"Disinformation on WhatsApp: Maldita.es' chatbot and the 'Frequently Forwarded' attribute" is a report that analyzes the importance and usefulness of a WhatsApp service for fact-checking organizations such as Maldita.es. The impact and progression of the WhatsApp chatbot developed by Maldita.es in July 2020 and the progress made by automation are also evaluated compared to the old manual WhatsApp Business system. In addition, it also looks at how the “Frequently Forwarded” attribute, also known as Highly Forwarded Messages (HFMs), can help fact-checkers through case studies. Maldita.es is the first fact-checker in the world to investigate its usefulness for the fact-checking community.
Our WhatsApp chatbot has won the European Press Prize 2021 in the innovation category.
Maldita.es' database contains the trace that disinformation has left in Spain in recent years. Every piece of data in these files provides specific information that allows us to know how disinformation works, when it appears, how it is shared or what techniques it uses to try to get as far as possible.
Back in July 2020 amid the COVID-19 pandemic, we launched our WhatsApp chatbot and began to collect data on how disinformation was being shared with us in an automated way, enriching our database and standardizing a methodology that until then had been manual.Today our chatbot is capable of receiving and identifying videos, audios, images and text; comparing them with our database and responding automatically if there is a debunk. In addition, it archives all the misinformation complaints it receives to measure its virality and its evolution over time.
Since March 2021, a new element has been included in that database: messages marked with the Frequently Forwarded (FF) attribute, also known as Highly Forwarded Messages (HFMs), which indicates those that have already been forwarded on WhatsApp five or more times. They are marked with a double arrow. According to WhatsApp, these messages can no longer be sent to various groups at the same time, but they can still be sent individually to other recipients.
The academic research team at Maldita.es has made a first analysis of the meaning and relevance of this new attribute. After studying the patterns of appearance of the FF during the first month in which it was active (March 2021), we have found indications that it is strongly linked to disinformation:
- 74% of the alerts with FF we received were linked to an investigation by Maldita.es' team.
- 78.72% of the investigated FF alerts were finally rated as a hoax or disinformation.
- Contents associated with the FF attribute are three times more likely to be monitored as probable disinformation by Maldita's journalists following our own journalistic and methodological criteria. Therefore, this is an attribute that can help fact-checkers prioritize their work in order to scale it.
An example of how this FF attribute indicates high rates of virality is this hoax in which the former Vice President of the Spanish Government is falsely accused of doing a legal trick to receive a very high compensation when leaving office: the curve of total alerts that reached our WhatsApp compared to the alerts associated with the FF attribute is coincident and that indicates that multiple FF alerts in a single day indicate high virality.
Additionally, in the patterns of appearance of the alerts with FF we observe that they tend to be included in hoaxes with a very high potential impact, either because they are likely to become very viral or because they are old hoaxes that already had a significant impact in the past and are back in circulation. There are also features that suggest that these types of alerts can signal the appearance of coordinated disinformation campaigns. This data opens the door to the development of an early warning system that allows fact-checkers to react earlier to this type of disinformation. These are some of the promising lines of research that are opened after the analysis of only one month of operation of the FF attribute. A broader and more in-depth study can confirm these hypotheses and expand the knowledge about the possibilities that FF alerts provide to facilitate the work of verifiers.
After nearly a year of functioning, over 26,000 people have used our automated service, of whom 61% have done so more than once. Our chatbot has sent over 400,000 messages, received 108,000 contents to fact-check and issued 143,800 daily debunks summaries in its text format and 18,400 in audio format. Just in 2021, 16,200 users have contacted us, with an average of over 950 daily conversations among users who sought to verify contents and those who requested other options in our chatbot.
Primera fecha de publicación de este artículo: 03/06/2021