Thoughtful with Social AI
A living bundle of principles and practice for AI-supported work with groups.
What is this?
Four documents on how to use AI thoughtfully in work with groups: facilitation, deliberation, neighborhood, team and community engagements. No tool reviews, no prompt tricks, but a working practice that protects people's voice when a language model sits in between.
The bundle is meant in two ways. You can read it if you do this work (or are thinking about it), and you can paste it whole or page by page into your own AI (Claude, ChatGPT) so that AI works thoughtfully on your projects. The principles replace one-off prompt instructions and become a behavior guideline.
Based on practice: years of facilitation, recently the Doesburg engagement (a bottom-up community engagement around a caring community) and the Social AI Field Guide (socialaiveldgids.nl). Continuously evolving, not finished.
The four documents
Give it to your AI
The docs are designed to be understood by a language model. Two ways to do that:
Option A: page by page. Open the raw markdown version of the document you need, copy everything, paste it at the top of your AI conversation with the instruction: "Apply these principles to everything you do for me."
Raw links (open in a new tab, copy everything, paste into your AI):
- SOCIAL-AI-PRINCIPES.md for the 15 principles
- PROMPT-BEST-PRACTICES.md for concrete prompt design
- OwnershipPrinciples.md for the ownership grammar
- BottomUpPrinciples.md for bottom-up system change
Option B: the whole bundle in one. en/bundle.md contains all four docs together, open it, copy everything, paste into your AI.
What an AI does with this: instead of explaining per task "please write in participant register, avoid consultant-speak, label inference as inference", the AI gets this as a behavior guideline up front. The rest of your conversation is about the content.
Visual map
Two layers, four docs. Practice on top, foundation underneath.
┌──────────────────────────────────── PRACTICE ────────────────────────────────────┐
│ │
│ ┌────────────────────────────────────┐ ┌────────────────────────────────────┐ │
│ │ SOCIAL-AI-PRINCIPES.md │ │ PROMPT-BEST-PRACTICES.md │ │
│ │ │ │ │ │
│ │ 15 principles, 7 clusters │ │ A concrete translation │ │
│ │ │ │ into prompt design │ │
│ │ Fundamental · Disposition · │ │ │ │
│ │ Method · Attentive · │ │ 4 facets · 4 core constraints │ │
│ │ Data ownership · │ │ 3 AI value levels · multi-pass │ │
│ │ Anti-decontextualization · │ │ review · DIRECT/INFERENCE · │ │
│ │ AI value levels │ │ quote density │ │
│ └────────────────────────────────────┘ └────────────────────────────────────┘ │
│ │
└──────────────────────────────────────────────────────────────────────────────────┘
▲
│ rests on
▼
┌─────────────────────────────────── FOUNDATION ───────────────────────────────────┐
│ │
│ ┌────────────────────────────────────┐ ┌────────────────────────────────────┐ │
│ │ OwnershipPrinciples.md │ │ BottomUpPrinciples.md │ │
│ │ │ │ │ │
│ │ Where it comes from │ │ How change emerges │ │
│ │ │ │ │ │
│ │ Psychological grammar · │ │ 7 core ideas · │ │
│ │ language principles · │ │ preparation + process · │ │
│ │ scoring methodology · │ │ step-by-step for beginners · │ │
│ │ from the Doesburg work │ │ step-by-step for advanced │ │
│ └────────────────────────────────────┘ └────────────────────────────────────┘ │
│ │
└──────────────────────────────────────────────────────────────────────────────────┘
Practice layer (Principles + Prompts): what you do when AI joins group work.
Foundation layer (Ownership + Bottom-up): why this works, and where the convictions come from.
Two entry points, choose your reading order
Top-down (start with practice): Principles → Prompts → Ownership → Bottom-up.
Choose this if you have a session tomorrow, are building a prompt, or are using AI in ongoing work and want to know which guidelines help. You read the Principles, translate them into concrete prompt design, and only dive into the foundation when you wonder why a principle is the way it is.
Bottom-up (start with the foundation): Bottom-up → Ownership → Principles → Prompts.
Choose this if you want to understand the whole field before you apply it. You first read how bottom-up system change works (the broader context Social AI fits into), then how ownership emerges psychologically and through language, and only then what that means for concrete principles and prompts.
Both paths arrive at the same place. The choice is about where your question is right now.
Living document
Living document, continuously evolving.
- Bundle published: 2026-05-20
- Oldest docs in this bundle: Ownership, Prompts and Bottom-up, all autumn 2025 (Ownership explicitly from 2025-08-26)
- Newest addition: Principles, first version 2026-05-12
Each page has its own "last update" date at the top.
Contributing & feedback
Issues, comments, improvements: open a GitHub issue or send Joost a message via jmfl.nl.
Repo: https://github.com/joostliebregts/thoughtful-social-ai
License
CC-BY-SA 4.0. Share, reuse, build on. Credit the source, use the same license for derivative works.
About the author(s)
By Joost MF Liebregts (jmfl.nl): bottom-up facilitation and the Social AI Field Guide. In collaboration with Finn, AI raid partner in thinking and writing.