About Schemas and Propositions

Where I'm going with this

"Schemas and Propositions" is a set of notes towards a better understanding of language and mental models of all kinds. Both human and artificial. I am interested in all the inexpressible tacit knowledge that makes constructing and understanding propositions possible.

I start with a simple provocative statement:

Schemas is all you need.

Or, to be a bit less cryptic:

Language models represent and process their knowledge as schemas but express it in the form of propositions.

The point of this series of notes is to explore the nature of what schemas and their tension with propositions. But really it's even more about the consequences of this dichotomy on how we think about semantics. But it's not just the semantics we have as humans but also the sort of semantics we find in various models of language, meaning and cognition - be they Large Language Models or just simple toy models of constructed by linguists, philosophers and logicians.

The idea of schemas is not new and part of my journey here will be looking at their history. But despite a long tradition in many fields that has found schemas to be essential to explaining how the mind works, most people who debate how humans and artificial language models deal with language seem to ignore the notion completely. Instead, they behave as if propositions were the only game in town and the myriad of contradictions that this throws up mere trifles to be sorted out later.

This Substack is an attempt to provide a better guide to the notion of schema and illustrate why it is so important.

I want to show why David Rumelhart was correct when he said in 1980:

Schemata are the building blocks of cognition. They are the fundamental elements upon which all information processing depends. "

But I also want to take up the question of why it is that next to noone has taken up this programme.

Where I'm coming from

On a personal level, this Substack is an attempt to a more programmatic writing. For over 10 years, I kept similar notes on MetaphorHacker.net but I haven't posted there for a few years because I've been focusing on many related but not quite relevant projects. Here are just some of them:

1. Random musings on practical epistemology:

   1. Promising Paragraphs

   2. X

2. Questions of Artificial Intelligence:

   1. Trends in AI

   2. Semantic Machines

   3. Practical AI

   4. Experiments in AI Creation

   5. AI in Academic Practice

3. Questions of practical pedagogy:
   1. Deliberate Practice as a Universal Learning Method
   2. Building Your Language Muscle
   3. Creating and Using Instructional Videos​
   4. Principles of Readability
   5. Digital Technologies for Academic Productivity
   6. Czech Language Navigator Companion
   7. Paragraph a Day for a Month: Academic Writing Boot Up Diary Template

And even more remains unshared in countless notes in various places and even in slides of presentations I've given but never pursued in writing.

So despite having written tens of thousands of words over the last few years, I've felt myself to be tongue tied. This is a place where I hope I can write towards a similar goal but under a more unified banner.

Because even though, my writing and other work has been all over the place disciplinarily, it is unified in its focus on how any of this is at all possible given the minds and language we have.

In short

Propositions is what we have in front of us but schemas is what we negotiate as we go through life. If you don't believe me, stick around and let's see if I can convince you.