By Claire Wardle

Despite endless headlines, reports and conferences on the topic of information disorder, the global news industry remains woefully unprepared to tackle the increasingly effective and dangerous tactics deployed by those intent on disrupting the public sphere. Journalists often write about the phenomenon at arm’s length, rarely considering how the current environment requires newsrooms to integrate new skills and adapt their routines, standards and ethics in order to prevent deliberate falsehoods and coordinated conspiracy from creeping into the mainstream.

From our work at First Draft over the past few years, our most important takeaway is understanding that many agents of disinformation see coverage by established news outlets as the end-goal. Their tactics center around polluting the information ecosystem by seeding misleading or fabricated content, hoping to catch out journalists who now regularly turn to online sources to inform their newsgathering. Having their deliberate hoax or manufactured rumor featured and amplified by an influential news organization is considered a serious win, but so is finding their work the focus of a debunk. It all amounts to coverage. As Ryan Broderick reported in the aftermath of the ‘MacronLeaks’ data dump on the eve of the French election in May 2017, users on 4chan celebrated when mainstream news outlets began fact-checking the controversy surrounding Macron’s financial affairs, boasting that the debunks were a “form of engagement”. Over the past year we have seen these tactics play out in various regional contexts around the world. The platforms might be different but the techniques are the same.

Drawing on a ‘Platform Migration’ diagram that was first created by Ben Decker, as well as the excellent work of Whitney Philips in her report, Oxygen of Amplification, I am sharing this ‘Trumpet of Amplification’. It is a simplified diagram showing the journey that disinformation often takes, as a way of kick-starting conversations in newsrooms. Disinformation often starts on the anonymous web (platforms like 4chan and Discord), moves into closed or semi-closed groups (Twitter DM groups, Facebook or WhatsApp groups), onto conspiracy communities on Reddit Forums or YouTube channels, and then onto open social networks like Twitter, Facebook and Instagram. Unfortunately at this point, it often moves into the professional media. This might be when a false piece of information or content is embedded in an article or quoted in a story without adequate verification. But it might also be when a newsroom decides to publish a debunk or expose the primary source of a conspiracy. Either way, the agents of disinformation have won. Amplification, in any form, was their goal in the first place.

Efforts to undermine and explain deliberate falsehoods can be extremely valuable and are almost always in the public interest, but they must be handled with care. All journalists and their editors should now understand the risks of legitimizing a rumor and spreading it further than it might have otherwise traveled, particularly in newsrooms developing misinformation as a ‘beat’ in its own right.

Most newsrooms now have dedicated social media monitoring teams, looking for tips, sources and eyewitness content. The problem is that not all of those teams have had the training to deep dive into the provenance of the posts, images or videos they find online. Where did they originate? What was the motivation of the publisher? Unless journalists are taught to really interrogate a source, they could miss how details first appeared on Discord two weeks previously, or were part of a coordinated campaign on 4chan to amplify certain messages, or emerged as a result of tactics designed in a WhatsApp group, or a narrative honed in conspiracy communities on YouTube.

The news industry is incredibly vulnerable. There are thousands of journalists globally, many of them independently monitoring and posting on social networks every day. Convince one journalist to publish the falsehood or fabricated content, and it gets pushed almost immediately across the wider community (many newsrooms do not check reporting by other journalists or outlets with the same rigor, assuming that thorough vetting has already been completed). This is all happening just as newsrooms are being stripped of resources, but competition for clicks is fiercer than ever. And when many journalists haven’t received the necessary training in the forensic analysis of digital sources or content, it’s easier than ever to be fooled.

The flipside is also becoming increasingly easy. As an agent of disinformation, if you don’t manage to succeed in tricking journalists to reference or link to false information, try getting them to debunk it instead. A debunk is better than nothing. It provides oxygen, and oxygen means more people searching for your strategically crafted keywords online, and more people searching means more people discovering your extended networks, all pushing your ideas, narratives or beliefs. When your audience is connected, simply driving search behavior can open up your world to thousands of new eyeballs.

Doing journalism in an age of disinformation is incredibly hard, and I simply don’t think the news industry has even started grappling with the difficult questions being raised in the current environment. Here are five lessons that I hope can at least act as conversation starters in newsrooms.

1. Be prepared: train your newsrooms in disinformation tactics & techniques

In 2019, make sure your training plans include sessions on digital verification skills, but ensure they’re not just focused on assessing whether something is true or not. Make sure the participants are trained on digital provenance, so they can track where a piece of content originated. Similarly, ensure that your journalists understand how to do this type of work safely. Have they been trained in carrying out work in anonymous spaces online? Do they have high levels of personal digital security, in terms of their privacy settings, use of VPNs, etc.? Does your newsroom have ethics guidelines about sourcing information from closed or anonymous spaces online? Be responsible: Don’t give disinformation additional oxygen. First Draft is expanding its existing verification curriculum to cover more of this information and training in Spring 2019.

