Recursive Representation Flow (RRF)

The Recursive Representation Flow (RRF) is a framework for understanding systems in which representations are repeatedly transformed into new representations, while periodically interacting with the reality they represent.

At its simplest:

Reality → Representation → Re-representation → { Re-representation OR Action → Reality } → …

A re-representation can therefore take two fundamental paths:

Recursive path:

Representation → Re-representation → Re-representation → …

Reality path:

Representation → Re-representation → Action → Reality → new Representation → …

Components

RRF consists of six fundamental components:

Reality: The external or internal state about which representations are formed and upon which actions can operate.
Representation System: Converts some aspect of reality or existing information into a representation.
Re-representation System: Transforms an existing representation into a new representation.
Action: Allows a re-representation to affect, probe, or interact with reality.
Host: The physical system in which representation and re-representation occur.
Energy: Sustains the physical processes required for the flow.

Two important properties can then emerge:

Recursion: A re-representation becomes an input for further re-representation.
Autonomy: The flow continues without requiring external initiation at each step.

The central question of RRF

What happens as a system becomes capable of repeatedly re-representing its own representations, acting on reality, observing the consequences, and continuing the flow?

Example 1: Human Cognition

A human implementation of RRF can be described as:

Language = Representation System
Mind = Re-representation System
Brain = Host
Metabolism = Energy
Human Behavior = Action

Consider:

Reality: Someone disagrees with you.

→ Representation: “She disagreed with me.”

→ Re-representation: “Maybe she doesn't trust me.”

→ Re-representation: “Perhaps my team doesn't trust me.”

→ Action: You ask her why she disagreed.

→ Reality: She explains that she agrees with your objective but questions one assumption.

→ Representation: “The disagreement was about the assumption.”

→ Re-representation: “My original interpretation was probably wrong.”

Reality has re-entered the flow and changed its direction.

Alternatively, no action may occur:

“She disagreed with me” → “She doesn't trust me” → “My team doesn't trust me” → “I'm losing credibility” → …**

The flow continues recursively without returning to reality.

Example 2: Agentic LLMs

An artificial implementation can be described as:

Language = Representation System
LLM = Re-representation System
Computing Infrastructure (Data Centers) = Host
Electricity = Energy
Tools / Agent Actions = Action

A simple LLM primarily transforms an externally supplied representation:

Prompt → LLM → Re-representation → Human

An agentic system can create a continuing RRF:

Reality → Representation → LLM → Re-representation → Action → Reality → Observation → Representation → LLM → …

For example, an agent may form a hypothesis, search a database, observe the result, revise its representation, perform another action, and continue.

The critical change, therefore, is not simply that an LLM can generate language. It is that an agent can close the loop between re-representation and reality.

Why RRF Matters

RRF provides a framework for examining recursion, action, grounding, representational distance, and autonomy.

It asks:

Recursion: How many re-representations occur before reality re-enters the flow?
Grounding: How frequently are representations checked against reality?
Transformation: What is preserved, removed, distorted, or introduced during re-representation?
Action: When do representations produce interventions in reality?
Feedback: How effectively does the system incorporate the consequences of its actions?
Distance: How far has the current representation moved from the reality it is intended to represent?
Autonomy: How long can the representation–action–reality loop continue without external initiation?

RRF therefore describes two competing forces:

Recursion pulls the flow deeper into representations.
Action and observation reconnect the flow with reality.

A well-functioning recursive system requires both.

The central concern is not recursion or autonomy themselves. It is autonomous recursion that loses touch with reality while retaining the ability to impact reality.