From Explainability to System Language: Why AI Accountability Requires a Formal Execution Vocabulary

<div> <p>Current approaches to AI accountability rely predominantly on explainability. By translating system behavior into human-interpretable narratives, explainability is expected to restore trust, enable audits, and support responsibility attribution.</p> <p>This paper argues that this expectation rests on a categorical mistake. Explainability operates at the level of interpretation, while accountability requires control at the level of permission. In probabilistic systems, explanations can rationalize decisions after the fact, but they cannot determine whether a decision was permitted to occur in the first place.</p> <p>The paper introduces the concept of <em>system language</em> as a structural counterpoint to explainability. System language defines machine-readable execution semantics for accountability, including decision boundaries, commit events, non-decision states, provenance, and reliability tiers. These concepts govern when generated language becomes an accountable system action.</p> <p>Rather than proposing a technical implementation, the contribution identifies a foundational requirement for accountable AI: accountability presupposes a formal execution vocabulary that precedes interpretation. Without such semantics, governance mechanisms remain external overlays, unable to enforce responsibility within the decision process itself.</p> <p>The paper positions system language as a prerequisite for auditability, certification, and liability in high-risk AI systems, and argues that explainability alone—however sophisticated—cannot fulfill this role.</p> </div>This paper is part of a series examining accountability, auditability, and operational viability in probabilistic and agentic AI systems. A German-language version is available on Zenodo with&nbsp;<span>DOI: 10.5281/zenodo.18663875</span>