A suitable aesthetic evaluation is an indispensable part of designing structures. However, it remains difficult for most engineers to design structures while reflecting aesthetic considerations due to their subjective nature. In order to reduce the burden of design, it is desirable to develop a technique which can support aesthetic evaluations. The present study provides the artificial neural networks approach for various aesthetic evaluations of structures, in particular concrete gravity dams. It has been found that the trained networks perform well. In addition, the relationship between the input and the output items has been identified quantitatively by a sensitivity analysis after training.