An Optimization Framework for the Adaptive Design of Robust Choice Questionnaires

We propose a general framework for adaptively designing choice-based conjoint questionnaires at the individual level. This framework uses complexity control to improve the robustness of the conjoint designs to response error and links the informativeness of conjoint questions to the Hessian of the loss function minimized in partworth estimation. It formalizes and generalizes several methods recently proposed both for questionnaire design and estimation. Simulations as well as an online experiment suggest that it outperforms established benchmarks, especially when response error is high.