Saturday, August 8, 2026

Having the Claude AI create a concept map and mindmap for a blog article or series of blog articles

Below is a prompt to give the Claude AI when creating a conceptual mind map (or conceptual mind maps) for a blog article:

Read the entire article carefully. Create the minimum number of SVG mind maps
necessary to represent the article's conceptual structure. Prefer one map; use
multiple only if the article contains genuinely distinct conceptual systems that
would overlap or become illegible in a single diagram.

CONTENT RULES:

  • Represent underlying knowledge structure, not paragraph summaries
  • Preserve causal/influence relationships, distinctions, and mechanisms
  • Remove minor examples, repetition, incidental detail
  • Do not add unsupported information
  • After drafting, re-check the map against the article and correct any omitted
    or distorted concept, relationship, or qualification before finalizing

VISUAL / LAYOUT RULES:

  • Output raw SVG only — no HTML wrapper, no <img>, no external libraries, no JS
  • Use a radial or left-to-right hierarchical tree layout (specify which)
  • Set viewBox explicitly (e.g. "0 0 1600 1000") and make width/height 100% so
    it scales responsively inside Blogger's fixed-width post column
  • Central/root node in the center or far left; branch outward by hierarchy level
  • Cap hierarchy at 3 levels deep (root → main branch → sub-node) — collapse
    anything deeper into the sub-node's label rather than adding a 4th tier
  • Limit to ~6-8 main branches and ~4-5 sub-nodes per branch max, to prevent
    visual clutter and node overlap
  • Wrap node text at a fixed character width (e.g. ~28 chars/line) using
    multiple <tspan> lines rather than letting text overflow its shape
  • Give every node a background shape (rounded rect or ellipse) sized to fit
    its wrapped text, not fixed-size boxes with overflowing text
  • Use a consistent color per hierarchy level (root, branch, sub-node) — 3-4
    colors max, applied via inline fill/stroke attributes (no CSS classes,
    since Blogger strips <style> blocks unpredictably)
  • Render causal/influence relationships as directional arrows (<marker> or
    manually drawn arrowhead paths) with a short inline label if needed
    (e.g. "shapes", "→ reinforces →")
  • Ensure no text or arrow crosses over another node — check coordinates
    before finalizing, don't estimate
  • Use a readable sans-serif font-family fallback stack (e.g. Arial, Helvetica,
    sans-serif) since custom fonts won't load in Blogger
  • Keep a consistent stroke-width and font-size scale across all nodes at the
    same hierarchy level for visual consistency

TEXT-FIT RULES:

  • Size every node's background shape from the actual measured width of its text at
    the target font/weight/size — not a character-count formula. Use a real font
    metrics source (e.g. a metric-compatible font like Liberation Sans Bold for
    Arial Bold, measured via a library such as PIL/Pillow, or an equivalent text-measurement
    tool) to get each line's true rendered width, then add ~15-20% horizontal
    padding as a safety margin.
  • Do not use the textLength/lengthAdjust="spacingAndGlyphs" attribute. It forces
    the renderer to stretch or compress a line's glyphs to exactly match the given
    width — if every line in a multi-line node is given the same textLength (e.g. the
    node's overall inner width), short lines get their letters stretched wide to fill
    it, producing visibly oversized, distorted words. If textLength is used at all, it
    must be computed per-line from that line's own measured width, never shared across
    lines in the same node — but given accurate width measurement above, it is not
    needed and should simply be omitted.
  • Vertical padding: node height must accommodate all wrapped lines at
    1.3× font-size line-height, plus at least 16px total top/bottom padding.

READABILITY / SCALE RULES:

  • Design the canvas assuming final display width will be ~600px on desktop
    and as narrow as ~380px on mobile — Blogger's column, not a full monitor
  • Set viewBox width no larger than ~1000-1100px for a single-column diagram.
    If the content requires more room than that to stay legible, split into
    additional diagrams rather than widening the canvas further
  • Minimum font sizes in source units (before any shrinking):
    root/title nodes ≥ 30px, branch/category nodes ≥ 24px, leaf/sub-nodes ≥ 20px,
    annotation/caveat text ≥ 17px — never smaller, even under space pressure
  • If node count forces text below these minimums to avoid overlap, that's a
    signal to remove a node, shorten its label, or split into another diagram —
    not to shrink the font
  • Cap total nodes per diagram more conservatively: ~4-5 main branches with
    2-3 sub-nodes each (12-18 nodes max), rather than 6 branches x 3 subs (18-24).
    Wide hexagonal/radial spreads eat horizontal space that mobile doesn't have —
    prefer a vertical or top-to-bottom layout over a wide radial one when the
    concept allows it

VERIFICATION STEP (mandatory, not optional):

