Skip to content
Thinking in Diagrams

Teaching with this book

Drafted for Aryeh's edit.

These notes are for a teacher who sets the book for a class or a workshop. The book is written for a reader working alone: every check in the lab names the section that teaches it, and every answer is shown as one defensible answer among several. The notes on each chapter give its timing, the misreadings its bad diagram, understanding checks, reading and exercises show, and the exercises to set for a session and for afterwards; every minute given for an exercise is an estimate.

Twelve chapters in five parts, read in order. Part I frames a situation: what a diagram shows that a paragraph doesn’t (chapter 1), a system context map’s boundary (chapter 2) and an influence map of what affects an outcome (chapter 3). Part II draws causal loop diagrams (chapters 4 and 5), Part III stock & flow diagrams (chapters 6 and 7), and Part IV feedback control diagrams (chapters 8 and 9). Part V names system archetypes (chapters 10 and 11) and closes with choosing which diagram answers a question (chapter 12). A course of twelve sessions takes a chapter each.

The case, a village primary school losing pupils, is the last exercise of every chapter, CS 01 to CS 12: one situation drawn as every type in turn. Its page gives the situation whole, the question each chapter’s drawing answers about it, and each reader’s saved drawings side by side. Set each chapter’s part for afterwards, and the class reaches chapter 12 with twelve drawings of one school.

Each exercise’s answer waits behind a soft gate: it opens once the reader has saved an attempt with something drawn, and Show me anyway opens it without one. The gate is the reader’s own, and no teacher sees who opened it, so a teacher who wants attempts first says so in the room. Step by step plays the answer back one step at a time, each step’s sentence beneath the drawing, never on a timer; on a projector, a class can say what comes next before each step. A reader’s drawing that differs from the answer can still meet the brief: the lab’s checks test the brief’s structure, not agreement with the answer.

Attempts are saved in the reader’s browser and nowhere else. Export saves one as an image or as JSON, which a reader can hand in.

The lab works on a desktop or a tablet. On a phone, the chapters, their understanding checks and their readings work as anywhere, and each lab shows a notice in its place: a class on phones reads in the session and draws at a larger screen.

For paper, the workbook’s PDF prints each exercise as a sheet, its brief to tick, its palette and a dot grid to draw on, with the answers at the back; the book’s PDF prints the chapters and the field guide.

In the lab, a reader can turn on written feedback: a language model’s reading of their names and effects, which the checks can’t give. It uses the reader’s own Anthropic API key. When the reader presses Ask for written feedback, their browser sends their key, the exercise, the diagram, its checks and the book’s answer straight to Anthropic; this site has no server and never sees them. Anthropic bills each request to the key’s account, at its prices. The lab doesn’t check the reply.

Tell a class: written feedback is optional and off until turned on, and every check works without it; each reader who wants it needs an Anthropic account and key of their own and pays for their own requests; a key shared with the class is billed to its owner for every request anyone sends with it. A key stays in the tab until it closes, or on the browser with Remember on this browser; on a shared computer, Forget my key removes it.