Is My Interface Beautiful?—A Computational Model-Based Approach

In this paper, we proposed a computational model to measure the beauty or aesthetics of an interface by analyzing the components associated with it. The three primary elements-text, image, and white space were considered for our study. Using support vector regression, we developed two different computational models to predict the aesthetics of the image and the text components present in an interface. To create the image aesthetics model, we considered the 11 features associated with image aesthetics. The 83 users rated a total of 250 images considered for our study. The average users' rating and the 11 feature values were used to train and test our image aesthetics model. The experimental results showed that our model could predict image aesthetics an RMSE of 0.68. Another computational model of text aesthetics was developed by considering the six features of text elements. We designed 95 text samples by varying these features systematically. The 185 users rated the text samples. The six feature values along with the average users' ratings were used to train and test our text aesthetics model. The model can measure text aesthetics with an RMSE of 0.58. We performed another empirical study to develop the white space aesthetics model. It was observed that the median scale is best suited for this. Finally, we created the interface esthetic model by integrating the three models-text, image, and white space. The 150 interfaces were designed to validate our final model. The experimental results showed that our proposed model could measure interface aesthetics with an RMSE of only 0.79.