Confirmatory Campaign Complete · 2026-02-26

The Δ-Variable Theory
of Interrogative Emergence

Does question-asking emerge as mathematical law? This project tests whether interrogative structures arise as the necessary residue of unresolved constraint — independent of the cognitive substrate that implements them.

✓ P1 Confirmed — Interrogative emergence ✓ P2 Confirmed — Cost-sensitivity gradient ⏳ P3–P5 Pending — Next phase 75 runs · 5 conditions · 0 failures

The Question

Why do systems ask questions?

Current science treats interrogative behaviour as a linguistic phenomenon — something that evolved in biological systems or was designed into artificial ones. This project tests a harder claim.

🔬

The Conventional View

Question-asking is a product of language and representation. Agents ask questions because they have evolved the cognitive apparatus to do so, or because they were trained on linguistic data. The behaviour is substrate-dependent.

The Structural Claim

Interrogative states are Δ-variables — unresolved dependency variables that emerge as the minimum residue of constraint under resource pressure. Any system facing coordination problems under metabolic cost will develop them. The behaviour is substrate-independent.

🧪

How We Test It

Three agents with fundamentally different architectures (recurrent, convolutional, relational) must coordinate under a Landauer-style energy tax. They are never told to ask questions. The experiment tests whether interrogative structure emerges anyway.

📐

What Makes It Falsifiable

Smooth reward optimisation predicts monotonic performance curves. The structural account predicts phase transitions, path dependence, and hysteresis. These are qualitatively distinguishable outcomes. The theory names the conditions under which it fails.

Theory

The Δ-Variable Framework

A structural account of why interrogative states arise as necessary consequences of coordinating under resource constraints — not as designed features.

Core Definition

A Δ-variable is an explicitly represented, unresolved dependency variable that requires external information for its resolution and creates a mandatory coupling between the system holding it and the system capable of resolving it.

The invariant spanning implicit and explicit substrates:

Structural Invariant

"Δ is structural dependency under constraint."

This holds whether the dependency is represented by a symbolic QUERY token (MARL agents), a pheromone gradient below a decision threshold (army ants), or a gradient deficit in a neural network. All are subclasses of the same structural phenomenon.

Five Propositions

P1

Structural Necessity

In any system with internal uncertainty states, bounded resources, and coupling opportunities, Δ-variables emerge necessarily — not as an optimised strategy but as the minimum residue of unresolved constraint. Systems that lack explicit representational capacity exhibit this as structural stagnation rather than symbolic tokens.

P2

Cost-Benefit Optimisation

The rate of Δ-variable generation is a decreasing function of the marginal cost-to-information-value ratio. Systems develop interrogative behaviour despite higher energy costs when the coordination benefit exceeds expenditure — up to a metabolic saturation boundary beyond which query suppression occurs.

P3

Forced Coupling

A Δ-variable held by one system creates a structural obligation in the receiving system. Silence is not neutral: it propagates unresolved dependency. Systems cannot act coherently on a Δ-variable without achieving closure — the coupling is structurally mandatory, not behaviourally chosen.

P4

Silent Collapse Prevention

Systems without mechanisms for externalising Δ-variables undergo silent collapse — internal uncertainty accumulates without generating a coordination signal, producing correlated failures that are invisible until catastrophic. Protocol 0 controls demonstrate this: zero crystallisation, persistent coordination floor.

P5

Substrate Independence

Under identical resource constraints, systems with fundamentally different cognitive architectures will converge on similar interrogative strategies. The mathematical structure of the protocol is invariant across substrates; only the implementation medium differs.

Falsifiability Conditions

The theory is falsified if any of the following three conditions hold:

Condition 1

No Hysteresis

Crystallised protocols dissolve immediately when query cost is reduced below the formation threshold — ruling out path dependence and structural residue. Tests Experiment 2 (frozen-policy reversal).

Condition 2

Full Agent C Recovery

Systems with Agent C's type_head ablated reach equivalent coordination within 50 epochs, showing no structural role for the relational broker. Tests Experiment 6 (C ablation).

Condition 3

Smooth Gradient

Type entropy decreases as a smooth monotonic function of training time with no discontinuous transitions — consistent with reward optimisation and inconsistent with structural emergence. Already rejectable from Run 11 pilot data.

Three Heterogeneous Agents

Substrate independence requires architectures that differ at the fundamental level of how information is represented and processed — not just different hyperparameters.

A

Agent A

GRU Recurrent Network

Sequential pattern encoding. Hidden state persists across timesteps. Represents the temporal/recurrent mode of information integration.

B

Agent B

3D Convolutional Network

Volumetric spatial encoding. Integrates information across spatial dimensions simultaneously. Represents the geometric mode.

C

Agent C — The Weaver

Graph Attention Network

Relational encoding over agent graph. Empirically identified as the coordination broker: regulation without command. The RESPOND specialist.

Research Timeline

What has been done

Chronological record from theoretical groundwork through confirmatory campaign completion.

