — humans with brains, and brains have limitations.
2. I’ll base my comparisons on personal experience as a human being learning how to learn. So take this as a critical limitation of the discussion.
3. Still, I claim that other brains behave similarly.
4. Brains have a limited “comprehension bus factor”: expecting too much comprehension at once often leads to frustration.
5. Brains learn only when we add a new concept on top of a well-established base. Adding many dangling concepts at once is, again, a major source of frustration.
6. Therefore, concepts should be heavily interconnected.
7. Simply regurgitating knowledge is boring. It creates the illusion of comprehension and benefits neither the speaker nor the listener.
8. Concepts should be discussed. Through discussion, you discover gaps in your understanding and test the strength of your conceptual roots — where they are strong and where they’re weak.
9. Brains are not naturally great at abstract thinking. It’s not that they can’t do it — on the contrary — but I argue that concrete instantiations of knowledge (via examples or mappings) are key to truly understanding something.

Putting all principles together

1. A learning system should be designed to maximize discussion of ideas.
2. Discussion is the primary process that generates comprehension.
3. Through discussion, you connect ideas.
4. But merely connecting ideas — as proposed in Active-Ideation — is not the ideal goal.
5. An idea might be linked to thousands of others and still be dangling — barely surviving in our memory.
6. This has happened with many concepts I often relate to others. But the connection is superficial — since neither concept was deeply understood. It becomes a shallow link and sometimes even detrimental, as we see links but can’t recall where they come from.
7. Nor is recall improvement — as proposed by Progressive-Summarization — an ideal goal.
8. I agree that notes should be designed to be recallable by your future self.
9. But this alone doesn’t serve as a good goal for a learning system.
10. It doesn’t differentiate between superficial knowledge and comprehension. Your brain will quickly learn how to game the system — giving you the illusion of comprehension because you’re “doing it right.”
11. You get stuck thinking, “The results will come if I keep doing this every day for years.” But you don’t see them — yet.
12. This falls into the Collector’s Fallacy: you’re conflating the collection of knowledge with comprehension. As we’ve seen, these are very different things.

A practical example: annotating books and papers

1. What’s the value of a highlight? “To mark the important parts of a text,” you say.
2. But why do that? “So I can remember the important parts.” But how does a highlight actually help with that?
3. Here is where the argument starts to fall apart. “If I read the highlight, I’ll remember the key part of the text and won’t need to reread the whole thing!” Wrong. Try it — or maybe you already have. Take a dense paper and highlight the key parts. Do this regularly. You’ll soon find yourself highlighting large blocks of text — because there are so many “key parts” and unknowns you need to revisit later.
4. Now return to the text a few weeks later. Did the highlights help you remember the key parts? I doubt it. It’s just a mess.
5. Highlights are not for remembering information! Not even close. Don’t be unfair to yourself — acknowledge the limitations of your brain.
6. Highlights are, first and foremost, visual cues. “Cues for what?” For marking that a discussion took place. When no discussion happened, no highlight — simple as that.
7. If you return to the text in a week and reread your discussion points instead, you’ll immediately be back where you left off. You’re up to date on the relationship between the text and your more deeply-rooted concepts.
8. If no discussion happened, you didn’t understand the text. That’s the key point.
9. It’s not your job to maintain a comprehensive version of the author’s ideas in your knowledge base. That’s the job of Wikipedia, LLMs, or textbooks.
10. Your job is to bring your own perspective into the equation. This aligns with Active-Ideation, but the focus is not just on relationships — it’s on discussions.

Discussion-is-the-Main-Goal!