[Paper Review] A Comparison of Common Users across Instagram and Ask.fm to Better Understand Cyberbullying

Homa Hosseinmardi, Rahat Ibn Rafiq+5 more|arXiv (Cornell University)|Aug 21, 2014

Hate Speech and Cyberbullying Detection12 references18 citations

TL;DR

This study compares common users across Instagram and Ask.fm to analyze differences in negativity and positivity in user-generated content, revealing that Ask.fm exhibits significantly higher negativity despite anonymity, which surprisingly correlates with lower negativity levels. The research uses linguistic analysis and behavioral correlation to understand cyberbullying dynamics across social networks with differing privacy models.

ABSTRACT

This paper examines users who are common to two popular online social networks, Instagram and Ask.fm, that are often used for cyberbullying. An analysis of the negativity and positivity of word usage in posts by common users of these two social networks is performed. These results are normalized in comparison to a sample of typical users in both networks. We also examine the posting activity of common user profiles and consider its correlation with negativity. Within the Ask.fm social network, which allows anonymous posts, the relationship between anonymity and negativity is further explored.

Motivation & Objective

• To understand how cyberbullying behaviors differ between two popular social networks, Instagram and Ask.fm, which are frequently used for such activities.
• To investigate the role of anonymity in shaping the expression of negativity and positivity in user posts on Ask.fm compared to Instagram.
• To analyze posting behaviors of users common to both platforms to determine whether shared users exhibit distinct behavioral patterns across networks.
• To explore the correlation between user activity levels and negativity in online interactions on both platforms.
• To examine whether profile owners and their 'friends' on Ask.fm show similar patterns of negativity and positivity, and how this compares to Instagram.

Proposed method

• Collected comprehensive data on four user types: normal Instagram users, normal Ask.fm users, and common users (active on both platforms).
• Performed linguistic analysis on user comments to quantify positivity and negativity using predefined word lists for negative and positive sentiment.
• Normalized sentiment scores by comparing common users to a sample of typical users in each network to control for baseline differences.
• Analyzed posting activity patterns (frequency, volume) of common users and correlated them with sentiment scores.
• Explored the relationship between anonymity and negativity in Ask.fm by comparing anonymous and non-anonymous comments on the same profiles.
• Conducted correlation analysis between profile owners' and friends' sentiment behaviors on Ask.fm and cross-network behavioral alignment.

Experimental results

Research questions

• RQ1 How does the level of negativity in user posts differ between Instagram and Ask.fm?
• RQ2 What is the relationship between anonymity and the expression of negativity in user comments on Ask.fm?
• RQ3 Do common users exhibit similar positivity and negativity levels on both Instagram and Ask.fm compared to non-common users?
• RQ4 Is there a significant correlation between user activity levels and sentiment expression in either network?
• RQ5 How do the sentiment behaviors of profile owners and their friends correlate on Ask.fm, and is there cross-network consistency in user behavior?

Key findings

• Ask.fm exhibits significantly higher levels of negativity in user posts compared to Instagram, despite both platforms being used for cyberbullying.
• There is no statistically significant difference in the level of positivity between users on Instagram and Ask.fm.
• Common users on both platforms show similar levels of positivity and negativity compared to non-common users within each network, indicating behavioral consistency.
• Anonymity on Ask.fm is associated with lower negativity in comments, a counter-intuitive result suggesting that anonymous users may use less aggressive language than identified users.
• A strong correlation exists between the negativity and positivity of profile owners and their friends on Ask.fm, indicating coordinated emotional expression patterns.
• Profile owners who are positive on one network tend to be positive on the other, and similarly for negative behavior, suggesting behavioral consistency across platforms.