Questions-Claim-Evidence/Discourse Graph in Obsidian?

Has anyone ever tried to implement Joel Chan’s Discourse Graph workflow in Obsidian, I guess using Dataview, maybe DB Folder etc?

Here’s a really nice demonstration of it in Tana: Advanced Note Taking in Tana with QCE! - YouTube

And Joel’s paper on it: Knowledge synthesis: A conceptual model and practical guide · Open and Sustainable Innovation Systems (OASIS) Lab

This is one of the reasons I (and quite a few people I know) never managed to fully make the leap from Roam to Obsidian, and why I’m currently tempted by Tana. But for many other reasons I’d actually rather be fully in Obsidian if I could get this working here. This workflow felt impossible/too hard to implement in Obsidian without a lot of friction though. Seeing how far some of the plugins have come, I’m guessing we might now be closer to managing something like the Tana example above.

Any thoughts/experiences?

You can already build something like that using metadata-powered plugins (dataview, dbfolder, metadata edit, etc etc). Obsidian itself will improve its support for this in the future

I wanna add a note that is tool-independent about this workflow.
If you do follow QCE, or any other highly structured system for your notes, you are essentially condemning yourself to do the work every time you read/write something, categorize it appropriately in this database.

You are not paid for having a neatly organized vault.
You are making the assumption that this daily chore will pay off at the time of producing your output (for example a paper, a post).
You are banking that:
(Time Spent Keeping The Database Organized + Time Using the database to produce an output) < (Produce an output with an unorganized Vault)

I am very skeptical of this assumption. I am skeptical of frontloading the work, because you do not know if said work will even be useful in the end (how many reviewed papers will end up in the final product?). You don’t know if the paper you categorize today will be useful 2 years from now.

Example from a CS Task: sometimes you need to perform a search in a list that grows over time. Is it better to keep the list sorted as your receive datapoints or to just search the unsorted list?
Keeping the list sorted is a small chore to do every time there’s new datapoint, searching a sorted list is fast. Searching an an unsorted list is slow(er).
Which is better in the end? It depends on the access pattern.

Another two side points:

  1. The more structures you use, the more you tie yourself to a tool that can parse those structures.
  2. Even if I don’t technically blame structured data for this, there is a potential detrimental psychological aspect to it: the dread it can produce every time you have to add something or the discomfort you may feel when you fail to keep things neatly organized (because life happens).

Whenever I think to introduce structure, I try to figure out if it’ll be worth it.


Is this an attempt to overthink it? In reality, QCE can just be linked notes for each element within Obsidian, that way, each note is useable across multiple contexts. ie, any evidence note could be linked to multiple claims etc.

A Discourse Graph Extension as it exists in Roam Research would be of great interest! +1