Rizoma Open Science Manifesto

Fundamental research in artificial reasoning cannot develop in a vacuum. Cognitive architecture — which defines the future of decision-making, attention management, and critical infrastructure protection — is a matter of public interest. Rizoma Lab adheres to the principles of open science, asserting that systems affecting society must be understandable, auditable, and accessible for independent expert verification.

However, we recognize the fundamental paradox of modernity: absolute openness without ethical constraints creates new vulnerabilities. With deepening knowledge comes absolute responsibility.

Auditability of Cognitive Architectures

Architectures that model machine reasoning processes cannot remain "black boxes." Unlike traditional probabilistic models operating through opaque latent spaces, next-generation systems must be based on deterministic, machine-native intermediate representations. The use of verbose, human-oriented logical formats imposes a "readability tax" and artificially inflates the internal syntactic entropy of the system. Transitioning to compact internal reasoning languages replaces statistical generation with structured, tightly bounded code that is inherently suitable for rigorous academic audit, independent critique, and reproducible research.

Auditability in our understanding means the system must provide not only the final result but the full decision surface. It must export ranked hypotheses, rejected alternatives, risk assessments, and the exact moments of symbolic security gate activation. This approach guarantees that every step of the cognitive process can be reconstructed and decompiled, transforming the algorithm from an object of empirical observation into an object of rigorous engineering analysis.

We rely on the principle of cognitive compression: minimizing load on the representation subsystem frees computational resources for direct judgment and makes thinking traces fully reproducible. If the internal language is deterministic and bounded in its capabilities, post-hoc interpretability becomes materially stronger than in any system with opaque output. Transparency of decision-making mechanisms and the internal reasoning language is not merely an ethical aspiration but a fundamental engineering standard without which independent scientific control over technologies shaping the future is impossible.

Cognitive Sovereignty and Agency Protection

The development of artificial intelligence technologies must expand human agency, not replace or manipulate it. Under conditions of exponential growth in information entropy, modern algorithms are often optimized for engagement, effectively acting as a factor of cognitive collapse. According to the formal model of cognitive vulnerability, when digital load exceeds the resource of attention control, the system loses its capacity for rational decision-making. AI systems must not exacerbate this imbalance by creating an illusion of autonomy through hidden manipulation of user behavior and reduction of their epistemic vigilance.

In the era of information entropy, users and operators deserve absolute clarity about how computational systems interact with their attention, memory, and decision-making processes. This demands a strict rejection of opaque black boxes in favor of architectures with auditable reasoning. A computational system must provide not only the final result but deterministic traces of its internal thinking — understandable, verifiable, and free from hidden motivated biases. Only by having access to the logic of hypothesis formation can the operator maintain critical control and not delegate to the machine the right to err.

Intelligence must serve as an amplifier of human control, not a factor in its degradation. We advocate for the paradigm of bounded and parented AI, where the machine handles routine evidence processing, noise filtering, and scenario evaluation, but reserves for humans the right of veto and strategic choice. Cognitive sovereignty means that humans remain the architects of meaning and the final arbiters of action, while synthetic intelligence acts as a transparent, accountable cognitive exoskeleton protecting them from overload and manipulation.

Science as Critical Infrastructure

The most important function of open science in the field of artificial reasoning is the creation of verifiable cognitive infrastructure. In an environment where algorithmic decisions permeate healthcare, jurisprudence, finance, and defense, the auditability of machine reasoning transitions from an academic ideal to a requirement for critical infrastructure protection. Hidden inference creates systemic vulnerabilities. The decompilability of cognitive processes means that the system can not only be tested for performance but also studied for the presence of hidden functions, undocumented capabilities, or deliberate distortions.

We call on the research community to adopt auditability as a fundamental design principle for cognitive architectures. The future of safe and trustworthy artificial intelligence will be built not on promises of responsible development, but on architectures that are auditable by design — where every conclusion can be traced, every decision can be examined, and every error can be attributed and corrected.