A causal AI and explainable optimization framework for social robot design

Mapping human-machine requirements in smart product design remains challenging. An integrated framework combining semiotic architecture product design (SAPAD), dual machine learning (DML), hesitant fuzzy quality function deployment (HFQFD), and multi-objective optimization is proposed. Home companion robots are used as a case study. User behavior is deconstructed via SAPAD to generate demand hypotheses. Causal inference is performed by DML to identify core demands impacting user experience, with priorities defined by integrating expert-assigned and causal weights. Evaluation uncertainty is addressed by HFQFD, through which validated demands are converted into design parameters. A hybrid approach integrating Cuckoo Catfish Optimizer (CCO), least squares support vector machine (LSSVM), and non-dominated sorting genetic algorithm II (NSGA-II) is developed for multi-objective optimization. Shapley Additive exPlanations (SHAP) analysis is employed to quantify each parameter's marginal contribution to conflicting requirements. A closed-loop process from behavioral observation to parameter optimization delivers transparent decision support for complex human-machine system design.