The Influence of Personalized Content Algorithms on Persistent Low Mood: Assessment and Intervention

Abstract

The study was sparked by clinical observations of clients whose symptoms closely resembled depression but did not fully match diagnostic criteria. These individuals experienced persistent sadness perpetuated by algorithm driven digital content, prompting the development of the Algorithm-Induced Low Mood Scale (AILMS) to better capture this distinct mood disturbance. 50 participants were randomly assigned to either a control group receiving standard cognitive therapy or an experimental group receiving therapy combined with support to disrupt algorithm-driven content patterns. Mood scores were recorded at three points: before treatment, after treatment, and at follow-up several weeks later. There was a clear improvement over time, F(2, 47) = 75.30, p < .001, η² = .76, with scores dropping from pretest (M = 22.08) to post test (M = 10.71), and slightly rising at follow-up (M = 13.61). The experimental group showed greater improvement and maintained progress better. At follow-up, the control group experienced a significant relapse, t(48) = 4.29, p < .001. AILMS scores moderately correlated (r = .681), with Beck Depression Inventory II (BDI II), but some items did not correspond with BDI II patterns, suggesting the scale captures a unique experience. The factor analysis showed that the mood disturbance measured by AILMS involves three correlated but different parts: feelings of sadness, how users react emotionally, and how algorithms affect these emotions. Although passive social media users had higher average scores, differences were not statistically significant, F(1, 48) = 2.10, p = .154. These findings support the validity of AILMS and suggest that helping individuals disrupt mood-matching digital content loops may aid emotional recovery.