Many evolutionary multi-objective optimization (EMOs) methodologies have been proposed and shown a great potential in approximating the entire Pareto front. While in real world, what decision makers (DMs) want is one or several solutions to satisfy their requirements. It has become a hot problem that dynamically using preference information provided by DMs during the optimization process guides the search of EMO algorithms. An interactive reference region-based evolutionary algorithm through decomposition is proposed, denoted as RR-MOEA/D in this paper, which focuses the search on the desire of DMs to save computational resources. MOEA/D, as a well-known multi-objective optimization algorithm, is used as a basic framework here. In MOEA/D, by dealing with the sub-problems in the preference region and ignoring uninterested ones, the solutions obtained can converge to the regions which the DM prefers on the Pareto front and the computational complexity can be saved to a great extent. At each interaction, a humanized and simple interactive condition is adopted so that the reference region can be changed in a very intuitive way if the DM is unsatisfied the results in the interactive process. A rapid interaction is designed and a set of rough solutions can be obtained quickly whenever the preference information is changed. The proposed algorithm is tested on several benchmark problems and the experimental results show that the proposed algorithm can take full use of preference information and successfully converge to the reference region due to its reasonable and simple interaction mechanism.