Recent research using machine decision makers has revealed that some leading interactive evolutionary multi-objective optimization algorithms do not perform robustly with respect to interactions with preference models (and biases) posited to be representative of human Decision Makers (DMs). In order to model preferences better, we propose an explainable interactive method that uses decision trees to automate (fast) pairwise comparisons based on trade-offs of two given solutions. To cancel out possible biases and errors in estimations, we use the trained tree in holistic comparisons to determine solutions that survive each generation. We test our new method with respect to two different preference models (Tchebychef and Sigmoid) on problems from 2 to 10 objectives, and control both the number of interactions available and various biases. The results suggest the superiority of our method in learning the DM’spreferences and in terms of the utility value of the final solution returned by the algorithm compared with some well-known interactive methods.Keywords: Interactive Evolutionary Multi-Objective Optimization · Decision Tree · Machine Decision