Park Goun This study diagnoses the recurring simplification and formalization of constructed-response assessments as “alignment collapse” and proposes a set-based task design framework to capture transfer performance as observable evidence. By integrating Evidence-Centered Design (ECD) and Concept-Based Inquiry Learning (CBIL), this study develops a three-layer design model that restores construct-task-evidence alignment and presents a design case of opinion poll critique task sets within the 2022 revised Korean Language Arts curriculum’s “Reading and Writing” course. The study operationalizes transfer as “conditional evidence-based reasoning” and proposes a task structure that accumulates cognitive pathways through text function division (conceptual foundation → contextual application → public norms) and the gradual transfer of responsibility across items. The rubric employs dual principles of perspective/criteria bifurcation and CBIL pathway-based qualitative differentiation, with backward validation procedures to ensure construct fidelity and control for construct-irrelevant variance (CIV). This study reconceptualizes constructed-response assessment as an “alignment-ready design system” rather than a mere item format, providing the theoretical and practical foundations for process-oriented assessment structures that overcome single-item limitations. Future research should pursue empirical validation through student performance data and inter-rater reliability assessment.