Can Neural Networks Learn by Experimenting on Themselves? Self-Interventional Learning from Functional Consequences to Predictive Self-Knowledge

Michał Tomaszewski

cs.LGAug 14, 2026

Can Neural Networks Learn by Experimenting on Themselves? Self-Interventional Learning from Functional Consequences to Predictive Self-Knowledge

Authors: Michał Tomaszewski

Organizations: Department of Artificial Intelligence Faculty of Computer Science Opole University of Technology Prószkowska 76, 45-758 Opole, Poland

Abstract

Machine-learning systems usually model external data, while their internal functional organization is analyzed by external observers. This work introduces Self-Interventional Learning (SIL), in which a neural system perturbs its own functional structure, observes consequences, learns a predictive self-model, generalizes to unexecuted interventions, and uses predictions to guide later structural action. In a construction-known synthetic system, SIL recovered critical structure, redundancy, and replaceability, while synergy was not reliably recovered. Across 30 fresh confirmatory seeds, increasing the pairwise intervention budget from 4 to 56 reduced held-out prediction error from 0.0335 to 0.0148 and increased Spearman correlation from 0.629 to 0.883. In a matched ablation, preserving the correct intervention--consequence mapping reduced prospective prediction error by 81.3%, while using the same learned self-model for action reduced normalized regret by 31.7% relative to ignoring it. However, model-guided action did not significantly outperform a direct empirical-memory policy, and powered CIFAR-10/ResNet validation showed no robustness advantage over equal-budget direct repair search. These results support SIL as an intervention-driven framework for learning predictive knowledge about a network's own functional organization, while showing that the self-model remains incomplete and is not universally superior to simpler direct strategies.

Figures & tables

Block	Primary scientific question	Confirmatory unit	Primary family	Final decision
Applied Kintsugi/CPPV	Does predictive self-interventional repair knowledge improve structural adaptation, and does that advantage transfer externally?	Fashion $N=35$ replication; CIFAR $N=82$ powered external	Frozen paired endpoint families	Fashion advantage replicated; CIFAR Learned $\approx$ Direct performance boundary; predictive mechanism transfers
Experiment A	Can a learner recover construction-known but hidden functional structure and predict interventions it did not execute?	20 fresh seeds	A-H1–A-H4; Holm FWER .05	3/4 primary PASS; synergy recovery FAIL
Experiment B	Does predictive self-knowledge improve systematically with increasing self-interventional experience?	30 fresh seeds	B-H1–B-H2; Holm FWER .05	2/2 primary PASS
Experiment C	Are intervention–consequence semantics necessary for prediction, and does learned self-knowledge improve action when used?	30 fresh seeds	C-H1–C-H3; Holm FWER .05	2/3 primary PASS; Full SIL $>$ direct memory not supported

Table 1: Final experimental program. Pilot studies were used only for technical promotion; inferential conclusions below refer to fresh confirmatory seeds.

ID	Contrast	$N$	Mean effect	Bootstrap 95% CI	Holm $p$	Decision
A-H1	Critical AUROC $>0.50$	20	$+0.5000$	$[0.5000,0.5000]$	$1.16\times 10^{-5}$	PASS
A-H2	Redundancy AP: pairwise SIL $>$ Direct	20	$+0.53020$	$[+0.35401,+0.70735]$	$9.55\times 10^{-5}$	PASS
A-H3	Synergy AP: pairwise SIL $>$ Direct	20	$+0.00558$	$[+0.00040,+0.01267]$	$0.1942$	FAIL
A-H4	Held-out MAE: pairwise SIL $<$ singleton SIL	20	$-0.01415$	$[-0.01633,-0.01166]$	$7.63\times 10^{-6}$	PASS
B-H1	MAE slope vs. $\log_{2}(B)<0$	30	$-0.004841$	$[-0.006571,-0.003265]$	$1.86\times 10^{-8}$	PASS
B-H2	Fisher- $z$ Spearman slope vs. $\log_{2}(B)>0$	30	$+0.151065$	$[+0.113044,+0.190683]$	$2.33\times 10^{-8}$	PASS

Table 2: Preregistered confirmatory hypotheses for Experiments A–C. Effect signs follow the stated contrast; negative values are favorable for MAE/regret hypotheses. All failures are retained in the main paper.

Claim	Status	Evidence boundary
Self-intervention consequences form predictive self-knowledge	Confirmed	Earlier CSM/CPPV plus C-H1: true intervention mapping outperformed a matched permuted control in 30/30 seeds.
Self-knowledge generalizes to unexecuted interventions	Confirmed	Fashion unmeasured repair selection and A-H4 under a strict held-out execution firewall.
Hidden functional organization can be recovered	Partially confirmed	Criticality, redundancy, and replaceability recovered; synergy confirmatory hypothesis failed.
More self-interventional experience improves self-knowledge	Confirmed	Both Experiment B primary trend hypotheses passed Holm correction.
Using self-knowledge can improve action	Confirmed with boundary	C3 reduced regret relative to C2, but C3 did not significantly outperform direct empirical memory C1.
Learned CPPV is universally superior to equal-budget Direct search	Not supported	Fashion success replicated; powered CIFAR-10/ResNet effect approximately zero.