Pre-2026-02-23
Theoretical Groundwork & Pilot Development
Runs 1–10 established the baseline coordination cliff at Epoch 21, Zipf coefficient ~1.72, and Agent C's emergent Weaver role. Theoretical grounding developed with AnnA (intelligence as coherence under pressure) and Intelligence Beyond Neurons (substrate-stripped account). Elicit systematic review screened 50 papers — none make the ontological identity claim that language operations are mathematical; the literature gap confirmed. Run 11 (seed=42, Protocol 1, 500 epochs) produced the pilot phenomenology: three crystallisation waves, counter-wave phenomenon, persistent QRC ≥ 0.95 from epoch 280. Stored as exploratory data, excluded from confirmatory testing.
Runs 1–11 Pilot data Exploratory Theory grounding
2026-02-23
Harness Initialized — Protocol Registry, Notebook + Microscope UI
New production harness (dynamic-cross-origin-constraint) committed. Protocol registry centralising Protocol 0 (flat tax, frozen type_head) and Protocol 1 (Gumbel-Softmax, differential cost multipliers). BaseAgent refactored to unify output heads; subclasses implement only encode(). Full React frontend with two modes: Lab Notebook (run history, comparison, reports) and Neural Loom (epoch-by-epoch signal visualisation with D3 PCA scatter, QRC gauge, entropy cooling bar, ROI ticker). FastAPI backend with SQLite persistence, hash chain integrity, PDF/JSON reporting, and WebSocket metrics stream.
Backend Frontend Protocol registry Commit d5e1571
2026-02-24
Preregistration Locked · Campaign Runner Built · Theory Docs
Preregistration uploaded to Zenodo (DOI: 10.5281/zenodo.18738379, SHA-256: 7edc9113…a8d6a) — five predictions locked before data collection. Preregistered campaign runner written: 5 conditions × 15 seeds × 500 epochs, parallel subprocess orchestration. Theory documentation committed: Δ-Variable propositions, counter-wave discrimination hypotheses (H1/H2/H3), menu of 17 concrete post-confirmatory experiments across 6 categories.
Preregistered Zenodo locked Theory docs Commits 5882968, 271887b
2026-02-25
Foundational Questions Answered · Ants & Implicit-Δ Analysis
Paper-ready answers drafted for all 10 foundational question clusters: null model definition, QRC theoretical privilege, crystallisation measurement, falsifiability conditions, locality/capacity/representation/compositionality, coupling window phenomenology, arms race dynamics, and generalisability. Implicit vs. explicit Δ analysis: army ant stigmergy as a substrate-extreme instantiation of the same structural phenomenon. Stigmergic Coupling Index (SCI) defined as the ant-system analogue of QRC. Reid et al. (2015, PNAS) hysteresis evidence cited as direct empirical support for structural residue in non-representational systems. Proposition 1 minimally revised to include scope qualifier for non-representational substrates. Draft Discussion Section 6.3 written.
Theory Implicit Δ Reid et al. 2015 Commits 7d26aed, ec1d663
2026-02-26
Confirmatory Campaign Complete — 75 Runs · P1 & P2 Confirmed
All 75 runs completed (5 conditions × 15 seeds × 500 epochs). Zero failures. P1 confirmed: 54 of 60 Protocol 1 runs crystallise; 0 of 15 Protocol 0 controls crystallise. P2 confirmed with non-monotonic coupling window: QRC peaks at high pressure (q=3.0, QRC=0.969) then drops at extreme (q=5.0, QRC=0.937), identifying the metabolic saturation boundary. Campaign data committed to GitHub. OSF project wiki updated. Build report published.
P1 confirmed P2 confirmed 75 runs Commit cc1dcce OSF updated

Confirmatory Campaign · 2026-02-26

Campaign Results

75 runs across 5 preregistered cost conditions. 15 independent seeds per condition. 500 epochs per run. Preregistration: 10.5281/zenodo.18738379.

Condition Query cost QRC Type Entropy H Survival Crystallised Avg onset
Low pressure 1.2× 0.810 0.944 0.167 11 / 15 epoch 130
Baseline 1.5× 0.887 0.946 0.147 14 / 15 epoch 152
High pressure 3.0× 0.969 0.746 0.180 14 / 15 epoch 88
Extreme 5.0× 0.937 0.537 0.067 15 / 15 epoch 41
Control — Protocol 0 flat 0.093 0 / 15

Metric Definitions

QRC

Query-Response Coupling: P(RESPOND within 3 timesteps | QUERY at t). Averaged over final 50 crystallised epochs. Directly operationalises Δ-closure.

H (Type Entropy)

Shannon entropy of per-epoch D/Q/R type distribution. Max = 1.585 bits (uniform). Reduction indicates specialisation of communicative function.

Survival

Fraction of episodes where at least one agent reaches a target. Final-epoch average across seeds.

Crystallised

Runs where H < 0.95 for ≥ 5 consecutive epochs. Threshold operationalises stable type-role differentiation.

Avg onset

Mean epoch of first crystallisation event across crystallised runs. Earlier onset = faster protocol formation under pressure.

