Article proposes causal software engineering framework

The author argues that data-driven software engineering tools relying on machine learning and large language models are limited by their reliance on correlation, making them prone to spurious results and untrustworthy in critical domains. To overcome this, the article introduces causal software engineering, which combines explicit causal models with data-driven inference to enable what-if analysis, counterfactual reasoning, and confounder-free decisions. The proposed conceptual framework integrates human domain knowledge directly into the reasoning process to improve the transparency and rigor of machine assistance. Open article →