Table 3: Frozen evidence boundaries after completion of the experimental program.

Panel of Figure 1 — Figure 1: Self-Interventional Learning (SIL) and the empirical evidence map. The upper pathway defines the operational mechanism tested in this work. The lower evidence grid summarizes the applied Kintsugi/CPPV origin and the three mechanistic experiments, including their principal negative boundaries rather than depicting SIL as universally superior.

Figure 2: Applied CPPV evidence and the external boundary. Learned CPPV independently outperformed equal-budget Direct CPPV on Fashion-MNIST/ElasticCNN, but the corresponding powered CIFAR-10/ResNet effect was approximately zero. Predictive protection knowledge and extrapolation nevertheless transferred, separating mechanism transfer from performance superiority.

Panel of Figure 3 — Figure 3: Experiment A. Pairwise SIL strongly recovered redundancy and replaceability while synergy remained weak. Direct singleton-additive prediction was extremely accurate on the global held-out set, which was dominated by near-additive pairs, but its error increased sharply on forced interaction-sensitive challenge relations. Error bars are bootstrap 95% confidence intervals over 20 confirmatory seeds.

Panel of Figure 4 — Figure 4: Experiment B learning curves for the preregistered random-nested acquisition policy. Singleton evidence and the held-out evaluation set were fixed while pairwise self-interventional experience increased from $B=4$ to $B=56$ . Prediction error decreased, ranking quality increased, and calibration improved. Error bars are bootstrap 95% confidence intervals over 30 fresh confirmatory seeds.

Panel of Figure 5 — Figure 5: Experiment C. Preserving the intervention–consequence mapping substantially improved prospective prediction under the matched control. Coupling the same learned model to action (C3) reduced regret relative to ignoring it (C2), while Full SIL did not significantly outperform the direct empirical-memory policy (C1). The decomposition shows why: relational self-knowledge was particularly effective for redundancy-sensitive decisions, whereas direct singleton-additive memory remained stronger on matched-general cases and synergy remained difficult. Error bars are bootstrap 95% confidence intervals over 30 confirmatory seeds.

Causal Structure Learning

Causal structure learning infers directed relationships among variables from observational, interventional, or time-series data.

See all 124 papers

From Prediction to Self: Developmental Conditions for Agency in Minimal Neural Systems
Jun 4, 2026Evan YeAgentic AI
Independent Researcher
Predict Before You Modify: Learning Improvement Dynamics for Self-Evolving Agents
Sep 26, 2026Zixuan Xie, Chang Xu, Yuhua Jiang +1LLM Agent Self-Improvement Self-Improving Agents
University of Virginia · Microsoft Research · Tsinghua University
Don't Forget What You Just Learned: Self-Directed Repair Turns Agent Mistakes Into Lasting Lessons
Oct 22, 2026Yuval Kansal, Karen Ullrich, Polina Kirichenko +4Self-Improving Agents LLM Self-Correction
Princeton University · Meta FAIR

Jun 4, 2026J/K move · Enter open · S save

From Prediction to Self: Developmental Conditions for Agency in Minimal Neural Systems

Evan Ye

Independent Researcher

Agentic AI

Open paper AskPDF arXiv

How does a system that merely predicts the world come to distinguish its own causal influence from everything else? We trace this transition in a minimal 192-dimensional GRU through 40 controlled experiments arranged as a developmental sequence, adding components one at a time and tracking whether the system can distinguish self-caused from world-caused changes. The developmental path reveals four conditions that must be satisfied in strict order: (1) persistent state forming stable attractors, (2) a causal action loop linking output to input, (3) proprioceptive feedback that makes implicit causal knowledge explicit, and (4) asynchronous awakening - perceptual learning must consolidate before action learning begins. We propose agency gain (A = Err_world - Err_self), the predictive advantage of knowing one's own action, as a metric to track this process. The self-aware predictor consistently outperforms the self-blind predictor across periodic (sinusoidal) and chaotic (Lorenz) environments, and the metric survives ablation of all auxiliary components. Only forward-sampled action selection produces meaningful agency gain; two gradient-based alternatives degenerate. Equally significant are 12 falsified hypotheses mapping where development stalls: predictive coding alone does not produce self-represent

Abstract

Causal Structure Learning

Explore similar work

From Prediction to Self: Developmental Conditions for Agency in Minimal Neural Systems

Predict Before You Modify: Learning Improvement Dynamics for Self-Evolving Agents

Don't Forget What You Just Learned: Self-Directed Repair Turns Agent Mistakes Into Lasting Lessons