Comparing Interfaces for Image Sorting

Abstract

Many tasks ask people to make categorization decisions about a large number of items based on images. There are many ways in which this task can be presented to the user, but there is little information available about how the decision technique affects performance. To investigate this topic three user studies were conducted, which compared different methods for user categorization of images: one based on the "swipe right / swipe left" mechanism commonly utilized in dating sites, four based on grids of images, one that presents a moving stream of images, and one that rapidly flashes a sequence. Additional studies were conducted to investigate the impact of altering certain aspects of the interface design, such as changing the speed of image presentation or adding extra features including zoom and image preview. These studies also investigated the impact of changing the context of the image categorization task, to make the task more familiar to the users. The experimental studies showed strong differences between the various techniques in time, accuracy, effort, and preference, with the small grid the clear winner on all measures. This work provides the first empirical evidence about different approaches for supporting image categorization, and identifies a technique that has distinct advantages over other methods.