How do you organize academic work in Obsidian? Here's my approach after 4 years

Knowledge management
,
1

Hi everyone,

I’ve been using Obsidian for academic work since 2020, and have settled on my own Zettelkasten-ish system that works for me. Recently, several non-Obsidian-using friends asked me to show them how I use Obsidian, and I decided to write a comprehensive guide documenting my approach. Since I learned a lot from lurking on these forums, I decided to host the guide online and post it here – hoping to help others & to get some feedback on my approach.

About the Guide

The guide covers how I use Obsidian for studying and research. Beyond just explaining what I do, I tried to also offer the principles according to which I do things, and the reasons for believing these to be good principles.

The guide is available here: https://obsidian-guide.neocities.org/guide

An exact table of contents is available on the page, but the basic flow of the guide is from theoretical approach (what are the goals of a vault, what valuable roles can a vault play, what are principles for vault-building & note-writing) to practical implementation (how to link, tag, use folders, implement note types etc.). I finish with opinionated takes on plugins, AI integration and collaboration.

I would be very interested to hear what you think and where you disagree (and how you do things differently). And I have a lot of more specific questions, including:

  • Would you choose a different set of principles? Do you think some are missing or superfluous?
  • What do you think about the vault roles?
  • What kinds of note types do you use? And (how) do you write reading notes?
  • What are your experiences with Obsidian AI integration, dashboards, and collaborative vaults?

The system I present in the guide works well for me, but I’d love to discuss & refine it. Thanks, and I hope the guide is useful to some of you!

12 Likes
2

This was a good read. I’m pretty new to obsidian (and taking notes in general; I’ve never been good at it), so I was kind of curious what an example of some of your notes look like. I’m trying to use obsidian for a lot of things but mainly for notes on subjects in my classes, and though I understand both your processing/collecting distinction and not categorizing your general knowledge by the class it’s from, I’m curious on what your strategy is on taking notes on, say, topics in a math class? Since much of it will of course be formulas copied straight from the class with annotations that make the most sense to me.

1 Like
3

Thanks, I’m glad you enjoyed it! I don’t have many math-y notes as math plays a small role in the philosophy (& other things) I focus on – so maybe someone with more experience can jump in to help. That being said: I would think that processing here just means adding annotations to the formulas. I assume the goal of these notes is to help you remember & understand & apply the formulas, and not to generate original ideas about them for papers or such – so processing will be more minimal.

An aside: I haven’t used this myself, but this plugin might be useful to you if you have lots of formulas in your vault: GitHub - RyotaUshio/obsidian-latex-theorem-equation-referencer: A powerful indexing & referencing system for theorems & equations in your Obsidian vault.

Here is a copy-paste of four of my more math-y notes (they tend to be more encyclopedic than other notes, more focused on allowing me to quickly understand than on my own ideas):

Bayes’ Theorem

Type:: #note
Status: #
Tags: #statistics #bayesian_epistemology #bayesian

Overview

[! Abstract] In a Nutshell
Bayes’ Theorem is about updating probabilities (/credences) based on new evidence. So: [[Prior Probability (Distribution)|priors]] + evidence = posterior probability distribution (/ updated credence).

Bayes theorem connects $P(H \mid E)$ (the direct probability of a hypothesis (or: “uncertain quantity”) given evidence) with $P(E \mid H)$ (the inverse probability of evidence given hypothesis).

  • “Though a mathematical triviality, Bayes’ Theorem is of great value in calculating conditional probabilities because inverse probabilities are typically both easier to ascertain and less subjective than direct probabilities.” SEP

“the Theorem’s central insight[:] that a hypothesis is confirmed by any body of data that its truth renders probable” SEP

[! info] Bayes’ Theorem (simple version):
$P(A \mid B) = \frac{P(B \mid A) \cdot P(A)}{P(B)}$

To understand the formula above, see [[Conditional Probability]] ($P(A \mid B)$) and [[Prior Probability (Distribution)]] ($P(A)$). Also: [[Probability]].

[! info] Bayes’ Theorem SEP
I add this because it uses credences (Cr), evidence (E) and hypothesis (H) and is thereby more in the vein of epistemological use of the theorem.

" Suppose that Cr is probabilistic and assigns nonzero credences to H and E, and that the Ratio Formula holds. Then we have:

$Cr(H \mid E) = \frac{Cr(E \mid H) \cdot Cr(H)}{Cr(E)}$

[…] This theorem is often useful for calculating credences that result from conditionalization on evidence E, which are represented on the left side of the formula."

