Literature

11 Literature

11.1 Closest conceptual antecedents

Tronick and colleagues’ 1998 account of dyadically expanded states of consciousness is a close antecedent to the central interactional proposal [L1]. The inspected publisher abstract links collaborative interaction and mutual affect regulation to more coherent and complex organisation. This prevents a broad priority claim that interpersonal engagement can expand conscious organisation. The present programme adds a different formal target: retained continuation-relevant organisation, explicit transfer contracts and its conditional relation to A0–A3. The full article was not retrieved, so more detailed comparisons with its developmental evidence remain to be checked.

De Jaegher and Di Paolo’s participatory sense-making treats interaction as a process that can acquire relative autonomy and transform meaning through its interplay with participants [L2]. Its embodied and affective account is directly relevant to relevance-guided development. The present formalism should not reduce that programme to message exchange. Its additional question is which retained distinctions permit later autonomous continuation or transfer under specified resources. Interactional autonomy, effective modularity and a newly assigned joint subject remain distinct targets.

Clark and Chalmers argue that external processes can participate in cognitive organisation when they play the relevant coupled functional role [L3]. This is an antecedent for artifacts, notebooks and environmental organisation contributing to thought. The present artifact account follows that general problem while separating cognitive participation from A1 subject qualification and A3 episode identity. A durable medium can reshape a later conscious organisation without itself being assigned experience.

These precedents do not invalidate an independently reached synthesis. They locate its appropriate claim. The useful contribution is the particular connection among recursive incorporation, bounded reasoning, continuation-preserving transfer and a declared consciousness constitution. The existence of familiar ingredients increases the importance of precise formal and empirical differences.

11.2 Predictive state, value and computational control

Littman, Sutton and Singh represent state through action-conditional predictions of future observations [L4]. This is a direct antecedent for using future tests rather than a hidden-state label as the target of representation. Recasting such predictions as developmental capability does not make predictive sufficiency new. The present diagnostic results ask when a specified finite score panel spans the relevant observable space and how its conditioning controls approximation.

Tzeng’s probabilistic-automaton equivalence work supplies a classical algorithmic antecedent for finite linear methods [L5]. The monograph already contains the exact finite word-span result used here. Only publisher metadata and abstract were inspected for Tzeng; no detailed attribution of a specific uninspected theorem is made. D1–D3 are supplied with self-contained proofs, and their status is deductions developed here rather than an asserted first discovery of observability methods.

Subramanian, Sinha, Seraj and Mahajan develop approximate information states for partially observed systems [L6]. Their framework distinguishes exact sufficient structure from controlled approximation and gives policy-performance bounds. It is a close formal neighbor of the continuation problem. The present target emphasizes declared native transcript laws, retained costs and the difference between task-value preservation and the complete object used by A2. Its all-finite diagnostic bound relies on a fixed exactly known finite family; it does not replace an approximate-model learning theory.

Alver and Precup study minimal value-equivalent partial models for lifelong reinforcement learning [L7]. The inspected proceedings abstract makes the decision-relevant target explicit. This supports a key distinction: an economical model can deliberately omit environmental features irrelevant to its planning criterion. Such omission is not automatically legitimate for a different future learning target or for a full native predictive identification. A claim of minimal developmental state must therefore state what is to be preserved.

Russell and Wefald’s metareasoning programme evaluates computational actions through their effects on decisions and their resource use [L8]. This is a close precedent for relevance-guided allocation of thought and for separating logical consequence from useful achieved inference. It also reinforces a constraint: evaluating the value of a computation is itself a computational problem. A perfect relevance oracle cannot be introduced without cost. Publisher metadata and abstract, an author publication list, and an archived opening page support this comparison; the complete journal article was not inspected.

Two further primary sources sharpen the developmental account. Kretch, Franchak and Adolph compared infants’ visual access during crawling, walking and sitting [L10]. Their head-mounted measurements show that bodily posture changes the available visual field and gaze opportunities. This is evidence for a change in access, not merely a change in the salience of identical input. It cautions against using a common environment as proof of equal sensory evidence. No broader phenomenal conclusion follows from those measurements.

Ellis and colleagues’ DreamCoder develops reusable program abstractions together with a learned search policy [L11]. Its component ablations and reported training costs make it direct prior work for acquired organisation improving later synthesis. The learned library and the procedure recognizing when to use it are distinct contributors. Training generally required roughly a day on 20–100 CPUs in the reported experiments, so the work also illustrates why discovery and runtime costs should be reported separately. This paper does not reproduce its experiments or infer consciousness from their success.

11.3 Causal abstraction and the contribution boundary

Rubenstein and colleagues formalize consistency between structural causal models through exact transformations that respect specified interventions [L9]. This is an antecedent for requiring an intervention translation when a relation becomes a higher-level component. The present P4-informed use is particular: behavioural conjugacy does not preserve the elementary physical interventions used to justify a consciousness realisation. A successful abstraction needs a stated intervention domain and cannot assume that every mathematical intervention is executable.

The checked literature supports restrained positioning. Neither predictive state, embodied meaning, dyadic transformation, bounded metareasoning, external cognition nor intervention-preserving abstraction is being introduced from scratch. The transfer theorems isolate a useful exact coding problem; the diagnostic results add explicit quantitative consequences within the inherited finite-model machinery; the philosophical synthesis states a developmental use of a pre-existing constitution. A wider systematic priority review would be needed before describing any theorem as historically novel.