AI for software engineering: from probable to provable

2025-11-28

AI Summary

AI programming faces two fundamental challenges: (1) the difficulty of precisely formalizing requirements—prompt engineering is, in essence, high-effort requirements engineering—and (2) model hallucination, which compromises the reliability of generated code. This paper proposes a novel paradigm that synergistically integrates AI’s creative capacity with formal methods, shifting from probabilistic code generation to *provably correct* program synthesis. Our core contribution is a *specification-driven AI programming framework*: it takes machine-readable formal specifications (e.g., in TLA⁺ or Coq) as input constraints and leverages lightweight program verification alongside interactive theorem proving to guide large language models in generating and iteratively refining code within strict semantic boundaries. Experimental evaluation demonstrates substantial reductions in hallucination rates and significant improvements in functional correctness and verifiability of synthesized code—establishing a trustworthy pathway for AI-assisted development in safety-critical systems.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Reasoning under Uncertainty: Probabilistic Programming

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Vibe coding, the much-touted use of AI techniques for programming, faces two overwhelming obstacles: the difficulty of specifying goals ("prompt engineering" is a form of requirements engineering, one of the toughest disciplines of software engineering); and the hallucination phenomenon. Programs are only useful if they are correct or very close to correct. The solution? Combine the creativity of artificial intelligence with the rigor of formal specification methods and the power of formal program verification, supported by modern proof tools.

Problem

• Specifying goals for AI programming is difficult
• AI-generated programs suffer from hallucination issues
• Programs must be correct or nearly correct to be useful

Innovation

• Combining AI creativity with formal specification methods
• Integrating formal program verification for correctness
• Utilizing modern proof tools to ensure program accuracy