Engineering and causality: A centenary response to Claudius

In July 1932, R. Howard Claudius published a notable meditation on the relationship between engineering practice and causal knowledge. His " Causality from the Point of View of the Engineer " insisted that engineering success depends on knowing how causes produce effects, not merely that they do. A century later, this foundational assumption faces an unprecedented challenge: modern engineers routinely build and deploy machine learning (ML) systems whose internal structure is largely opaque. The tension is clearest with large language models (LLMs), whose capabilities arise from exploiting statistical patterns rather than by encoding causal theory. This centenary examination argues that engineering has bifurcated into two distinct practices: traditional causal engineering that vindicates Claudius's thesis, and a radically new non-causal engineering that succeeds without causal understanding. This transformation forces us to reconceptualize what engineering is and whether causal knowledge remains essential to its practice.