Toward a Representational Architecture of Reality
18/03/26
Over the past recent period, I have been working on a paper that explores a simple but far-reaching question:
What do all models of reality have in common?
Across disciplines (physics, economics, artificial intelligence, law) we build models to describe, predict, and understand the world. These models differ widely in language, assumptions, and purpose. Yet they all face the same fundamental task: to relate what is described to what is observed.
The work I’ve just submitted for publication proposes that, at a structural level, many of these models share a common feature that is often left implicit: they depend not only on a description of a system, but also on the context in which that system is interpreted.
In everyday terms, this is intuitive. The same situation can lead to different outcomes depending on conditions, perspective, constraints, or background assumptions. What my submitted paper explores is how this intuition can be expressed more precisely and how making it explicit can help us compare models that would otherwise appear incommensurable.
The central idea is straightforward:
Models are not just descriptions of systems; they are descriptions of systems in measurable context.
When context is omitted or under-specified, models can appear complete while still producing inconsistent or misleading results. When context is made explicit, differences between models often become easier to understand, analyse, and sometimes even align. In study terms, a practical proposal for this is outlined here.
While I can’t share much more about at it at this stage, the paper develops a general framework for thinking about this structure, with the aim of supporting clearer comparison across domains rather than replacing existing theories. It draws connections between areas such as probabilistic modelling, quantum measurement, and artificial intelligence, showing that context-dependent structure is already present in many successful approaches.
Importantly, the work does not attempt to reduce different fields to a single theory. Instead, it asks whether there is a shared representational layer that allows them to be related without forcing them into the same ontology.
If this is right, then part of the fragmentation across disciplines may not come from fundamentally incompatible theories, but from differences in how context is handled, represented, or left implicit.
This is an ongoing line of work, and I’ll share more once the review process is complete. In the meantime, I’m interested in how others think about context in modelling. This is the first time I have asked readers here for response or context but I think it is valuable especially in cases where seemingly similar systems behave differently under different conditions.
Because often, that’s where the most interesting structure is hiding. Anyway, that’s what I have been busy doing. Normal service of writing on this blog resumes.
S.
