🤖 AI Summary
Traditional programming models lack structured mechanisms for governing side effects, making it difficult to reliably oversee the behavior of autonomous systems. This work proposes an intent-driven computing model in which programs generate only declarative “intents” describing proposed actions. These intents are then vetted by a managed runtime against a decidable policy language and, upon approval, recorded in a tamper-evident ledger before execution. By shifting governance from the undecidable domain of program semantics to the decidable domain of intent data, the approach enables event sourcing, intent replay, structured auditing, and high human interpretability. An implementation atop the BEAM virtual machine integrates Rocq formal verification—encompassing 454 theorems—and over 70,000 property-based tests, achieving full verification of the specification’s null hypothesis lemma and strong alignment between implementation and formal specification, thereby establishing the first programming model for autonomous systems with built-in governance capabilities.
📝 Abstract
Programming languages assume programs directly execute effects. When autonomous systems generate behavior dynamically, this assumption becomes problematic: there is no structural mediation point between deciding to act and acting. We define intent-driven computing: a programming model where programs produce intents (finite data values describing proposed actions) rather than directly executing effects. A governed runtime examines each intent against a decidable policy language, records every decision in a tamper-evident ledger, and only then realizes the effect. The language provides no alternative path to effects. The model does not decide arbitrary behavioral properties of programs (which Rice's theorem shows is impossible). Instead, it constrains the language so that all effectful interaction is reified as finite intent values, shifting governance from the undecidable domain of program semantics to the decidable domain of intent data. This yields emergent properties: event sourcing by construction, governance simulation via intent replay, structural audit completeness, and improved human comprehensibility. We specify the model formally, implement it in a concrete language compiling to the BEAM virtual machine, and verify key properties in Rocq (454 theorems, 36 modules, zero admitted lemmas). Property-based testing (70,000+ random inputs, zero disagreements) validates that the implementation matches the specification.