Causal Inference in Photopolymer Process Development: Interventions, Confounding, Transportability and Mechanistic Claims

Photopolymer process datasets frequently contain strong correlations among formulation, machine, exposure, geometry and final properties. Correlation can support prediction but does not by itself establish a causal mechanism. This review translates causal-inference concepts into photopolymer process development. It distinguishes interventions from observations, identifies common confounders and mediators, and explains why feature importance, regression coefficients and predictive accuracy cannot automatically be interpreted as causal effects. A process-structure-property graph is proposed for formulation, optical dose, thermal history, conversion, network state, geometry and measured performance. The framework requires causal assumptions to be explicit, tests transportability when moving between printers or material states, and treats mechanistic claims as stronger than predictive claims. Published materials and additive-manufacturing work demonstrates growing use of causal and physics-guided analytics, but photopolymer-specific causal identification remains an evidence gap requiring designed interventions. ER-403.