COGNITIVE CONTROL / REPRESENTATION / DYNAMICS

What holds a goal in mind?

A goal is built from the same representations as everything around it.
Its power comes from how those representations are held together.

One representational system.
Many possible organizations.

CHOOSE WHAT GOVERNSSame evidence. Different priorities. 123

The held graph & its landscape

STRUCTURERepresentations & relationshipsTop-down implementation ↓
DYNAMICSA goal-conditioned landscapeIllustrative
Abstract representational spaceDrag to rotate
MAINTAINED GOAL

Be accurate

The same evidence now favors verification: inspect the evidence, check assumptions, and compare explanations.

Protect the selected goal from contextual capture.

Increase the pull of a competing priority.

Traverse the hierarchy

Goals constrain subgoals and feature selection.

Goal held; relevant paths remain influential.
FOLLOW THE ORGANIZATION

A route through shared representations

Changing the goal changes the relevance of the evidence, the subgoals that receive priority, and the trajectory through the landscape.

I

The harness is a relationship

Maintenance, selection, and causal influence give a representation its controlling role. The boundary marks functional protection from interference.

II

Composed, then recomposed

Goals, subgoals, and context draw on shared features. Goal changes alter their weighting and coupling while their identities stay in place.

III

Two directions, distinct roles

Downward traversal implements a goal. Upward traversal uses evidence to evaluate or revise it. This does not reverse physical causation.

ABOUT THE MODEL

A maintained organization
of shared representations.

This visualization explores the proposal that goals are hierarchical relational structures: composed of shared representations, sustained as control states, and capable of reorganizing ongoing processing.

What the graph represents

Goals connect to subgoals through means–end relations; subgoals connect to features through compositional relations. Highlighted routes indicate control modulation. These relation types have different meanings; not every link is a physical causal connection.

What the boundary represents

The “harness” is a functionally maintained pattern of influence. Its boundary illustrates resistance to irrelevant input. It does not imply a separate anatomical store or an additional inner agent.

What changes the landscape

Each goal defines an illustrative potential well. The selected goal, maintenance, interference, and background assumptions change the relative depth of the wells. Particles follow a damped gradient with small fluctuations. The landscape uses abstract coordinates and arbitrary units.

How priors enter

“Stable causal relationships” preserves relatively consistent dependencies. “Context-sensitive relationships” changes the relative usefulness of verification and alternative search. These are two illustrative world-model assumptions; no probability distribution over metaphysical beliefs is being estimated.

Explanation and introspection

The route below the graph makes the model’s selected dependencies inspectable. Human verbal explanations need not faithfully reveal the processes that produced them; graph access alone would not establish introspective accuracy.

Scope of the proposal

The shallow three-tier graph, priority weights, potential function, and gate are design choices for exploring this account. They make assumptions visible and manipulable. They do not establish a complete theory of cognition or a neural implementation.