Sources and representational access

2 Sources and representational access

2.1 Evidence, inferential reach and achieved understanding

Available evidence and achieved understanding are distinct. A system can retain premises that logically imply a conclusion while lacking a procedure that reaches it under the available time, memory and access constraints. A newly learned ordering, abstraction or search policy can make that conclusion accessible. This is a change in organisation even when the original evidence is unchanged. The information carried by the source, the information available through a particular interface, the consequences derivable in principle, and the consequences actually derived are four different objects.

An assertion that two participants have the same knowledge must therefore specify what is equal. Equal initial propositions do not imply equal observations during interaction. Equal message strings do not imply equal access to counterfactual queries. Equal unrestricted deductive closure does not imply equal bounded performance. Conversely, an interaction may supply genuinely new evidence as well as improve its organisation. The developmental claim is clearest when these possibilities are measured separately.

Experience can affect the organisation and use of knowledge without making truth subjective. A person may learn to notice a dependency, recognize an exception, or deploy an already known rule at the appropriate time. Such changes can improve contact with a common source. They can also produce overgeneralization, rigidity or misplaced salience. An emotionally compelling explanation may be more likely to be retained while remaining false. Salience is a causal property of processing; evidential warrant is an epistemic assessment. Neither should be used as a substitute for the other.

2.2 Source, aperture, readout and constructed representation

Let $\Omega_t$ name a nominated source state, $\Delta_t$ the available access relation, and $r_t=\Delta_t(\Omega_t)$ an accessible readout. These are model variables, not a representation of the unsplit ontological prior. A representation $\widehat\Omega_t$ is constructed from readout, retained organisation and processing. It may preserve a genuine source relation without being identical to the source or recovering it exhaustively.

Mediation does not entail distortion. A thermometer can accurately report a temperature while omitting almost everything else about the measured object. A proof can preserve a logical relation without reproducing the physical process by which someone discovered it. Equally, a readout can be ambiguous, noisy, selectively sampled or interpreted under a mistaken model. The appropriate question is which source properties are identified by the admitted evidence.

P2 provides a precise inherited answer. Relative to an experiment grammar, a source attribute is identifiable exactly when it is constant over models with the same admitted experiment laws. Nonidentifiability of the whole source does not imply nonidentifiability of every attribute. This supports a position between exhaustive transparency and global skepticism: warranted knowledge is possible at a stated resolution and under stated access assumptions.

The source–readout distinction also clarifies agency. Some activities change the representation while leaving the source and aperture fixed. Others change the aperture through a question, instrument, perspective or intervention. Still others change the source conditions themselves. A scaffold-derived design can be implemented in a tool or institution, modifying the world that later participants encounter. The next readout is consequently shaped partly by earlier representations that have become causes. The feedback need not be benign: a misleading classification can reorganize an institution so that its later data partly reflect the classification’s effects.

2.3 Perspective, awareness and metacognition

Primitive awareness, phenomenal organisation, a localized perspective, a personal model and metacognition should not be collapsed. In the adopted constitution, awareness is an ontological commitment; phenomenal organisation is assigned to a qualifying native organisation; a personal model represents an individual’s attributes and history; metacognition concerns monitoring or regulating cognitive activity. A system can improve metacognitive performance without the present analysis establishing a new subject. A subject can have experience without satisfying an added requirement of reflective self-description.

A finite perspective has incomplete access and a particular organisation of relevance. This is operational closure relative to an interface and a current state, not a proof of absolute metaphysical isolation. Interaction can change what is available through that interface. A model of one’s own processing can become a new object of reasoning, but further descriptive levels are not automatically deeper consciousness. They may increase control, incur overhead, or generate another inaccurate representation.

2.4 Functional perspective and internal arbitration

A functional perspective can be modelled as the organisation through which competing distinctions become consequential for one continuing process. It need not have a spatial centre. Its actions can include allocating attention, admitting a memory, revising a prediction or running an internal simulation. Restricting control to overt movement would omit these cases. A distributed controller may coordinate them without a single representation receiving every signal.

One common output does not prove a global interpretive bottleneck. Independent modules can concatenate their answers, and a nonseparable output can be a simple common descendant without a rich perspective. A meaningful arbitration claim must identify conflicts, admissible alternatives, the mediator that resolves or preserves them, and the future consequences. Defining a bottleneck as whatever already performs the desired integration would make the necessity argument circular.

Native qualification, developmental plasticity and self-model centrality are independent axes. A qualifying core may have rigid dynamics; a flexible externally serviced learner may lack a certified internal return; a personal model may lose privilege while native qualification persists. The failure of a universal derivation of recurrence does not refute this richer functional architecture. It locates it as a proposed organisation within the broader constitution.

Unpresented processing can also shape a scene. Foreground attention and explicit self-description do not exhaust A2’s complete endogenous predictive object. Whether that object includes too much apparently unconscious processing is a substantive challenge to A2, not something to resolve by silently retaining only attended content.