leia este post em português

Today at Global Fact 7, the largest worldwide gathering of fact-checkers, global technology not-for-profit Meedan announced a case study of a global fact-checking project using the WhatsApp Business API. The groundbreaking study is based on nearly 5,700 fact-checks across India, Brazil, South Africa, Kenya and Nigeria and was made possible with support from WhatsApp and Facebook.

Download the full report: End-to-End Fact-Checking

For the project, Meedan developed several software features on Check, a web application for journalists to verify and annotate online information for open-source investigations, that were customized specifically for global fact-checking organizations and deployed by the AFP (in Brazil and India), Africa Check, BOOM and India Today. They have  been running since October 2019 across 4 languages and more than 180,000 submissions, including ongoing submissions today.

The report, called End-to-End Fact-Checking, details the insights, field research and product features made that maximized the value of the API for fact checkers around the world; it’s name is a reference to the program’s ability to successfully fact-check on the end-to-end encrypted service.

"Fact-checkers working globally have a variety of needs, so we built tools that optimize workflows and the nuances of AI in order to create a roadmap for improving fact-checking and information integrity across the web," said Ed Bice, CEO, Meedan. "This effort has been highly collaborative, and we appreciate the support and guidance that we were given from the teams at AFP, Africa Check, BOOM and India Today in particular."

The report offers the following core insights:

Bots are the new audience engagement paradigm.

A WhatsApp fact-Checking bot can serve three purposes: (1) help audiences understand how to interact with the fact-checking team and see commonly-requested content, (2) serve as a bridge between a human editor and someone submitting content to a tipline and (3) cut down on unactionable content.

Machines in the loop augment quality human efforts.

Already designed for human work, Check now automates everything from content matching to intake, helping fact-checkers do their jobs more efficiently and therefore effectively.

Follow the news cycle.

Newsrooms learned that a topic-oriented approach to tipline promotion following the news cycle yields stronger audience engagement and submissions. After the outbreak of COVID-19, we saw an uptick in the percentage of quality, actionable content, in turn generating better fact-checks.

Fact-checkers need no-nonsense, no-code interfaces in multiple languages.

We adapted our toolkit to help fact-checkers focus on the fact-checking rather rather than technical tasks. This enabled teams to customize their workflows.

"We used to get inundated by the thousands of messages that people would send us on WhatsApp," said Balkrishna, fact-checking editor at India Today. "Check has not only made it possible for us to sift through the fact-check requests quickly, it also helped deliver most of our readers with fast and personalized responses. This helps us build a direct bond of trust with our audience."

"Check really improved the way we communicate with our readers on WhatsApp," said María Clara Pestre, a fact-checker at AFP Checamos (Brazil). "Now, we can be sure anyone who reaches out through the app will have an immediate and consistent reply, while having time to dedicate ourselves to other parts of the verification process."

"We’re excited to work with partners like Meedan to help address the challenge of misinformation within private messaging. Our approach is to give users tools to easily get accurate information and support fact checking," said Ben Supple, head of civic engagement at WhatsApp. "We appreciate and value the work done to maximize the WhatsApp Business API for fact checkers and ensure more people have access to accurate information at the click of a button."

Meedan presented this study at Global Fact 7, with representatives from WhatsApp, the International Fact Checking Network, Boom and Africa Check. See the video from the panel here:

Meedan's Shalini Joshi moderates a discussion about the evolution of fact-checking on WhatsApp at Global Fact 7

To support this work, Meedan developed a number of key software features that maximized the WhatsApp Business API with a mix of AI, interface and multilingual innovations, including:

Multilingual WhatsApp Bot: a customizable bot that presents both intuitive menu items and human conversations with fact-checkers in multiple languages

Workflow Automation: a triage system to recognize, review and prioritize claims and responses

Similarity Analysis: algorithms that detect content patterns and trends based on variations of the same claims

Response at Scale: a publishing system to send fact-check reports to tipline users who send duplicate or similar requests

"Our goal with Check," explained Meedan’s Director of Product Pierre Conti, "is to build a scalable human and AI system for fact-checkers to efficiently size the demand for verification directly from their audience, cut through the noise, and maximize impact by helping them to prioritize their response to the most pressing misinformation."

  1. Online conversations are heavily influenced by news coverage, like the 2022 Supreme Court decision on abortion. The relationship is less clear between big breaking news and specific increases in online misinformation.
  2. The tweets analyzed were a random sample qualitatively coded as “misinformation” or “not misinformation” by two qualitative coders trained in public health and internet studies.
  3. This method used Twitter’s historical search API
  4. The peak was a significant outlier compared to days before it using Grubbs' test for outliers for Chemical Abortion (p<0.2 for the decision; p<0.003 for the leak) and Herbal Abortion (p<0.001 for the decision and leak).
  5. All our searches were case insensitive and could match substrings; so, “revers” matches “reverse”, “reversal”, etc.
Words by
No items found.
No items found.
Words by
Published on
June 25, 2020
April 20, 2022