Abstract
We investigate whether wording, stylistic choices, and online behavior can be used to predict the age category of blog authors. Our hypothesis is that significant changes in writing style distinguish pre-social media bloggers from post-social media bloggers. Through experimentation with a range of years, we found that the birth dates of students in college at the time when social media such as AIM, SMS text messaging, MySpace and Facebook first became popular, enable accurate age prediction. We also show that internet writing characteristics are important features for age prediction, but that lexical content is also needed to produce significantly more accurate results. Our best results allow for 81.57% accuracy.
Introduction
Predicting the age of an anonymous author is a classic challenge in natural language processing (NLP), with applications ranging from targeted marketing to forensic linguistics. This research moves beyond treating age as a static category, instead focusing on generational shifts. The authors hypothesize that the "Generation Y" (Millennial) cohort developed a unique digital dialect because they were the first to be "college-aged" (18–22) when transformative social technologies—like Instant Messaging (AIM), SMS, and early social networks (MySpace, Facebook)—became mainstream.
Methods and Dataset
The study utilized a corpus of approximately 24,500 blogs scraped from LiveJournal, focusing on US-based authors born between 1940 and 2000. LiveJournal was selected for its rich metadata and long-term posting histories.
A supervised machine learning approach using Logistic Regression was employed. Feature engineering was divided into three categories:
- Online Behavior: Metadata-driven features like friend counts, post frequency, and time of day.
- Lexical-Stylistic: "How" people write, including emoticons, internet slang, and capitalization patterns.
- Lexical-Content: "What" people write, using a Bag-of-Words (BOW) approach to identify significant word pairs and Part-of-Speech patterns.
Key Insights: The "College Effect"
The most accurate age splits (1979 and 1984) correspond to when authors would have been in college during the launch of AIM (1997) and MySpace/Facebook (2003/2004). This suggests that the technology used during the transition to adulthood permanently shapes a user's social "writing voice."
Stylistic features like emoticons and slang were found to be more "portable" markers of age across platforms compared to lexical content. Notably, as the birth year gets later (younger authors), average sentence length decreases while the use of emoticons and "internet-speak" increases linearly. Additionally, younger bloggers tend to use first-person singular pronouns (I, me) more frequently, whereas older bloggers favor first-person plural (we).
Results
The model achieved its peak performance—81.57% accuracy—when distinguishing between authors born before and after 1984. For the 1979 threshold, the model achieved 79.96% accuracy. While behavior or style alone yielded accuracies in the 60-70% range, combining them with Lexical Content provided a significant boost, proving that age is best predicted by a holistic view of a user's digital footprint.
Conclusion and Future Work
This study demonstrates that generational shifts in technology adoption create distinct linguistic markers. While platform-specific features like "Interests" are predictive, stylistic features offer a more universal approach to age classification. Future research may apply these techniques to higher-level tasks, such as identifying opinion or power dynamics in online text, which are often influenced by the speaker's age.
Summary updated