A computer-implemented method of determining visual preference of a user for selecting item images from at least one item database according to the visual preference of the user, the method being implemented by one or more computer processors and comprising:
- selecting a display set of item images from the at least one item database;
- causing the item images of the display set to be displayed at a user interface in a scrollable item array at respective array positions, wherein the item images are arranged in item groups within the item array based on at least one visual feature type, such that item images with similar visual features of that visual feature type are grouped together;
- detecting engagement events at the user interface between the user and engaged-with item images of the display set, wherein the engagement events comprise at least one passive engagement event which indicates passive engagement with one of the item groups without the user selecting any image of the passively engaged-with item group;
- using the detected engagement events to determine a visual preference hypothesis for the user based on visual features extracted from the engaged-with item images;
- and
- updating the display set with new item images selected from the at least one item database by comparing visual features extracted from the new item images with the at least one visual preference hypothesis;
- wherein subsequent engagement events between the user and engaged-with new item images of the display set continue to be detected and used to refine the at least one visual preference hypothesis, and the display set continues to be updated as the visual preference hypothesis is refined, wherein the at least one passive engagement event comprises a scroll event and the visual preference hypothesis is determined or updated based on a direction of the scroll event at the user interface relative to the passively-engaged-with item group.