2. Be responsible: Don’t give disinformation additional oxygen

Our work suggests that there is a tipping point when it comes to reporting on disinformation. Reporting too early gives unnecessary oxygen to rumors or misleading content that might otherwise fade away. Reporting too late means the falsehood takes hold and there’s really nothing to do to stop it (it becomes a zombie rumor — those that just won’t die).

There is no one tipping point. The tipping point differs by country but is measured when content moves out of a niche community, starts moving at velocity on one platform or crosses onto other platforms. The more time you spend monitoring disinformation, the more the tipping point becomes clearer, which is another reason for newsrooms to take disinformation seriously. It is also a reason to create informal collaborations so newsrooms can compare concerns about coverage decisions. Too often newsrooms report on rumors or campaigns, for fear that they will be “scooped” by other newsrooms, when again, this is exactly what the agents of disinformation are hoping for. Having every newsroom publish a QAnon explainer back in August after people turned up at Trump rallies with Q signs and t-shirts was exactly what the Q community had hoped would happen.

3. Be aware: Understand the implications of a networked audience

This is connected to the point above. Debunking or explaining these conspiracies, falsehoods or rumors, gives them not only legitimacy, but a set of keywords for your audiences to use to search for more information. Even relatively small, disparate communities can appear significant online. Before the internet, such remote communities struggled to connect because it was so difficult to meet face to face. Now such communities can flourish. (For an excellent report into the power of keywords to shape reality, read Francesca Tripodi’s Searching for Alternative Facts).

4. Explain: Don’t act as stenographer

When so many people get their news from just tweets, Facebook posts, as headlines on Google News, or push notifications, the responsibility for how headlines are worded is more important that ever. It doesn’t matter if the 850 word article provides all the context and explanation to debunk or explain why a narrative or claim is false, if the 80 character version of that context is misleading, it’s all for nothing. Academic research shows that simply repeating the falsehood in the headline is problematic. Finding alternative ways to word a headline is difficult. We face this problem frequently at First Draft, but we have to be smarter about the ways we word headlines, tweets and posts.

5. Inoculate: Do more reporting that helps explain the issues that are often subjects of disinformation campaigns.

Rather than simply reacting to the falsehoods when they appear, any reporting that preempts some of the most common and powerful narratives might help inoculate some of the disinformation. From our analysis of information in the lead up to the Brazilian and U.S. midterm elections, we know that the following topics are most common: attempts to undermine the integrity of the election system; attempts to sow hate and division based on misogyny, anti-semitism, islamophobia, and homosexuality; attempts to demonize immigrants; conspiracies about global networks of power.

Reporting in age of disinformation is difficult. Academic research on best practices for writing headlines is still being published and we have yet to reach firm conclusions. Newsrooms are struggling to find money for training. Most senior editors, many of whom developed their journalistic skills before digital, have no idea of the risks faced by their journalists every day. But the decisions being made by newsrooms around the world impact how disinformation spreads. The professional media is a critical part of the information ecosystem, and currently it is vulnerable and being used to spread false and misleading information.

I hope 2019 is the year when newsrooms start discussing how to take concrete actions to fight back, whether that is rolling out a new training course, writing new ethics and standards documents, or simply including a discussion in editorial meetings of the potential impact on the audience of reporting on disinformation.

5 Lessons for Reporting in an Age of Disinformation was originally published in First Draft Footnotes on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Sérgio Lüdtke (Versão em português brasileiro)

Competition can be a great motivator, but a shared purpose for all is better, which is what I discovered after 12 weeks of working collaboratively on Comprova. During this period, journalists from 24 competing media companies worked together to investigate the veracity of dubious claims and content that circulated on social networks during the Brazilian presidential campaign of 2018.

The fluidity of collaborative work depends on engagement of the organizations taking part and the clarity of the project’s goals. With Comprova, having the editorial processes defined and a project scope was instrumental in creating a solid foundation for our work together.

Before the start of the project, journalists gathered for a 2-and-a-half-day bootcamp where they received training and met their new collaborators. Working together like this is a fundamental step for collaborative work, especially if that collaboration relies on teams that will work remotely. Personal contact helps to create relationship networks, but above all it is an opportunity to reveal affinities and reinforce the bonds of trust that proved to be necessary for Comprova.

At the bootcamp, Comprova journalists learned about network-monitoring methodologies and were trained to use a range of tools like CrowdTangle, NewsWhip’s Spike and Torabit, that allowed them to search for questionable content during the campaign. Face-to-face training with hands-on activities and group-work is necessary to strengthen relationships and to promote an exchange of experiences. Verifiers complemented their training with several hours of online courses, but it was the time when we were together that most contributed to partner engagement during the project.

