Using GraphViz to Build Complex Models Bottom-Up • Zettelkasten Method

System

Using GraphViz to Build Complex Models Bottom-Up • Zettelkasten Method

When you can't see the whole of a complex issue yet, model it bottom-up: keep a GraphViz note of reinforcing and balancing relationships, research each link in its own note, and let the model of the whole assemble as a byproduct.

Read the full story here

harr

I also had positive experiences with diagram-as-code tools like GraphViz, PlantUML and Mermaid. They nicely complement a digital Zettelkasten because one can quickly visualize relationships. They nicely complement Markdown because they are text first.

In Obsidian I use built-in Mermaid for that effect. This is how the post's example could be implemented in Mermaid:

flowchart TD
    Will -->|+| Deterrents
    Will -->|+| Incentives
    Will -->|+| Action{"Action"}

    Deterrents -.->|-| Action
    Deterrents -.->|-| TN["Threshold for<br/>negentropy perception"]

    Incentives -->|+| Action
    Incentives -->|+| ID["Incentive density"]

    ID -->|+| TN

    Action -.->|-| Energy
    Energy -->|+| Action

    Action -->|+| Reward
    Reward -->|+| Conditioning
    Conditioning -->|+| HabitAccount

    Action -->|+| HabitAccount
    HabitAccount -->|+| Action
    HabitAccount -->|+| DP["Desensitization<br/>to pain"]
    DP -->|+| SE["Self-esteem"]

It looks like this:

Personally, I prefer words to +/- symbols. I learned from Novak's concept maps that a lot of insight can be gained by verbalizing the relationship. The quality of the relationship is still indicated by the stroke. For example like this:

Diagram-as-code tools are useful. However, they can't handle too much complexity. At some point one needs to draw a chart manually for a more refined layout.

harr

I found a community plugin that renders GraphViz diagrams in Obsidian. Same idea:

Here's the source code:

digraph G {

    Will -> Deterrents [taillabel = "+"]
    Will -> Incentives [taillabel = "+"]
    Will -> Action [taillabel = "prompts"]

    Deterrents -> Action [taillabel="discourage" color=red style="dashed" labelfontcolor=red]
    Deterrents -> TN [taillabel = "reduce" color=red style="dashed" labelfontcolor=red]
    TN [label="Threshold for\nnegentropy perception"]

    Incentives -> Action [taillabel = "encourage"]
    Incentives -> ID [headlabel = "increase"]
    ID [label="Incentive density"]

    // cf. the dopamine hypothesis [[201808210933]]
    ID -> TN [taillabel = "increase"]

    Action [shape="diamond"]
    Action -> Energy [headlabel = "costs" color=red style="dashed" labelfontcolor=red]
    Energy -> Action [taillabel = "fuels"]

    Action -> Reward [headlabel = "as"]
    Reward -> Conditioning [headlabel = "positive\nreinforcement"]
    Conditioning -> HA [taillabel = "strengthens"]
    HA [label="active\nhabits" shape="doublecircle"]

    Action -> HA [label = "mutually\nfacilitate" dir="both"]
    HA -> DA [taillabel = "can lead to"]
    DA [label="Desensitization\nto pain"]

    // [[202307221041]]
    DA -> "Self-esteem" [headlabel = "increases"]
}

I learned that GraphViz is an excellent alternative to Mermaid in Obsidian. Thanks!