  • Before delivering, rasterize the SVG to a PNG at both full size and a 400px-wide
    mobile simulation — do not rely on mental estimation of whether text fits.
  • Use a renderer that honors the same SVG text/attribute behavior as a real browser
    (e.g. rsvg-convert/librsvg, or an actual headless browser). Do not rely solely on
    cairosvg for this check — it silently ignores some attributes (including
    textLength) that browsers apply, so a clean cairosvg render does not guarantee a
    clean render on the actual publishing platform.
  • Visually inspect the rendered PNG and confirm: no character in any node
    touches or crosses its shape's edge, no line's text is stretched or compressed
    relative to other lines in the same node, no text or arrow crosses over another
    node, and every label is legible at 400px width.
  • If any overflow, overlap, or distortion is found, widen the affected node, shorten
    the label, or adjust spacing, then re-render and re-check before finalizing.
    Repeat until the rendered image is clean — do not deliver on the first pass
    without this visual check.

DELIVERY:

  • Output must be a single self-contained <svg>...</svg> block, valid XML,
    ready to paste directly into Blogger's HTML view with no further editing
  • If more than one map is genuinely needed, output each as a separate
    complete <svg> block with a plain-text heading above it (not inside the SVG)


Questions to ask other AIs (ChatGPT and Gemini) if Claude is not available

Where it gets less certain:

  • The mandatory render-and-inspect step assumes the model can actually rasterize the SVG and look at it before delivering — that requires a code-execution or image tool in that specific chat. I know Claude has this here; I honestly don't have reliable, current knowledge of what tools ChatGPT or Gemini expose by default in a given conversation, since that changes over time and by plan/mode. If neither has that capability active, they'll "confirm" the verification step happened without actually doing it — which defeats the point of that rule. Worth testing once with a real render rather than trusting it.
  • Without that visual check, they're back to estimating text width the same way I was before you caught the problem — textLength still guarantees no overflow mathematically, but layout choices (spacing, node sizing) that aren't governed by a hard attribute are more exposed to the same kind of estimation drift.

Practical suggestion: since this is just a fallback for Claude being unavailable, I wouldn't over-invest in testing it now. If that day comes, hand the same instructions to whichever one you're using and specifically ask up front whether it can render and inspect the SVG before delivering — if it says no or dodges the question, you'll know to eyeball the output extra carefully yourself before pasting into Blogger.

Using Conceptual Mind Maps to Learn This Blog’s Article Series on Beliefs

For this blog’s 32-article series, The Psychology of Belief: 32 Part Series – How Your Inner World Shapes Your Outer Life, I think conceptual mind maps are particularly useful.

My Recommendation

For this 32-article series, I would use them in three stages.

Stage 1 — Now:
Create an AI-generated conceptual mind map beneath each article.

Stage 2 — After the series is mature:
Create a master conceptual map connecting the major concepts across the articles.

Stage 3 — Learning layer:
Use the maps to generate retrieval questions or flashcards for the most important concepts and relationships.

That gives you:

Articles → Individual conceptual mind maps → Master conceptual map → Retrieval practice

That is a remarkably efficient way of turning a collection of articles into an actual structured learning system rather than simply a collection of web pages.

I Would Actually Use Two Levels of Conceptual Maps

1. A Conceptual Mind Map for Each Individual Article

Put it directly below the article, exactly as proposed.

For example, article #13, How Beliefs Are Formed, could have a map organized around:

Belief Formation

  • Quality of thinking
  • Repetition
  • Emotion
  • Authority
  • Experience
  • Interaction among the five forces
  • Implications for changing beliefs

The purpose is not to reproduce the article in diagram form. The purpose is to give the reader a compact representation of its conceptual structure.

A good map should help answer questions such as:

  • What are the major concepts?
  • How are they related?
  • What causes or influences what?
  • What distinctions does the article make?
  • What mechanisms does it describe?
  • What practical conclusions follow from those relationships?

2. Eventually, One Master Conceptual Map for the Entire Series

This is where I think the project becomes particularly interesting.

The 32 articles are not simply a collection of unrelated topics. They form a broader conceptual progression.

What beliefs ARE
→ definition
→ layers
→ beliefs vs. truth vs. knowledge

How beliefs FORM
→ quality of thinking
→ repetition
→ emotion
→ authority
→ experience

How beliefs OPERATE
→ mindset
→ interpretation
→ behavior
→ habits
→ self-fulfilling loops

How beliefs CHANGE
→ evidence
→ updating
→ identity
→ CBT
→ cognitive restructuring

How beliefs INTERACT WITH LIFE
→ emotional regulation
→ assumptions
→ social transmission
→ performance
→ work and self-employment

That is essentially a conceptual architecture of belief.

