Thoughts and stuff...

The Limits and Potential of Using ChatGPT for Understanding Zettelkasten: A Personal Experiment

Introduction

A major system I use in my life is the Zettelkasten note-taking system (don't worry, I'll explain what this is in a second). I use it to guide my thinking and build a more holistic picture of the world. It helps me uncover new insights and connect ideas that I never thought I'd be able to connect. Needless to say, it plays an important part of my life.

Despite my Zettelkasten serving me pretty well, I was feeling it was a bit disorganized this week. I have quite a bit of notes that touch on various topics in the field of humanities. I have notes on philosophy, sociology, psychology and human behavior, and knowledge acquisition. The problem that I was having was in thematically grouping these notes together in a way that made sense. My notes were scattered, and I needed to impose some order on this chaos. Organizing these diverse notes thematically is key for me to maximize the system's benefits, especially given my broad interests.

Given the buzz around ChatGPT, specifically in its ability to put words to general ideas and patterns, I thought it a perfect solution to my problem! I would just put all of my notes into ChatGPT and get the perfect solution – all of my problems would be solved! I expected ChatGPT not just to organize my notes but to revolutionize how I think about these complex topics.

In this post, I am going to do a deep dive into this personal experiment I ran with Zettelkasten and ChatGPT. I wanted to see how it could help me revamp this major system in my life. I'm going to give a brief introduction into Zettelkasten itself, discuss the experiment, the results, where I was impressed, and where ChatGPT really fell short (Hint: it was a big disappointment for me).

Buckle up!

What is Zettelkasten and Folgezettel?

Zettelkasten is a German term that translates to “slip box”, and it's a unique note-taking system popularized by a sociologist named Niklas Luhmann. In this system, individual notes or “zettels” are written on index cards or digital equivalents. These notes are not just stashed away but are linked to one another in a non-linear fashion, which enables you to create a web of interconnected ideas. This helps in forming new insights and provides a structured way to delve into complex topics over time (Source: Ahrens, Sönke. “How to Take Smart Notes”).

Folgezettel, another German term meaning “follow-up slip,” further amplifies the utility of the Zettelkasten system. The idea is to create follow-up notes that build on or extend previous notes, thereby developing a chain or sequence. This approach promotes an evolving understanding of topics and fosters a deeper, more nuanced perspective. It essentially turns your collection of zettels into a self-contained learning environment where one idea naturally leads to another (Source: Luhmann, Niklas. “Communicating with Slip Boxes”).

In practice, Folgezettel employs a numerical or alphanumerical system to build upon previous notes. Let's say you start with a note titled “1.1” about, for example, the concept of “Mindfulness.” As you read more or ponder further, you might have additional insights that expand upon or refine your understanding of mindfulness.

To capture these new insights, you'd create a follow-up note and title it “1.1a.” This note could focus on a specific aspect of mindfulness, like “Mindfulness in Daily Activities.”

Then, perhaps you decide to delve into another major topic, titling this new zettel “2.1.” This could be about “Cognitive Biases.” If, along the way, you find a specific cognitive bias that you want to explore in-depth, a follow-up note “2.1a” could focus on that, say “Confirmation Bias.”

This numerical system doesn't just keep things neat; it makes it easy to trace the evolution of your ideas. You know that note “1.1a” is directly related to “1.1” and expands upon it, while “2.1” and “2.1a” form another thematic cluster. The beauty of this is that you can go as deep as you like into subtopics, without losing sight of the big picture.

With this in mind, let's delve into my experiment, which involved using my real notes from my Zettelkasten and a conversation with ChatGPT.

The Experiment: A Long, Deep Conversation

My notes, using the Folgezettel system described above, are organized from 1.x all the way to 4.x. I started off my experiment by copying and pasting 20 notes into ChatGPT, using notes from each section to give ChatGPT a broad view of my Zettelkasten.

After pasting a handful of notes from my 1.x section, I asked ChatGPT to come up with some categories or themes it detected. I repeated this prompt with notes in my remaining sections (2.x, 3.x, and 4.x). Admittedly, I was pretty impressed with the results here! ChatGPT had successfully put words to some of the ambiguous themes that I couldn't properly articulate.

To illustrate this a bit more, below is a list of note titles from my 1.x section:

Here's a snippet of ChatGPTs response to my notes in the 1.x section...

Based on the notes you've shared from the 1.x section, it appears that there are several overarching themes and categories:

Themes:

  1. Mental States: This theme includes your discussion on mental solitude, its connection to anxiety, and the physical states like FOMO. Notes like “Mental solitude can exist independent of physical solitude” and “Less Screen time likely leads to lower levels of anxiety” fall under this category.

  2. Information Management: Notes such as “Fear of missing out (FOMO) contributes to digital hoarding of information” and “Intuition is a powerful filter for capturing information” focus on how we manage, filter, and consume information.

