Source-linked AI summary
Collective attention in the age of (mis)information
Delia Mocanu, Luca Rossi, Qian Zhang, Màrton Karsai, Walter Quattrociocchi
TL;DR
The paper asks how information quality affects collective attention and political interaction on Facebook, where false information is widespread. It analyzes categorized Facebook pages, user interaction patterns, and responses to injected troll posts, finding that attention patterns are similar across information types and that alternative-news-focused users are more prone to interact with false claims.
Problem
The paper addresses limited understanding of how information quality affects attention processes, including information virality, lifespan, and consumption patterns, amid pervasive social-media misinformation.
Method
The study categorizes Facebook pages and users by interaction patterns, then measures responses to 2788 false-information posts from a troll page.
Results
Attention patterns are similar across political activism, alternative information, and regular news, while users strongly preferring alternative sources are more susceptible to false information.
Takeaways & Limitations
Unsubstantiated claims can reverberate as long as more verified information, and alternative-news-focused users are especially responsive to false claims.
Takeaways & Limitations
The analysis observes only active participating users and excludes page subscription lists, regardless of users’ privacy settings.
Abstract
from arXiv · showhide
In this work we study, on a sample of 2.3 million individuals, how Facebook users consumed different information at the edge of political discussion and news during the last Italian electoral competition. Pages are categorized, according to their topics and the communities of interests they pertain to, in a) alternative information sources (diffusing topics that are neglected by science and main stream media); b) online political activism; and c) main stream media. We show that attention patterns are similar despite the different qualitative nature of the information, meaning that unsubstantiated claims (mainly conspiracy theories) reverberate for as long as other information. Finally, we categorize users according to their interaction patterns among the different topics and measure how a sample of this social ecosystem (1279 users) responded to the injection of 2788 false information posts. Our analysis reveals that users which are prominently interacting with alternative information sources (i.e. more exposed to unsubstantiated claims) are more prone to interact with false claims.
Introduction
The paper examines how information quality shapes attention and political discussion on Facebook, amid widespread false information and unclear effects on informed debate. It focuses on Italian Facebook communities around alternative news, political activism, and mainstream media before the 2013 elections.
- False information is pervasive on social media and can foster collective credulity despite expectations that digital technologies encourage informed debate.
- Unsubstantiated claims, including conspiracy theories that reject scientific evidence, are proliferating across the Internet.
- A central open question is how information quality affects attention processes such as virality, lifespan, and consumption patterns.
- The study investigates relationships between information sources and online political debates during the period preceding Italy’s 2013 elections.
- Political activism and alternative-information pages proliferated, while troll pages used parody, satire, and fabricated statements that sometimes became viral evidence in online debates.
- The analysis compares alternative news, online political activism, and mainstream news, finding similar attention patterns and greater false-information responsiveness among users focused on alternative sources.
Methods
The study builds a public Facebook dataset by categorizing pages according to their social functions and information, then analyzes user interactions and responses to troll-generated false posts. Its data cover six months of activity across 50 pages, supplemented by 2788 troll post ids.
- The dataset contains all posts and user interactions from 50 public pages collected over six months, from September 1, 2012 to February 28, 2013.
- Pages are classified as mainstream newspapers, alternative information sources, or political activism according to their social functions and disseminated information.
- User activity is measured through interactions with public posts, specifically likes, shares, and comments.
- The dataset breakdown distinguishes mainstream news, alternative news, and political activism by their page types and social purposes.
- 2788 posts from a troll page supplied caricatural political and alternative-news content containing false information for measuring ecosystem responses.
Results and Discussion
Across alternative information, political activism, and mainstream news, users displayed similar attention and post-lifetime patterns despite differences in topic and information quality. Users active across these communities were classified by interaction behavior, and those strongly affiliated with alternative information sources were most responsive to injected false information.
- Attention patterns: User interaction distributions were nearly identical across alternative information, political activism, and mainstream media pages.The similarity remained when comments and likes were analyzed separately.
- Attention patterns: Post interest lifetimes showed similar distributions across all three page types.Lifetime was measured as the temporal distance between a post’s first and last comment.
- Cross-community interaction: Political discussion and alternative news users interacted with each other more dominantly than with mainstream media users.Users from the first two groups were also represented in mainstream newspaper audiences in comparable ways.
- User classification: The analysis classified users from six months of likes, comments, and likes to comments, using likes and non-overlapping 95% confidence intervals to assign strong topic labels.Only active participating users were analyzed, and users’ page subscriptions were not available.
- User classification: Alternative-information and political-activism pages had dominant labeled-user fractions of 45% and 49%, respectively.Mainstream media pages showed a more balanced distribution of user classes.
- Response to false information: Users with strong preferences for alternative information sources were more susceptible to interacting with false information.The response analysis counted users who liked 2788 troll posts, including posts containing multiple false statements.
Conclusions
The study finds strong interaction between political discussion and information sources, with similar consumption patterns across qualitatively different information. It also examines how Facebook contributes to the diffusion of false beliefs when truthful and untruthful rumors coexist.
- The study provides an outline of Facebook’s role in diffusing false beliefs when truthful and untruthful rumors coexist.
- Political discussion interacts strongly with both alternative and mainstream information sources in the analyzed Facebook ecosystem.
- Consumption patterns are similar despite differences in the nature of the information.
- Parodistic troll memes sometimes foment animated debates and diffuse through the community like other information.
Authors Contribution
The authors divide responsibilities across experiment design, experimentation, data analysis, tools, and writing.
- WQ and DM designed the experiments; WQ performed them; WQ, DM, and QZ analyzed the data; and WQ, DM, LR, QZ, and MK wrote the paper.
1 Supporting Information
The supporting information describes Facebook Graph API data collection, its access constraints, and supplementary analyses of user interactions and post lifetimes across page categories. It also documents the page-category examples and statistical tests used to compare distributions.
- Data collection: Data were collected from public Facebook pages through the Graph API, including posts, user likes, comments, authors, timestamps, and post types.The observation window covered September 1, 2012 through February 28, 2013.
- Data collection: A post pointing to an existing photo or video was distinguished from the linked object, whose interactions were downloaded separately.Posts and objects can accumulate separate likes, comments, and shares, while objects may accrue interest over longer periods.
- Data collection limitations: Graph API rate limits and privacy settings produced incomplete interaction data, with approximately 20% of observed user actions estimated to be invisible.The API limited access to the last 5000 shares, while likes and comments could also be partial; this estimate was specific to the studied users and pages.
- Data collection limitations: The complete sharing tree of an object with more than 5000 shares was virtually impossible to reconstruct because upstream access was unavailable and privacy restrictions applied.The API allowed access to branches of share actions but not the reverse upstream process.
- Post lifetime analysis: Supplementary analyses define post lifetime as the interval between the first and last comment and compare normalized distributions with quantile-quantile plots and tail-truncated tests.The supporting figures compare alternative news, mainstream news, and political movements; KS and t tests cut normalized-distribution tails at 0.1.
- Page categories: The analyzed page set included alternative information, mainstream media, and political activism categories, with examples ranging from conspiracy-oriented pages to national newspapers and activist groups.The listed examples include pages discussing chemtrails, vaccines and autism, political lobbying, government actions, and national news coverage.