[Paper Review] WhatsApp, Doc? A First Look at WhatsApp Public Group Data

Kiran Garimella, Gareth Tyson|arXiv (Cornell University)|Apr 4, 2018

Complex Network Analysis Techniques15 references22 citations

TL;DR

This paper presents a novel, reproducible methodology for collecting large-scale public WhatsApp group data, demonstrating feasibility through a dataset of 178 groups with 45K users and 454K messages. The study reveals high diversity in group topics, activity levels, and geographic spread, enabling new social science research on communication dynamics, misinformation, and emerging digital institutions.

ABSTRACT

In this dataset paper we describe our work on the collection and analysis of public WhatsApp group data. Our primary goal is to explore the feasibility of collecting and using WhatsApp data for social science research. We therefore present a generalisable data collection methodology, and a publicly available dataset for use by other researchers. To provide context, we perform statistical exploration to allow researchers to understand what public WhatsApp group data can be collected and how this data can be used. Given the widespread use of WhatsApp, our techniques to obtain public data and potential applications are important for the community.

Motivation & Objective

• To explore the feasibility of collecting and analyzing public WhatsApp group data for social science research.
• To develop a generalizable data collection methodology for large-scale messaging platform data.
• To characterize communication patterns, diversity, and biases in public WhatsApp groups.
• To demonstrate the potential of WhatsApp data for studying misinformation, social institutions, and cross-cultural interactions.
• To release an anonymized dataset and open-source code for reproducible research.

Proposed method

• Scraping publicly listed WhatsApp groups from third-party websites such as joinwhatsappgroup.com.
• Automatically subscribing to groups via WhatsApp's public group invitation links.
• Monitoring and logging all messages in a structured, time-stamped schema for analysis.
• Applying text analysis and clustering techniques to categorize group topics based on titles and content.
• Using word clouds and statistical summaries to visualize topic distribution and communication patterns.
• Anonymizing user identifiers and releasing the dataset with full source code for reproducibility.

Experimental results

Research questions

• RQ1 To what extent are WhatsApp groups a broadcast, multicast, or unicast communication medium?
• RQ2 How do interaction dynamics and message frequency evolve over time in public WhatsApp groups?
• RQ3 What is the geographical distribution of users in public WhatsApp groups, and how does location affect communication patterns?
• RQ4 How do multimedia messages influence interaction and content spread in WhatsApp groups?
• RQ5 What is the potential of WhatsApp group data for studying misinformation, digital institutions, and social innovation in underrepresented regions?

Key findings

• The study successfully collected a dataset of 178 public WhatsApp groups, comprising 45,454 users and 454,000 messages, demonstrating the feasibility of large-scale data collection from WhatsApp.
• The largest group, 'DISFRUTA AL MAXIMO', contained 11,000 messages and was primarily based in Colombia, indicating strong regional engagement.
• Groups were highly diverse in topic, with major categories including politics (17 groups), sports (12 groups), and education/job discussions (23 groups), reflecting broad thematic coverage.
• Geographic diversity was significant, with notable user presence in developing regions such as India and Nigeria, suggesting WhatsApp's utility for studying underrepresented populations.
• The dataset revealed that WhatsApp groups are not limited to bilateral communication but support complex, multi-party, multimedia interactions, especially around shared interests.
• The study found that WhatsApp groups can serve as platforms for emerging social and economic institutions, such as informal job markets and community-based information networks.