Expanded formula (Bayes’ Theorem, 2nd form)
![[Screenshot 2022-08-26 at 14.04.33.png]]

Normally use the simpler version. If the denominator is unknown (here: P(A)), use the expanded formula to determine the denominator.

  • also used for medical [[Sensitivity and Specificity]]

Application in Epistemology

Key idea:

  • Probability as expressing the [[Credence]] in an event.
  • Bayesian epistemology studies norms governing credences, including how one’s credence should update in response to evidence

For more information, see [[Bayesian Epistemology]]

References

Conditional Probability

Type:: #note
Status: #
Tags: #statistics #bayesian_epistemology #probability #bayesian

Overview

[! abstract] In a Nutshell
$P(A \mid B)$ is a conditional [[probability]]. This means that it is the probability of event A occuring given that B is true. It is sometimes also called the posterior probability (of A given B).

The conditional probability can be calculated as follows:
$P(A \mid B) = \frac{P(A \land B)}{P(B)}$

($P(A \land B)$ is the joint probability of A and B)

Conditional probability of A given B can be transformed into likelihood of B given A:
$P(A \mid B) = L(B \mid A)$

[[Bayes’ Theorem]] is about [[Conditional Probability]] (see [[Bayesian Epistemology]]).

See [[Probabilistic Coherence Playground]] for examples of conditional probability calculations (need Numerals plugin).

References

Probability

Type:: #note
Status: #
Tags: #statistics #probability

Overview

[! Abstract] In a Nutshell
Notation: $P()$. Decimal representation, where 0 is minimum probability and 1 is maximum probability.

More information about probability is found at [[Statistics]].

See [[Bayesian Epistemology]] for a philosophical application of probability calculus.

Probability Calculus

Some rules of probability calculus:

  • $P(\neg A)$: [[Four Basic Rules of Descriptive Statistics#1 Complement Rule|Complement Rule]]
    • $P(\neg A) = 1-P(A)$
  • $P(A \lor B)$: [[Four Basic Rules of Descriptive Statistics#3 Addition Rule|Addition Rule]]
    • A&B mutually exclusive: $P(A \lor B) = P(A) + P(B)$
    • A&B not mutually exclusive: $P(A \lor B) = P(A) + P(B) - P(A \land B)$
  • $P(A \land B)$: [[Four Basic Rules of Descriptive Statistics#4 Multiplication Rule|Multiplication Rule]]
    • A&B independent: $P(A \land B) = P(A) \cdot P(B)$
    • A&B not independent: $P(A \land B) = P(A) \cdot P(B \mid A)$
    • Independence means that $P(A \mid B) = P(A)$ (see [[Statisticial Independence]])
  • $P(A \mid B)$: [[Conditional Probability]]
    • $P(A \mid B) = \frac{P(A \land B)}{P(B)}$
    • [[Bayes’ Theorem]] is an equivalent transformation of this formula.

An important question is: How do we determine the prior?

  • $P(A)$: [[Prior Probability (Distribution)]]
    • If we have n equally likely outcomes: [[Four Basic Rules of Descriptive Statistics#2 Rule for Equally Likely Outcomes]]
    • Can also determine it empirically.
      • According to the [[Miller’s Principle]], our [[Subjective Probability]] should match [[Objective Chance]] whenever possible.
    • According to [[Bayesian Epistemology#Objective Bayesians|objective bayesianism]], priors should be both coherent and free from bias. One way to encode this freedom from bias is the principle of indifference:
      • “A person’s credences in any two propositions should be equal if her total evidence no more supports one than the other (the evidential symmetry version), or if she has no sufficient reason to have a higher credence in one than in the other (the insufficient reason version)” SEP

Probability vs Likelihood

From https://www.statology.org/likelihood-vs-probability/:

  • Probability refers to the chance that a particular outcome occurs based on the values of parameters in a model.
  • Likelihood refers to how well a sample provides support for particular values of a parameter in a model.

From https://www.psychologicalscience.org/observer/bayes-for-beginners-probability-and-likelihood:

  • “Probability attaches to possible results; likelihood attaches to hypotheses.”
  • Possible results are mutually exclusive and exhaustive; hypotheses are often neither.

“To decide which of two hypotheses is more likely given an experimental result, we consider the ratios of their likelihoods. This ratio, the relative likelihood ratio, is called the ‘Bayes Factor.’”