Back in the newsrooms, communication between the group was vital to the project. The initial intent was to use Slack, however, Brazilians newsrooms have not heavily adopted the platform and having other tools to learn — the CMS, CrowdTangle, NewsWhip, etc. — the group opted for the familiar platform WhatsApp to communicate with each other about debunks and what we might work on together that day or week. WhatsApp does not allow for threaded messages so conversations can be lost and search can be clunky. Even so, the group generated 18,500 messages during the project, so while WhatsApp was not ideal for partner communications, it did work for the purposes of our collaborative project.

Obviously, at times, when there was more than one subject under discussion and many people involved in the conversation, the communication became a bit confusing. During the period that we investigated a video that showed “Bolsonaro 2018” written in the fields of a plantation, we exchanged 545 messages in the group in a single day.

In collaborative work, many conversations are generated. In the case of Comprova, we chose the management model of consensus, which encouraged partner participation. In this horizontal model, it was important that everyone’s opinion was respected and that space for disagreements was provided. In the project, verifiers were encouraged to comment, intervene, and propose solutions even for verifications in which they were not directly involved. The decision process adopted does not necessarily require more time since consensus is often not reached by a vote, but rather can be simply observed. A consensus model does require mature decisions, which is most welcome in investigative work that aims to be rigorous.

Traditional newsrooms have more rigid hierarchies with more vertical decision-making processes. A consensus model tends to give more value to the knowledge, opinions and participation of all. A survey of the verifiers at the end of the project showed that almost 90 percent of them did not have prior collaborative networking experience. All said they were satisfied with the consensus management model and with collaborative work, and almost 80 percent said they were fully satisfied with the level of collaboration from colleagues.

The openness to participation of all implies more commitment, distributes the responsibilities and creates a greater zeal for the final result. With Comprova, engagement among partners helped to ensure the accuracy of our work and to more easily achieve the zero-error goal together.

The scope and objectives for Comprova were clear from the beginning, and questions about any doubts were quickly solved. With transparency and openness to participate, the verifiers, in addition to collaborating with colleagues, became auditors for the objectives of Comprova.

(Versão em português brasileiro)

Previous articles about Comprova:
- Waves of disinformation in the Brazilian elections
- What WhatsApp “API access” meant for Comprova
- Comprova wraps in Brazil
- In an age of misinformation, sustainable newsroom collaborations should become a priority

How collaborative journalism worked with Comprova was originally published in First Draft Footnotes on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Sérgio Lüdtke, (Versão em português brasileiro)

Misinformation in the 2018 Brazilian presidential campaign moved in waves and like the tides obeyed the gravitational force of the public agenda. For 12 weeks, I sailed in a sea of ​​misinformation trying to trap rumors and lies that emerged on the web with astonishing speed and strength.

In three months, a team of two editors and I, along with reporters from 24 newsrooms, participated in Comprova, a coalition of media outlets that verified and debunked rumors and lies circulating during the Brazilian presidential campaign. We found these rumors by monitoring social networks in search of problematic content that might damage the electoral process.

During the project, I saw rumors and lies move like waves. After flooding an environment, the same rumors recoiled to return later with force, to reach different spaces and people. The currents that commanded this flow followed the public agenda and the moral issues that dominated online conversations. For example, fake election polling content appeared shortly before and after the real poll. Its creators had an environment where there were low levels of voting information; just enough that people knew that a new poll had taken place but without knowing its results. But when lies circulating among friends and family in closed WhatsApp groups, that lie can be more easily accepted.

We saw increasingly sophisticated content that was created to misinform the public during the project, which took two paths: complexity and simplification.

There was a complex conspiracy theory that reached more than 1.5 million views in just over 24 hours via a viral video. In the video, a man doubts the result of the elections and says that he will announce the results based on a mathematical equation. For six days, journalists from four Comprova partner organizations investigated the video and concluded that it was impossible to determine fraud from the equation presented in the video. The Comprova report was published one day before the date promised by the man in the video for disclosure of ballot fraud and, we hope, reduced the impact of the rumor.

Some rumors used simplified tactics, like lowering the image resolution to make verification more difficult and to contextualize the information differently. One rumor used a video that was recorded two years earlier during demonstrations against President Dilma Rousseff’s impeachment. The video was used to illustrate a claim that a car had been destroyed by Workers Party protesters for having a sticker supporting Jair Bolsonaro. The resolution was reduced and only red flags and a crowd around a white car was identifiable in the video. The video was enough to offer “proof” to support the false narrative. In this case, that seemed to come to a dead end, we were lucky and found the original content in a single Google search.