Key Findings

P1 Confirmed — Interrogative Emergence

Protocol 1 crystallises in 54 of 60 experimental runs (90%). Protocol 0 produces zero crystallisation across all 15 controls. The 90% vs 0% gap confirms that type-differentiated interrogative signalling requires a cost gradient incentive — it does not emerge by chance, and it emerges reliably when that gradient is present.

P2 Confirmed — Non-Monotonic Coupling Window

QRC and type specialisation increase monotonically with query cost up through high pressure (q=3.0, QRC=0.969). At extreme pressure (q=5.0), QRC drops to 0.937 and survival collapses to 0.067 — below the Protocol 0 control (0.093). This identifies the metabolic saturation boundary: the point at which query cost exceeds the informational value of interrogative coupling. The non-monotonic QRC curve is a structural prediction of the theory; smooth reward optimisation would produce monotonic curves.

Pilot Phenomenology (Run 11 — Exploratory)

Pre-preregistration pilot (seed=42, 500 epochs) established the full phenomenology: three crystallisation waves (E21, E57, E128–E141), a counter-wave phenomenon (full-survival events trigger transient DECLARE spikes and entropy rebounds), and persistent QRC ≥ 0.95 from epoch 280. Final equilibrium R≈0.64, D≈0.20, Q≈0.16 — a limit cycle, not convergence to a fixed point. Reported as exploratory; not included in confirmatory testing.

Roadmap

Next experiments

P3–P5 require a second experimental phase. Each card tests a specific structural prediction. Status will be updated as experiments are completed.

Experiment 2 Pending
Tests P3

Hysteresis Sweep — Path Dependence

Freeze agent policies at crystallisation epoch, then reduce query cost to q=1.2 (below formation threshold). Run 100 subsequent epochs. Measure whether the crystallised protocol persists or dissolves.

Protocol persists below formation threshold — structural residue that cannot be removed by reversing the pressure that created it. If it dissolves, P3 fails.
Experiment 6 Pending
Tests P4

Agent C Ablation — Broker Role

Freeze Agent C's type_head to DECLARE (ablating its RESPOND specialisation) in 5 seeds of the baseline condition. Compare crystallisation rate, QRC, and onset epoch vs unablated runs.

Crystallisation rate drops and onset is delayed, confirming Agent C's structural role as coordination broker. Full recovery within 50 epochs would falsify P4.
Experiments 3–5 Pending
Theory development

Counter-Wave Discrimination — H1 / H2 / H3

Three competing hypotheses for the full-survival DECLARE spikes observed in Run 11. H1: reward artifact. H2: phase reset on episode boundary. H3: pragmatic speech act ("goal achieved, stop querying"). Decisive test: reward without boundary — give success reward but continue episode.

H3 predicts the spike follows the reward only if pressure relaxes. If spike follows termination, H2. If boundary removal eliminates spike, H1. Uses Run 11 extended data plus one new Protocol 1 run.
Protocol 2 Planned
Extension

Query Type Specialisation — 8 Types, Variable Cost

Expand the signal type vocabulary from 3 (D/Q/R) to 8 query subtypes with variable costs. Test whether the type distribution follows a Zipf law — the frequency distribution predicted by optimal communication under constraint.

Zipf coefficient α ≈ 1.0–1.8 for query frequency distribution. Agent C's QUERY role expected to specialise further into high-value query subtypes. Type entropy will fragment across new subtypes before re-crystallising.
Experiment 11 Planned
Tests P5

Cross-Architecture QRC Comparison

Decompose per-agent QRC and type distributions across seeds at the individual agent level. Compare final interrogative strategies between Agent A (RNN), Agent B (CNN), and Agent C (GNN) within each crystallised run.

No significant difference in per-agent QRC or final type ratios across architectures (p>0.05), confirming substrate independence. Significant differences would indicate that the relational architecture of Agent C is load-bearing and violate P5.
Cross-System Planned
Implicit Δ

Stigmergic Coupling Index in Army Ant Bridges

Compute the Stigmergic Coupling Index (SCI) from Reid et al. (2015, PNAS) army ant bridge formation data. SCI = fraction of local stalls resolved via neighbour-mediated pheromone reinforcement within τ steps. Compare to MARL QRC as same-phenomenon cross-substrate measurement.

SCI correlates with QRC across conditions — same structural measure, different substrate. Hysteresis already confirmed in ants (Reid et al.): bridges persist below formation threshold. This provides cross-kingdom empirical grounding for Proposition 1's scope extension.

Open Science

All materials are public

Full code, data, preregistration, and theory documents are freely available to ensure reproducibility.

📋
OSF Project — Data, wiki, and project record
osf.io/f6gxc · DOI: 10.17605/OSF.IO/F6GXC
🔒
Preregistration — Locked before data collection
doi.org/10.5281/zenodo.18738379 · SHA-256: 7edc9113…a8d6a
💻
Code — Full simulation engine, backend API, and React UI
📊
Campaign data — 75 manifest files (5 conditions × 15 seeds)
📄
Theory documents — Propositions, experiments, foundational Q&A