The individual maps would represent the architecture of each article. The master map would represent the architecture of the entire series.

Why the Two-Level Approach Is Useful

The individual maps and the master map serve different purposes.

The individual map answers:

“What are the important concepts in this article, and how do they relate?”

The master map answers:

“How do the ideas from all of these articles fit together into one larger system?”

This distinction becomes especially valuable as the number of articles increases.

A reader who encounters article #20 might otherwise have to remember what articles #3, #8, #13, and #17 were about. The master map can provide a higher-level representation showing where the current article fits within the larger intellectual structure.

In other words, the master map can give readers a mental framework into which new information can be placed.

The Learning Layer

The maps can also become the foundation for a more powerful learning system.

Instead of turning every sentence in an article into a flashcard, important concepts and relationships from the conceptual maps can be converted into retrieval questions.

For example:

Concept map:
Emotion → influences → belief formation

could generate a retrieval question such as:

“How can emotion influence belief formation?”

Similarly:

Belief → influences → interpretation → influences → behavior

could generate:

“Explain the pathway through which beliefs can influence behavior.”

This creates a useful progression:

Article

Conceptual mind map

Important concepts and relationships

Retrieval questions

Spaced retrieval

The map therefore isn't replacing a flashcard system. It is helping determine what is worth remembering and how the pieces fit together.

Eventually, the Maps Can Become a Knowledge Architecture

There is another potential advantage to building the maps incrementally.

As more articles are mapped, recurring concepts will become visible.

You may discover that a concept introduced in an early article becomes a mechanism in a later article, that two apparently different topics are connected, or that several articles are actually examining different aspects of the same underlying process.

The master map can therefore evolve as the series develops.

Instead of thinking of the blog as:

32 separate articles

you can eventually represent it as:

One interconnected conceptual system expressed through 32 articles.

That is a much more powerful structure for learning.

What I Would Have the AI Do

I would not simply tell the AI:

“Make a mind map of this article.”

That instruction is too vague. The resulting diagram might simply reproduce the article's headings or produce a visually attractive summary without accurately representing the relationships among the ideas.

The more detailed prompt at the beginning of this article is better because it tells the AI what the map is actually supposed to accomplish.

In particular, the AI should:

  • read the entire article;
  • identify its underlying conceptual structure;
  • identify important relationships among concepts;
  • preserve important distinctions and qualifications;
  • remove minor examples and repetition;
  • avoid unnecessary nodes;
  • compare the completed map with the original article;
  • correct omissions or distortions; and
  • avoid introducing information that is not supported by the article.

The goal is not to make the largest or most elaborate diagram possible.

The goal is to make the smallest map that accurately represents the important conceptual structure.

Conceptual Mind Maps Rather Than Formal Concept Maps

There is an important terminology issue worth clarifying.

A traditional mind map is generally a hierarchical diagram organized around a central topic. A formal concept map places greater emphasis on explicit relationships among concepts, often using labeled connections and cross-links.

The maps described in this article occupy something of a middle ground.

They use the hierarchical structure and software commonly associated with mind mapping, but their purpose is more conceptual: representing the important ideas and relationships in an article.

For that reason, conceptual mind map is a useful description of the approach.

The terminology is less important than the function. The objective is to transform a linear article into a visual representation of its knowledge structure.

The Overall System

For this 32-article series, I would therefore use the following architecture:

32 Articles

32 Individual Conceptual Mind Maps

1 Master Conceptual Map

Important Concepts and Relationships

RemNote/Anki Retrieval Questions

Spaced Retrieval

Application and Integration

Each layer has a different purpose.

The articles provide detailed explanations.

The individual maps provide compact conceptual structures.

The master map shows how the articles fit together.

RemNote and Anki provide retrieval practice and spaced review of important knowledge.

Application tests whether the knowledge can actually be used.

Bottom Line

I think adding conceptual mind maps beneath the articles is a worthwhile improvement to the series, particularly because the articles form a larger conceptual system rather than being completely independent pieces.

The most efficient implementation is to build the individual maps now, rather than waiting until the entire series is finished. Once enough articles have been mapped, create the master conceptual map that connects the major ideas across the series.

Then use that conceptual structure to create retrieval questions and integrate the material into a spaced-repetition system.

The resulting architecture is:

Articles → Conceptual mind maps → Master conceptual architecture → Retrieval practice → Application

That turns a collection of blog articles into something considerably more useful: a structured, interconnected learning system.

No comments:

Post a Comment

Epistemology: What is it? Why is it important? How to have more epistemological discipline

Every belief you hold—about God, politics, markets, other people, and yourself—rests on some method for deciding what counts as good enough ...