Behavioral effects of affective filtering

Determine whether the perceptual arousal reductions produced by learned affective image adaptations translate into behavioral outcomes, such as reduced engagement with social media content, through longitudinal, in-the-wild studies.

Background

The paper demonstrates that the learned parametric filter reduces self-reported arousal in a controlled study of 54 participants and performs comparably to grayscale filtering while receiving higher perceived-quality ratings at an appropriate conditioning strength. However, the experiment uses single exposures to 12 images in a controlled setting and measures self-reported affect rather than actual behavior. The authors therefore identify whether these perceptual changes reduce engagement or otherwise affect real-world social-media use as the central unresolved issue, motivating longitudinal and in-the-wild evaluation using the deployed real-time system.