In which I collect my thoughts on many topics but mainly about systems engineering, software engineering, and system/software architecture

Ontic Accessibility vs Epistemic Probability

Why This Matters

The intent of this post is to stabilize interpretation of probabilities that we introduced in the previous post and lays the groundwork for the conceptual foundation for architectural basins in future posts.

We start with the principles upon which we introduced plausibility in the previous post. Those principles are

  1. Probability does not define architectural possibility.
  2. Probability operates within a space of possibilities that architecture constrains.

The Core Idea

We have said that architecture constrains what futures are reachable. Previously we introduced the idea that probability expresses uncertainty over which reachable future may occur. These are fundamentally different categories. Although causal chains and Bayesian networks are both often represented as directed acyclic graphs, the semantics differ in the following ways

  • Causal structure describes admissibility
  • Probability describes expectation

Bayesian models may be defined on top of causal structure, but causal structure is not itself probabilistic.

Structural Formulation

We defined architecture as an induced causal structure represented by

(E,\prec)

For architectural state x \in E, we define the admissible future region as

F(x) = \{ y \in E \, \vert \, x \prec y\}

We then say that F(x) represents all reachable differentiations of the system. Note that in this expression accessibility exists independently of knowledge or uncertainty. Probability may then be defined over this structure via

P(y \, \vert \, x)

but probability presupposes that admissibility has already been established. From this we can say that architecture determines the structure of possible futures. Probability operates within that structure.

Ontic vs Epistemic Interpretation

We use the terms ”Ontic” and ”Epistemic.” Ontic refers to the properties of the system itself, while epistemic refers to properties of knowledge about the system. Causal relations belong to the ontic category. Probability belongs to the epistemic category.

Confusion arises when epistemic language is used to describe structural restriction. For example: ”high probability region” may be interpreted incorrectly as ”structurally stable region.” Yet probability concentration may occur even when structural accessibility is low. Conversely, structurally rich future regions may have low probability simply because little information is available.

Capacity Formulation

To quantify preserved architectural freedom without imposing probabilistic semantics, we define a valuation on admissible futures. Let

A \subseteq F(x)

and define

V_x(A)

We impose the minimal conditions

V_x(\varnothing) = 0

and

A \subseteq B \Rightarrow V_x(A) \le V_x(B)

V_x is a monotone set function which we identify as ”capacity”. We do not require additivity in this case

V_x(A \cup B) \neq V_x(A) + V_x(B)

Future regions may thus interact structurally. For example, two alternatives may

  • Share enabling structure
  • Compete for architectural slack
  • Reinforce integration flexibility
  • Restrict reconvergence options

Accessibility value may therefore be subadditive or superadditive.

Interpretation for Architecture

Capacity expresses preserved structural freedom by maintaining accessibility to future regions. We will see later that technical debt can be interpreted as deformation of accessibility valuation

\Delta V = V_x(A) - V_y(A^{\prime})

where x \prec y. Loss of architectural freedom corresponds to contraction of valued accessible futures.

Canonical Distinction

Probability ranks futures by expectation. Capacity ranks future regions by preserved accessibility. Expressed differently we can say that epistemology evaluates paths within the cone while architecture determines the shape of the cone.

What’s Next

In the next post we will provide a worked example to demonstrate

  • Interacting development paths
  • Reconvergence constraints
  • Structural preservation of integration feasibility

Transition to Architectural Basins

All of this discussion leads to the notion of architectural basins. Architectural basins emerge when continuation within a region preserves local accessibility while alternative futures become structurally constrained. Entrenchment reflects deformation of accessibility structure rather than concentration of belief. As mentioned, we will discuss this in more detail in a future post.

Leave a Reply

Discover more from System Architecture, Systems Engineering, and Other Thoughts

Subscribe now to keep reading and get access to the full archive.

Continue reading