The other sophisticated element to these rumors is the absence of verifiable information: the creation of lies by suggestibility and no online content. A rumor that Comprova was unable to verify was that “a large weekly magazine” had received 600 million in bitcoins to destroy leading candidate Jair Bolsonaro. By insinuating this rumor, it then made the first magazine to publish an unfavorable report about Bolsonaro considered to be the recipient of the bitcoin. This bitcoin rumor contained no actual content so our team could not prove it to be false.

There were a few dozen other dubious claims checked by Comprova that were never published because it was impossible to reach a definitive conclusion. Even so, the team improved its ability to identify and accurately verify misleading content, which allowed us to publish 147 stories in 12 weeks. Most of the stories published — 135 — were found to be false. In some cases, we managed to stem the spread of rumors; in others, the rumors returned in different formats. Unfortunately, most of the misleading content Comprova debunked remains online.

There are still serious challenges for any team that investigates rumors:

• Respect for privacy and freedom of speech prevents us from finding the DNA of content that travels in apps like WhatsApp, which is capable of creating disinformation tsunamis.
• Finding original content and establishing provenance is the most challenging problem for journalists working on verification.
• In Brazil, candidates and parties create their own verification sections on campaign sites and some rumors debunked by Comprova were “denied” by candidates and parties. These denials eventually gained strength and made us aware that “artificial victimization” may be used in the future in favor of candidates.

Constant monitoring, a more active stance by journalists to intervene when necessary, and the sharing of experiences on projects like Comprova will be the key by which together we can reach a more healthy and controlled information environment.

(Versão em português brasileiro)

Previous articles about Comprova:
- What WhatsApp “API Access” Meant for Comprova
- Comprova wraps in Brazil
- In an age of misinformation, sustainable newsroom collaborations should become a priority

Waves of disinformation in the Brazilian elections was originally published in First Draft Footnotes on Medium, where people are continuing the conversation by highlighting and responding to this story.

The collaborative journalism project was the first to have special access to one of Brazil’s most popular apps

By Aimee Rinehart, (Versão em português brasileiro)

Our collaborative verification project, Comprova, brought together 24 newsrooms in Brazil to debunk mis- and disinformation around the national elections that took place in October 2018. First Draft wanted to facilitate a project in Brazil because it’s the largest democracy in the Global South and is becoming an increasingly polarized country. We also wanted to take a closer look at how mis- and disinformation is exchanged on the closed messaging application WhatsApp. Brazil has had a massive adoption of WhatsApp with an estimated 120 million people on the platform.

API sounds fancy but what did it actually mean for Comprova? For the first time, WhatsApp gave a journalism non-profit the ability to receive questions from the public on an intermediary website, Zendesk. Had we just used WhatsApp for Business, we would have been tethered to one phone to try and respond to the almost 70,000 questions the project received. Zendesk was integrated into our business account and allowed multiple journalists and student journalists to review “tickets,” or questions from the public. For most of the project, four journalists monitored questions. For the last month of the election, there were 10 people assigned to sort and sift tickets. Through this process, the team began to see patterns of mis- and disinformation and sent these on to editors for consideration to debunk. The team also replied to the person who sent the original message with a simple note or a link to the published article on the topic.

An early-in-the-project dashboard for Zendesk, which allowed the Comprova team to sort “tickets” or questions from the public.

On the days surrounding the first election on October 7, voting integrity was a concern among people asking questions to the Comprova WhatsApp number. Photos of ballot voting receipts declaring a winner early began to circulate online. Based on public interest and noting a possible trend to misinform voters, the Comprova partners collaborated on verifying the authenticity of the photo and published a report to clarify photos of the ballot. The photos were authentic, but the “winner” described on the ticket was for the regional area only, and not the final general-election outcome. The team responded to questions about this voting receipt with a link to the Comprova debunk, (pictured left and translated by Google Translate). Within Zendesk, we created a “macro” (pre-formatted responses stored in a drop-down menu) so with one click, we could send the debunk to each person who asked about the ballot papers; this response was sent 900 times to different people who asked about it.

The collaborative project CrossCheck France ran for 10 weeks in spring 2017 and received 600 questions from the public by email from a submission form on the website. We were warned by Brazilian journalists that we could expect to see 600 questions in the first hour from Brazilians. They said so without a laugh and they were right about an engaged citizenry: Comprova received almost 70,000 questions from the public. During the first week of October leading up to the first election, the business account received an average of 3,000 questions daily. The Zendesk project manager reported that the platform typically receives 2,000 questions per month for its most-active clients, so Comprova struck them as a particularly “viral” experience.

