Blog-Reading Histories as Networks: A Proof-of-Concept Prediction of High Parenting Stress

Blog writing reportedly reflects authors’ characteristics. However, readers’ blog-reading histories may also reflect their characteristics, an area that remains underexplored. In this context, we aimed to construct a reader–blog network based on reading histories to examine whether blog-reading histories contain information predictive of high parenting stress (HPS). The participants were 980 parents who held an account ID with Ameba (a popular blogging and social networking service in Japan) and were raising at least one child under 18 years of age; they completed the Parenting Stress Index–Short Form. We classified them into high (n = 425) and low (n = 555) parenting stress groups. We analyzed 881,354 blog posts read over a 3-month period and 1,262,033 edges linking participants to the posts they had read. Compared with the low parenting stress group, parents in the HPS group read significantly more content containing tokens indicative of negative emotions, sentences expressing negative affect in the family domain, and posts (articles) characterized by marital dissatisfaction, parenting difficulties, and self-criticism. The graph-based models showed above-chance but modest discrimination between the HPS and LPS groups (AUC = .662). These findings provide proof-of-concept evidence that blog-reading histories contain predictive information about parenting stress; however, the observed discrimination is not sufficient to support practical screening or intervention applications.