Re: Representing Procedural Knowledge: Examples and Use Cases

Hi Adam,

Thank you for kicking off this discussion! The examples you shared 
(schema.org's HowTo/Recipe, BPMN/BPEL) are excellent references for 
workflow and instruction representation.

PM-KR's focus complements these by addressing a slightly different 
challenge: **dual-client procedural knowledge** where the same canonical 
source serves both humans (readable) and AI systems (executable).

**Key distinction:**
- schema.org/HowTo → describes *how to perform* a task (instructions for 
humans)
- PM-KR → defines *knowledge as executable procedures* (like TrueType 
fonts, where glyphs are Bézier programs, not pixel arrays)


**A mathematical analogy (building on Milton Ponson's foundational work):**

Consider how high school mathematics works: the basic operators (+, −, 
×, ÷, =) and variables (x, y, z) are atomic symbols with defined 
procedural meanings. These compose into the quadratic equation (ax² + bx 
+ c = 0), which composes into systems of equations, which compose into 
calculus, linear algebra, and beyond.

**We don't duplicate the definition of "+" every time we write a new 
equation.** We reference it. That's the genius simplicity PM-KR 
formalizes: atomic knowledge units (symbols with procedural definitions) 
compose into arbitrarily complex structures without duplication.

It seems almost magical, but it's actually **simplistic elegance** — a 
solution that was always there, hiding in plain sight across 
mathematics, typography, and every compositional system. Milton's 
n-dimensional framework provides the mathematical foundation for this; 
PM-KR implements the 3D spatial case with empirical validation.

**PM-KR-specific use cases:**

1. **Educational knowledge** - Textbooks where the same procedural 
source renders visually for students AND executes computationally for AI 
tutors
2. **Game mechanics** - Rulebooks as executable programs (not just text 
descriptions)
3. **Scientific protocols** - Experimental procedures that humans read 
AND machines execute
4. **Mathematical knowledge** - Symbols/formulas as procedural 
definitions (composition, not duplication)
5. **Spatial knowledge** - Geometric primitives as programs (LINE, 
CIRCLE, RECT) that compose into complex structures

**Re: centralized trusted services** - Agreed! That's exactly where 
standards matter. PM-KR aims to provide the foundational layer for such 
services.

I've published more details on our mission and scope here:
https://www.w3.org/community/pm-kr/procedural-memory-knowledge-representation-pm-kr-community-group/

Looking forward to exploring overlaps with schema.org, BPMN, and other 
procedural knowledge formats!

Best regards,
Daniel Campos Ramos
PM-KR Co-Chair

On 2/24/26 1:11 PM, Adam Sobieski wrote:
> PM-KR Community Group,
>
> Hello. With respect to representing procedural knowledge, I would like 
> to share the following examples and use cases with the group.
>
> Firstly, existing schemas for procedural knowledge (and recipes) 
> include schema.org's:
>
> * https://schema.org/HowTo
> * https://schema.org/HowToSection
> * https://schema.org/HowToStep
> * https://schema.org/HowToDirection
> * https://schema.org/HowToTip
> * https://schema.org/Recipe
>
> Secondly, existing formats for business processes and workflows include:
>
> * BPMN ( 
> https://en.wikipedia.org/wiki/Business_Process_Model_and_Notation )
> * BPEL ( 
> https://en.wikipedia.org/wiki/Business_Process_Execution_Language )
> * XPDL ( https://en.wikipedia.org/wiki/XPDL )
>
> Thirdly, a use case for representing procedural knowledge involves 
> enabling and advancing centralized, trusted services for providing 
> procedural knowledge to people, organizations, AI agents, and 
> multi-agent systems.
>
> * https://www.openfn.org/
> * 
> https://www.theverge.com/news/785193/google-deepmind-gemini-ai-robotics-web-search
>
>
> For discussion, are there any other examples or use cases for 
> representing procedural knowledge to consider? Thank you.
>
>
> Best regards,
> Adam Sobieski
>

Received on Tuesday, 24 February 2026 16:21:24 UTC