We’ve had a learning curve figuring out the capabilities and limitation of a WhatsApp for Business Account. For example, there is no status update nor the ability to have a scrolling feed of information that would make it easier for people to share. Depending on the user path from the project website, the person is presented a field to either sign up to project updates or ask their question. We aimed to respond within 24 hours but we received so many messages our average response time was about 36 hours. While working with Zendesk to establish this integration, many of the analogies were related to airlines, as they have been the largest adopter of the WhatsApp Business Account API. These airline analogies also indicate that we need to develop technical integrations designed explicitly for journalism challenges.

Representatives from WhatsApp and Zendesk checked in, often on the weekend, to see how we were getting along with the API. The team is clearly aware of the challenges of misinformation circulating on the platform and sees projects like Comprova as one way to help.

There is no way the Comprova team could have addressed the almost 70,000 tickets on a single phone, so the WhatsApp API was critical to the ambitions of the project, which was to inform the Brazilian electorate. We hope Comprova will be back in 2019, but in the meantime we’re using the same WhatsApp API integration with our newest project, CrossCheck Nigeria. We suspect WhatsApp traffic in Nigeria will rival Brazil, but whatever happens, the learning continues.

(Versão em português brasileiro)

Comprova website

Previous articles about Comprova:
- Comprova wraps in Brazil
- In an age of misinformation, sustainable newsroom collaborations should become a priority

What WhatsApp “API Access” Meant for Comprova was originally published in First Draft Footnotes on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Claire Wardle (Versão em português brasileiro)

It’s not often that when you hear seasoned journalists admit something has rekindled their love for the craft. This year has been very difficult for journalists globally, and most of what we hear relates to journalists under physical, verbal and online attacks from politicians and trolls.

So earlier this month when a stream of WhatsApp messages from Brazilian journalists popped up on my screen, full of fist pump emojis and love hearts, it was clear something was going on. The messages were being shared in a WhatsApp group used by the 62 journalists who powered our collaborative election project, Comprova. On a daily basis, the group shared hundreds of messages as they decided what pieces of misinformation to debunk, how to verify the rumors, doctored photos and misleading memes, and how to word headlines responsibly. But as voting ended and the project concluded, the journalists started sharing with each other what the project had meant to them personally. And the messages were astonishing.

These “debunking” projects are often judged by whether or not they slow down or stop misinformation. Understanding this impact is almost impossible to do without data from the platforms, but we are in the process of measuring and evaluating the impact of the project using data we collected (more about this in a subsequent post).

But as well as the impact on the audience, we also need to acknowledge what these projects do to a news industry. This type of “CrossCheck” project was piloted in France during the election in Spring 2017. Our post-project evaluation showed that the project had significantly improved verification skills in newsrooms large and small, but perhaps more importantly, built significant trust between journalists in normally competitive newsrooms. Comprova has shown similar results.

You can read more about the details of Comprova here, but the basic premise is that journalists from 24 Brazilian newsrooms worked collaboratively every day to find, debunk and report on different rumors, false and misleading content circulating ahead of the Brazilian elections. Slack hasn’t taken off in Brazil so the reporters collaborated in a WhatsApp group they called the Comprova Verificadores (the nickname they gave themselves). The group was so busy that on some days there were more than 1,000 messages shared. One journalist would write up the report, and the other journalists were asked to then “crosscheck” the reporting. If they agreed with the quality of the work, they added their newsroom’s logo, and once three newsrooms had done this, the report was published on the central site.

The magic of the CrossCheck model, is that while Comprova did build up somewhat of a following, over a three month period we could not (and should not) compete with the newsrooms who were part of the coalition. So along with the report being seen on the Comprova website, partners that were happy with the “crosscheck” published their own versions on their sites, or broadcast stories on their own radio or television channels. For example, the central Comprova website had about one million unique page views. Just one of our 24 partners reports traffic of almost 7 million uniques.The amplification machine was powerful.

In total 147 reports were published on the main Comprova site and many many more were published via the partners (we’re in the process of collecting the final number). Jornal do Commercio, based in Recife in north eastern Brazil, published a collection of the debunks in a weekend edition several times during the project replicating the site for the print audience.

One of the elements that the journalists are most proud of is that not one correction was made during the project. As we saw in our previous project in France (which similarly made no errors in 10 weeks), forcing newsrooms to work together increases standards. Being held to account by your editor is one thing but being held accountable by your peers in competing newsrooms is another.

Having these cross-newsroom communication channels also led to in-depth discussions about the ethics of covering a particular piece of content, or how to frame the headline. We saw the same conversations play out in France. Doing journalism in an age of misinformation is increasingly challenging. Newsrooms are having to figure this out as they go.