From somewhere else:
Conditional probability of A given B can be transformed into the likelihood of B given A:

  • $P(A \mid B) = L(B \mid A)$

Notes

For more on probability & randomness (including examples), see [[Mlodinow, Leonard (2008) - “The Drunkard’s Walk”]]

References

Gibbard’s Proof

Type:: #note
Status: #
Tags: #philosophy #philosophy_of_language #conditionals

Overview

[! Abstract] In a Nutshell
Seems to demonstrate that:

“any [[Conditionals|conditional]] operator $\to$ satisfying (i) and the two additional rather obvious principles (ii) and (iii) reduces to material implication.” ([[Kratzer, Angelika (2012) - “Modals and Conditionals”|Kratzer 2012]], 87)

“if indicative conditionals have truth conditions, they cannot be stronger than material implication” ([[Khoo, Justin (2013) - “A Note on Gibbard’s Proof”|Khoo 2013]], 153)

[! Info] Assumptions
(i) $p \to (q \to r)$ and $(p \land q) \to r$ are logically equivalent. (This is also called [[import, export]])
(ii) $p \to q$ logically implies the corresponding material conditional. That is, $p \to q$ is false whenever $p$ is true and $q$ is false.
(iii) If $p$ logically implies $q$, then $p \to q$ is a logical truth.

Objections

[[Kratzer, Angelika (2012) - “Modals and Conditionals”|Kratzer 2012]]: Gibbard assumes that “if…then in English corresponds to a two-place propositional operator.” According to [[Kratzer’s restrictor analysis of conditionals]], this is not the case.
It follows from [[Stalnaker’s ‘close-possible-worlds’ Account of Conditionals#Some consequences of the theory|Stalnaker’s account of conditionals]] that (i) is false. I believe that [[Lewis’ ‘modal’ account of conditionals]] also invalidates (i).

References

#AllanGibbard (Gibbard 1981)
[[Khoo, Justin (2013) - “A Note on Gibbard’s Proof”]]
[[Kratzer, Angelika (2012) - “Modals and Conditionals”]]

4

I’m already familiar with these concepts, but you’ve done a terrific job of distilling note-taking with Obsidian. Clear writing, well laid out, practical, opinionated but not dogmatic. I think the Obsidian team could usefully point to this as a key resource for their software.

Q: is this produced using Obsidian Publish? I like the right-hand marginalia a lot.

1 Like
5

Thanks, I’m really happy you think so! No, I made a custom css/html/js design using neocities – I didn’t really need the features Obsidian Publish offers because the website is basically one page.

6

I enjoyed reading your guide. It’s laid out well.

What are your thoughts on note titles? I have found them to be an important structural element of my vault structure.

A note writing principle I would add is ‘Conversation with yourself’. If at some future point you find that you disagree with a note, don’t make major edits to the existing note. Instead, create a new one discussing your new train of thought and add a link in your old note.

For practical implementation, ‘keywords’ wind up being important to me for searching. If I am searching for an idea, I sometimes rely on certain phrasing, keywords, or verbiage to locate them. It’s not something I stress over when writing a note, but if I have a hard time finding a note I will make strategic edits to it based on search terms I used trying to locate it.

1 Like
7

The form of my note titles is partly dependent on note type:

  • My reading note titles have a fixed form: Surname, Name (Year) - “Title” (I prefer having more information in the title than just surname & year)
  • My regular & structural notes I just try to name descriptively; and I mostly use concept-titling (“Note Atomicity”), and sometimes thought-titling (“notes-should-be-atomic”). Still experimenting with which I like more when.
  • For some note types that I use a lot more rarely, I prefix the note type in the title: “Course Notes - Course name” for course notes, and another prefix for my meta notes about the vault itself.

But now I’m very interested in how you’re approaching note titles if you find them to be important structural elements – does your titling have something to do with note type, or does it encode different information like how they’re connected to other notes?

Regarding “conversation with yourself”: It makes a lot of sense to me to not delete all your notes you now disagree with, but to preserve the evolution of your thoughts! But two thoughts:

  • I suppose how exactly I deal with the note depends on the circumstances?
    • If something was just factually wrong, I might edit it without a trace, or, if it’s an error I see myself doing again, edit and put a callout warning me of the error? (I’ve done this a couple of times with concepts I was prone to confusing; apologies if this is an obvious/boring distinction)
    • If it’s my thoughts on something that changed, either create a new note or edit the note in a way that preserves my old thoughts too (I suppose the decision depends on note granularity / vibe considerations?)
  • In the way I set it up, I’m wondering whether this would maybe best be integrated into note dynamicity rather than being its own principle?

