Abstract: In particular embodiments, a computing system may receive an image comprising one or more virtual elements associated with a virtual environment and one or more real-world elements associated with a real-world environment. The system may determine a first metric and a second metric indicative of a measure of clutter in the virtual environment and the real-world environment, respectively. The system may determine gaze features associated with a user based on a user activity and predict, using a machine learning model, a reaction time of the user based on the gaze features. The system may determine a third metric indicative of the measure of clutter in the image based on predicted reaction time. The system may compute an overall clutter metric based on the first, second, and third metrics. The system may perform one or more actions to manage the clutter in the image based on the overall clutter metric.