In the U.S., newsrooms make poor judgments daily in terms of giving unnecessary oxygen to conspiracies or fringe personalities, or wording a headline or tweet in a way that reinforces a falsehood. These often result in a Twitter eruption of indignation but little changes. These types of supportive inter-newsroom channels allow for thoughtful consideration of these complex issues and new challenges.

The overwhelmingly positive experience for those journalists who took part in Comprova means that most want the project to continue. We will be traveling to São Paulo this week for a well-earned “wrap party” and to make plans on how to establish Comprova 2.0 for an ongoing collaborative effort to fight misinformation in Brazil. Unsurprisingly, misinformation didn’t stop on election day.

This project was ambitious and at times we were close to giving up. There were many sleepless nights and five overnight flights to Brazil. But it has all been worth it. Comprova has created a sustainable response to misinformation, created a team of journalists who could now train other journalists on how to discover, verify and responsibly debunk misinformation, and for some it has rekindled their love for journalism.

And for the academics in the room, we now have the most incredible misinformation datasets, which has always been the point of these projects. We will ask in our research how can we do good work in the field, but work that ultimately helps the academic community learn more about the information ecosystem across platforms?

This election project is the sixth one we’ve created; we have tried different methodologies and have learned a great deal. Our key takeaway after №6 is: While elections are the time when funders and newsrooms get excited about what is possible, we need to build ongoing sustainable collaborations in preparation for when the next elections come. Misinformation is not just a problem during election seasons. We need to provide newsrooms with training on social discovery, verification, and responsible reporting techniques when it comes to misinformation. We then need to establish ongoing collaborative projects to work so those newly found skills don’t die, and we need to build networks of “ambassadors” willing to spread quality content in WhatsApp groups. We have a lot of work to do, but journalists in France and Brazil in particular have shown what is possible.

Even I have had my enthusiasm for this topic rekindled. And it’s been a tough year!

(Versão em português brasileiro)

Comprova website

Previous articles about Comprova:
- Comprova wraps in Brazil
- What WhatsApp “API Access” Meant for Comprova

In an age of misinformation, sustainable newsroom collaborations should become a priority was originally published in First Draft Footnotes on Medium, where people are continuing the conversation by highlighting and responding to this story.

By Aimee Rinehart (Versão em português brasileiro)

Comprova, our largest collaborative journalism project to date, wrapped after the second round of voting in Brazil on October 28. The project brought together journalists from 24 newsrooms, who worked together virtually, using a shared CMS and an incredibly busy WhatsApp group, to find, verify and collaboratively report on 149 examples of misinformation over the ten weeks leading up the second runoff for the Presidential election.

Throughout the project, against a tidal wave of misinformation and several roadblocks, our partners persevered. From the very beginning of the project, there were challenges. The kick-off bootcamp, taking place in the Brazilian countryside coincided with a fuel strike and gas shortage. In addition, the 10 weeks of the project was marked by a number of breaking news events. The team worked early mornings and late into many nights working on Comprova as well as covering these big stories:

August 15: Grupo Abril, a large media group, which included three Comprova project partners, filed for bankruptcy protection and laid off 900 employees.

August 31: the top electoral court ruled that former president Lula da Silva could not run for a third term for president while serving his 12-year sentence on corruption.

September 2: A huge fire ravaged Museu Nacional.

September 6: Jair Bolsonaro was stabbed at a campaign rally.

Over the course of the 12-week project that started August 6, partners selected examples of misinformation that had passed the tipping point of virality (the team measured this using different tools to assess whether a rumor or piece of misleading or fabricated content had jumped out of the online community where it started, or had crossed platforms and was moving with velocity.) One of the greatest challenges of these projects is working out this tipping point. If you report on a small rumor, you give it unnecessary oxygen. If you leave it too late, it’s very difficult to slow down or stop.

One partner would begin a verification report, documenting the checks they had made, the tools they used and the evidence they had gathered. Then the project editor would send a message to the other partners via the main WhatsApp group and other partners would review the work and confirm or offer suggestions on any additional information that was needed. To publish a report, at least three partners had to confirm the findings and assign the organization’s logo to the bottom of the report in the CMS. On average there were about 9 partner logos listed at the bottom of the reports.

WhatsApp has very high penetration in Brazil and we knew that it would play a central role for Comprova. We are proud to be the first journalism non-profit to work with WhatsApp using enterprise software, which enabled us to sort through the almost 70,000 questions and tips we received from the public asking about content online they found questionable. For the last month of the election, we were receiving almost 2000 messages per day which meant we had 10 journalists and journalism students reviewing these questions, finding patterns and passing along the most pressing questions to the central team.

Comprova was very much a project driven by our Brazilian partners. We provided early training on the tools, the methodology and best practices for reporting on disinformation, but the project was a success because of the dedication and enthusiasm of our partners. We want to say a huge thank you and to name all of those who were involved.

