We examined “Internet Fatigue” as a “Protective Signal” in visibility-intensive social media. We analyzed 871 free text from young Social Networking
Sites (SNS) users in Japan (mean age = 21.9) using an embedding-assisted content analysis. Sentences were encoded with Multilingual E5 (Wang et al.,
2022; 2023; 2024) and clustered with k-means. We inspected the elbow curve (k = 2–30) to delimit a candidate range, then selected k by the average
silhouette within that range. Overall, 747/871 (85.8%) re-ported at least one fatigue experience. Seventeen interpretable clusters emerged—e.g., AltAccount Exposure and Conformity Pressure; Problematic Use and Time Dis-placement; Relational Maintenance Pressure—many tied to heightened
visibility and inflow. We argue that internet fatigue often functions as a protective signal, prompting boundary-setting and other self-regulatory actions.

Studies on Internet Addiction often interpret “spending more time online” as a sign of addiction. This study tests this time-based paradigm with a survey of adolescents and young adults in Japan (N = 313). We measured daily Internet use, screen time, and scores on a brief diagnostic questionnaire (DQ), and we recorded participants’ use of major social media platforms. We conducted correlations and multiple regression analyses to examine how strongly time spent online and social media use predict DQ scores. Time spent online showed only a small positive association with DQ scores (adjusted R² = 0.030; R² = 0.067). Most social media variables were not statistically significant. Gender had a small effect. These results suggest that time spent online alone is a weak indicator of problematic use. We discuss the importance of including qualitative symptoms (such as loss of control, distress, and interference with daily life) and psychosocial factors (motives, stress, and loneliness) in assessments. Furthermore, we outline implications of these findings for caregivers, schools, and platform design.
Keywords
Internet Addiction, Time-based Paradigm, Adolescents, Social media, Assessment

The rapid expansion of SNS use among adolescents has raised concerns about its potential impact on mental health and academic performance. In particular, Internet addiction, characterized by excessive and problematic Internet use, has become a critical issue. However, previous studies have not reached a consensus on whether SNS use indirectly affects Internet addiction through increased Internet use time or whether SNS use directly leads to addiction. This study compares two models: (1) a mediation model where SNS use affects Internet addiction through NetTime, and (2) a direct model where SNS use directly influences Internet addiction. Using standardized regression coefficients (β) from path analysis, we analyze the impact of SNS platforms and gender on Internet addiction. The findings indicate that Discord has significant direct and indirect effects on Internet addiction, whereas Instagram shows only a weak or even negative association. These results offer valuable insights for policymakers and educators aiming to develop targeted interventions to promote responsible SNS use among adolescents.

Adolescent PIU (Problematic Internet Use) has been studied in relation to
sleep, specific applications, and school life, but few studies have integrated these
indicators to characterize a comprehensive profile. In this study, a web-based survey
was conducted among 5,049 from fourth-grade elementary school to third-grade
junior high school students from a central city in Japan. Assessment items included
average daily internet use time, average daily sleep time, school absenteeism in
the past year, and the eight-item Young Diagnostic Questionnaire (DQ). After Zstandardization,
𝑘-means clustering was performed. The number of clusters was
selected using the elbow method and the Silhouette coefficient (𝑘 = 6). Regarding
app preferences, multiple app entries were binarized and aggregated into each
cluster. Correspondence analysis (CA) was performed using the app contingency
table for each cluster (apps with fewer than 10 users were grouped as "other"). The
results showed that two low-risk clusters (57% of students) spent approximately
2.39-2.56 hours of internet time per day, approximately 8.18-8.31 hours of sleep
per day, approximately 3.69-3.93 days of absence per year, and had a low DQ
(approximately 1.09-1.15). The remaining higher-risk clusters included (1) students
with short sleep and high DQ (approximately 3.58-4.40) and (2) students with very
long internet time (approximately 11.91 hours). CA analysis revealed clear app
mapping: the low-risk cluster tended to use games/portal apps, the high DQ cluster
tended to use Instagram/TikTok, and the very long-time cluster tended to use LINE.
By integrating DQ, time usage, sleep, and absence, diverse student subtypes with
distinct app profiles were identified. Our findings suggest that the "always-on"
state may occur even when DQ scores are not uniformly high, and the blurring of
the boundaries between online communication and everyday life warrants further
investigation.