  3. Knowledge and Learning: “Knowledge is information with personal enrichment” and related notes explore what constitutes knowledge and how one can transition from gathering information to attaining knowledge.

  4. Writing and Thinking: “Writing allows one to be the observer of their thoughts” and “Writing is a great medium for thinking” concentrate on the role of writing in cognitive processes.

Categories for Future Exploration:

  1. Screen Time and Well-being: You could explore how screen time affects not just mental solitude and anxiety but also other aspects of well-being like sleep, productivity, etc.

  2. Emotional Triggers in Information Gathering: Considering you mentioned intuition and fear as triggers, you could further investigate other emotional or psychological triggers that affect information consumption.

  3. Methods for Knowledge Internalization: Given your interest in turning information into knowledge, you might explore various methods to internalize or enrich raw data.

  4. Cognitive Benefits of Writing: Since you’ve mentioned that writing is a good medium for thinking, you could delve into how writing can be used as a tool for problem-solving, creativity, etc.

As you can see, ChatGPT came up with 4 potential themes that these notes touch on. I resonated with a few of these themes, but couldn't definitively pick a theme to encompass all of my 1.x section. However, that's not a shortcoming of ChatGPT. That's likely a problem with how I've organized my notes.

The results I got for sections 2, 3, and 4 had me feeling similarly. I agreed with a lot of the suggested themes presented by ChatGPT, but I couldn't definitively pick any single theme to encompass all of the notes in a particular section. Again, not a shortcoming of ChatGPT.

After getting ChatGPT's input on all of my notes, I asked it how it a few more interesting questions, such as:

This is where ChatGPT really delivered! By taking the entire context (notes from different parts of my Zettelkasten), ChatGPT was able to put concrete words and themes to otherwise scattered and chaotic notes. It actually helped me realize that I can condense my 4 sections of notes into 3 sections of notes while preserving the original spirit of the notes.

This is important because it unblocked me and opened the door to meaningful expansion of my Zettelkasten.

At this point, I was feeling ambitious. ChatGPT had done such a good job with what I threw at it, that I decided to give it one final task: take our entire conversation and turn it into a blog post.

The Final Test: A Blog Post

The set of prompts I used to generate the blog post were:

For those interested in the actual blog post, I've copied and pasted it into a separate thread with ChatGPT here. Feel free to take a look!

Analysis: Why It Fell Short

I'll start off by saying that I was pretty impressed with ChatGPT's outline and title. It seemed to capture the essence of our conversation pretty well. It broke down the outline fairly logically, starting off with the problem I was having, how it helped, and focused specifically on the mechanism of action (Folgezettel).

Where I was left extremely disappointed was that it didn't sound like me. It lacked my style, my tone, and my self-perceived charisma.

I will concede a lot here though. I have to take a step back and ask: is the problem of ChatGPT not sounding like me actually a problem? It might not be. From the perspective of purely conveying information, does voice matter? If ChatGPT can take a bunch of information, unearth really incredible insights, and then summarize the entire process in a logical and clear manner, does it matter that my voice is entirely removed from the output?

For me, it felt icky, and that feeling is entirely subjective based on my values. I value my voice coming through my writing at the expense of logical consistency and clarity. Is that ego talking? Maybe. Is it selfish? Possibly. However, these are the sorts of philosophical questions I think we will be grappling with as ChatGPT becomes more prevalent.

Takeaways and Future Prospects

I learned a lot from this short experiment. The biggest lesson I learned is that ChatGPT is amazing at kick starting the thinking and writing process. Thinking from a blank slate is hard. Writing from a blank slate is hard. Doing anything from a blank slate is hard.

ChatGPT was able to take my thoughts (as written in my Zettelkasten, and organized by my Folgezettel) and unearth patterns, allowing me to continue making meaningful progress in my work. It was then able to generate a solid outline for a blog post. If I didn't have ChatGPT, I'd be working from a more empty canvas.

I'm going to continue using ChatGPT to facilitate thinking and writing, but cautiously. I'm also going to continue to grapple with these harder, more ethical questions around ChatGPT. Lastly, I am going to recommend everyone try it out. It's pretty clear that it is here to stay, and people should likely take this seriously.

Conclusion

To wrap up: I started off this experiment from a place of curiosity. I wanted to see how this new tool, taking the world by storm and threatening to change the world, could help me out in a small, concrete way. I had high expectations — and it exceeded most — while falling short on a few.

I ended this experiment with more questions than answers, though. While the benefits seem too promising to ignore, I feel like I need to find my footing within the ethical debate of AI before continuing to seriously use ChatGPT. Or maybe I don't have to! Maybe I'll ask ChatGPT what I should do.

Thanks for reading!