Comprova had an editorial team of five that worked out of the offices at the Fundação Armando Alvares Penteado campus. Sérgio Lüdtke led the editorial direction on the team. He was supported by part-time associate editors José Antonio Lima and Rafael Garcia. For the first time on our collaborative project we hired part-time spots for a social media creator and social media strategist. We had three social media creators during the project: Marcela Tavares, Ricardo Pieralini and Helio Miguel Filho. Our social strategist, Guilherme Conter, worked off site and with our web development team on SEO and ensured that all of those tracking pixels were in place for the project. Pedro Burgos pitched in to establish our WhatsApp number and continued supporting the project by creating tools to help us analyze Comprova’s reach. We were lucky enough to have Harvard Kennedy School Digital Fellow Yasodara Cordova with us in the U.S. to work on questions send in to our WhatsApp number; Yaso created shortcuts, trained journalists on how to respond and kept us all apprised at the types of questions and rumors the number received. These are the people who on a daily basis kept Comprova running smoothly and made sure our WhatsApp number was promoted to a wide audience.

There has been help and encouragement on the ground by so many of the top journalism leaders in Brazil. Adriana Garcia has worked with us since February to bring partners together, help plan events and keep the project going when it looked — as it often does during collaborative projects — like it might all fall apart. Angela Pimenta, director of Projor, was one of our first points of contact and she said about this project, “when I decide on something, I decide.” It was that type of focus that kept us all moving forward together. Daniel Bramatti, a journalist at O Estado de S. Paulo and president of Abraji, has been a project stalwart. He kept the partners focused on journalism fundamentals and offered Abraji to be the coordinating organization on the ground, which helped tremendously. There were also a number of organizations that lent their technology or services to Comprova. Institutional Partners include Abraji, ANJ (National Association of Newspapers in Brazil), Aos Fatos, FAAP, Ideal H+K Strategies, the Brazil office of Harvard University’s David Rockefeller Center for Latin American Studies, Projor, RBMDF Advogados, Eleições Sem Fake and Textual Comunicação. Technology Partners include Chartbeat, CrowdTangle, NewsWhip, Stacker, Torabit, Twitter and WhatsApp. Google News Initiative and Facebook Journalism Project provided the funds to make this project a reality and for that, we are forever grateful.

There is ample evidence that mis- and disinformation will continue to propagate in Brazil and collaboration is the only way to combat this problem. Comprova is committed to seeing what is possible for a 2.0 version in 2019. We hope you’ll continue to follow our Twitter handle and Facebook page for project updates.

(Versão em português brasileiro)

Comprova website

Previous articles about Comprova:
- In an age of misinformation, sustainable newsroom collaborations should become a priority
- What WhatsApp “API Access” Meant for Comprova

Comprova wraps in Brazil was originally published in First Draft Footnotes on Medium, where people are continuing the conversation by highlighting and responding to this story.

The July rescue of 12 Thai soccer players and their coach trapped in a cave attracted global attention, with many photos and videos of the rescue shared millions of times. The video to the left— ostensibly of the team’s rescue — was viewed widely across online platforms. However, it is actually footage of a cave dive in Wisconsin from 2012. Storyful, a First Draft founding partner and verification agency detailed fakes like this and others that emerged about the Thai cave rescue. But how many journalists and journalism students working today know the steps to take to verify and identify the original footage?

First Draft’s online courses are built to bridge the knowledge gap some newsrooms and classrooms have with a 5-unit course for working journalists and college classrooms, and a 1-hour course for newsroom managers and others interested in verifying information online. The courses are currently available in English, Spanish and Brazilian Portuguese; by the beginning of next year, the courses will also be in Arabic, French, German, Indonesian, Italian, Japanese, Korean, and Russian languages.

Take and teach the courses

Dr. Carrie Brown, director of the Social Journalism master’s program at City University of New York, used First Draft materials in her class on advanced reporting. She said that the interactive exercises helped students prepare for real-world projects like Electionland and Documenting Hate.

“It’s hard to find a really good example that students can work with that isn’t so hard, but also not super easy,” said Brown. “With this course, they can actually get the practice going through the checklist and doing all the steps.”

Associate Professor at Francisco de Vitoria University Myriam Redondo has also used First Draft’s observation challenges in her course on information management in social networks. She said that, like Dr. Brown, she struggles to find real-life examples in which her students can practice their verification skills.

“Maybe one day you see a story you like, but then when you take it to the classroom everyone knows it’s a hoax,” Redondo said. “I tried First Draft interactives and then used them in my classroom.”The online courses offer a combination of video tutorials, case studies and hands-on activities. They are designed to provide working journalists, professors and students with a solid foundation in image and video verification, and detail tools needed to investigate the authenticity of a social account.