Adding keywords after tough searches sounds like a good simple trick, and I’ll implement it myself.

Thanks for sharing – I’ll integrate some of this into the guide with my next edit!

8

very well put together and very helpful! thank you for sharing.

I especially found your discussion on folder minimizing and note types helpful. I am a chronic nested folders user, this reading is already helping me a lot

1 Like
9

Thank you for this! I really like how you reference future and past mention using (G1) or (N2) and so on.. I wonder, if you are willing to share, how to do actually anotate your own note? is it a special callout type (aside)?

There is a lot of nice information and it is well laid out, thank you, going to do my own cheatsheet base on it :slight_smile:

1 Like
10

Glad it helps! I’m not entirely sure which annotations you mean – are you referring to the (G1), (N2) cross-references or right margins in this guide, or to how I annotate my actual Obsidian notes?

If it’s the guide stuff, that’s just HTML/JS on the webpage. Though you could do similar cross-references in Obsidian with internal links and hover previews, and something like margin notes with footnotes and the footnotes view core plugin.

Feel free to clarify if I misunderstood your question :slight_smile:

PS: I just updated the guide with some minor additions and implemented @Montblanc’s suggestions!

11

Has anyone yet asked for details about the AI Implementation? I would like to read more about that

12

what is the process of publishing obsidian to neocities ?

1 Like
13

I’m not sure how you’d best do that. I did my thinking/writing in Obsidian and then built the Neocities side by hand.

Maybe this is helpful?: Obsidian Neocities integration

14

Really appreciate how you structured this from principles → implementation. That order alone already avoids a common trap I’ve seen (implementation before intent).

One thing I’m curious about:

How much of your current structure emerged organically from accumulated notes, versus being defined early as a guiding architecture?

In other words — did your system grow out of friction you repeatedly encountered, or did you begin with a conceptual model and refine toward it?

I’ve been looking at recurring beginner struggles, and one pattern that stands out is premature architecture — people importing sophisticated structures before their own usage patterns stabilize. Your guide feels different in that it’s clearly battle-tested.

Also curious:

If someone tried to adopt your framework today, what would you recommend they not implement immediately?

Where would you suggest they start smallest?

1 Like
15

Hey, interesting project! I very much share the impression that many beginner struggles are related to: (i) implementing too much complexity from the get-go & (ii) focusing on implementation details over what one wants to get out of the vault. And I agree that one should adapt as one’s usage patterns stabilise.

I did the same in the beginning and ended up abandoning Obsidian for a while because it ended up being too much work “next to” all the other things I had to do for university. Then when I came back to it, I completely scrapped my prior system and mainly focused on making note-taking as effortless as possible. So I suppose the answer to your question would be: I developed the core of the current system in response to all kinds of frictions encountered with the old system, when I already had like 200 (?) notes and some working experience with Obsidian. And then I adapted some more things over time; e.g. further reducing tagging (I didn’t feel like I was getting any value out of tags beyond a core set) and adding (/honing) note types.

On the most minimal starting point: Tough, because I was already trying to be minimal in the guide :grin: I think I’d say:

  • The principles (2.4 & 2.5) are kind of one package, not sure what to take away here. Maybe I’d say: Really take the non-perfectionism to heart, and e.g. don’t be too obsessed with atomicity in the beginning. From my experience, having very high standards for every note is a big/typical inhibiting factor in the beginning.
    • And the core would be: Focus on what notes need to look like to be useful to you. Both useful while writing and useful when reading them later. In other words: Think about why you’re taking notes.
  • And relating to implementation:
    • Forget about plugins & settings in the beginning (unless it’s really fun & motivating for you)
    • Just have one vault, and one big folder where you create every note (and maybe add a template-folder & a file/attachment-folder, if needed)
    • Ignore tags unless you have very specific use-cases for them (like tagging note types or authors); but don’t ignore linking (it’s easier in the beginning anyway, because there are less notes to potentially link to)
    • Note-types I honestly don’t know what’s best to start with because I started thinking more about this when I was already along a bit. Maybe “start with one note type and add more if it turns out to be necessary” is good advice. But for academic use-cases at least reading notes seem to be an extra kind of note type relevant from the get-go. But e.g. structural notes don’t become relevant until the vault is a bit bigger.

Ok this got a bit longer than I intended… Hope this answers your questions :slight_smile:

1 Like
16

This is an interesting approach. You wouldn’t happen to have a sample/starter vault that implements this?

I understand better when I can see the methodology in practice.