Instructors can enroll all of their students and track student progress. Alternatively, they build their own course from First Draft resources; each unit has resources that can be downloaded individually and incorporated into personal curriculum. Organizations have used the discussion prompts for brown-bag lunches. First Draft also offers video conferences for newsrooms and classrooms to support and extend course learnings.

All of First Draft resources are free to use and incorporate into any trainings and are available under Creative Commons license.

Take and teach the courses

By Claire Wardle and Aimee Rinehart

São Paulo, August 6 — First Draft, a project of Harvard Kennedy School’s Shorenstein Center is thrilled to announce the first day of publishing for Comprova, a new collaborative journalism project that brings together 24 newsrooms in Brazil to verify misinformation online. Project partners will work together to authenticate information related to elections in Brazil, which take place on October 7 and 28 this year.

Over the past six months, First Draft has been working closely with Abraji, the Brazilian Association of Investigative Journalists, to provide training and project design support. Abraji is leading Comprova in Brazil, and the impact of the collaboration will be studied by a team of researchers at the John F. Kennedy School of Government at Harvard University.

Comprova will be the first journalism consortium to use the WhatsApp Business API, which will help reporters at newsrooms across the country review and respond to incoming messages at scale. The WhatsApp number is +55 (11) 97795–0022, and the public will be prompted to submit tips about confusing or false content related to the election.

The WhatsApp component is new for election projects facilitated by First Draft, which previously designed CrossCheck, its award-winning collaboration focused on the 2017 French election. The WhatsApp number will accept questions and tips from the public, and will facilitate the detection of trends in misinformation reported around the country. Researchers at the Harvard Kennedy School will be investigating how misinformation circulates on WhatsApp in the lead up to the election, and will be testing effective debunks on the platform. The findings will help inform newsroom best practices for requesting tips about misinformation as well as disseminating debunks to the public via WhatsApp.

“With about 120 million Brazilians using WhatsApp, it is the primary communication platform for most people, and as a result all kinds of information flows across the platform, including misinformation,” said Claire Wardle, who leads First Draft at the Harvard Kennedy School’s Shorenstein Center. “Comprova gives us a real opportunity to understand the role WhatsApp plays in Brazil, particularly how voters use it as a source of information during an election campaign.”

Comprova means “proof,” or when written as “com” and “prova,” it means “with proof” in Brazilian Portuguese. The project’s aim is to identify and defend against sophisticated manipulation and amplification techniques online. The workflow has been designed deliberately to encourage collaborative investigation among newsrooms that will continue beyond the election. The project website is optimized for a mobile audience, and report templates have been designed so they can be easily shared and consumed on social platforms, including WhatsApp. While it might not be possible to answer all questions posed on WhatsApp, Comprova journalists will strive to address content that has the potential to misinform a large proportion of the electorate.

The website — the engine that powers the collaborative element of Comprova — has a proprietary content management system designed specifically for this project, and will be used in future election projects around the world. Comprova is distinct in that no report can be published until three separate newsrooms agree with the verification work behind it, a process known as a ‘CrossCheck’.

The public can see the reports on a central Comprova website that has been designed to make it easy for people to share the content produced by participating newsrooms on Facebook, Twitter, and WhatsApp. Individual partners will also produce longer reports on their own sites. All information is published on the website under a Creative Commons License to encourage other publications to use the information — in whole or in part — with credit to Comprova.

Comprova Information:
WhatsApp: +55 (11) 97795–0022
projetocomprova.com.br
facebook.com/ComprovaBR
twitter.com/comprova
youtube.com/comprova

Newsrooms involved in Comprova include: AFP, Band News, Band TV, Canal Futura, Correio do Povo, Exame, Folha de S.Paulo, GaúchaZH, Gazeta Online, Gazeta do Povo, Jornal do Commercio, Metro Brasil, Nexo Jornal, Nova Escola, NSC Comunicação, O Estado de S.Paulo, O Povo, Poder360, Rádio Band News FM, Rádio Bandeirantes, Revista Piauí, SBT, UOL and Veja.

Comprova is being coordinated by Abraji, a Brazilian investigative journalist association, and has the support of Projor, which reports on the media and works to strengthen journalism in Brazil. The Google News Initiative and Facebook’s Journalism Project have helped to finance the project, and both companies are providing technical support and training for the teams involved.

Institutional Partners include Abraji, ANJ (National Association of Newspapers in Brazil), the Brazil office of Harvard University’s David Rockefeller Center for Latin American Studies, Projor and RBMDF Advogados. Technology Partners include CrowdTangle, NewsWhip, Torabit, Twitter and WhatsApp.