How did we report this and what standards did we use? Read our Ad Decoder methodology
The Miami Herald is collecting and analyzing political advertisements and sponsored content relating to the 2020 election.
The Herald’s Election Ad Decoder will tell you where an ad came from, who paid for it and any other information that may be useful to Herald readers — such as whether ads or their associated websites collect personal data. Additionally, the Herald will flag ads that spread disinformation, use hateful speech, promote violence, display manipulated images, or are linked to foreign attempts to interfere with the election.
Herald reporters define “disinformation” as false, misleading or decontextualized information intentionally produced to deceive its audience. For the purposes of this project disinformation breaks down into three, sometimes overlapping categories: false content, misleading content and manipulated images. These three categories are defined to align with the ways most people see and understand media content.
Intentionality is a key component of disinformation, as opposed to false information spread organically through ignorance (often called “misinformation.”)
The Herald considers all sponsored/promoted material to be intentional. Therefore any ad/sponsored content that presents untrue or misleading material constitutes disinformation, whether the errors were intended to deceive or were the result of negligent fact-checking.
“If a PAC or a campaign — which are very knowledgeable about elections — are spreading veritably false information, we know how dangerous that can be in suppressing voters, intimidating voters and misleading voters,” said Dipayan Ghosh, a researcher at the Shorenstein Center on Media, Politics and Public Policy at the Harvard Kennedy School.
The aim of the project is to provide a useful service to our readers as they determine what is true and false as a wave of information floods voters before the election.
The Herald worked with the following experts to develop the concept of the Election Ad Decoder. They will be consulted on an ongoing basis through the election.
Susan Benesch, American University
Emerson Brooking, Atlantic Council’s Digital Forensic Research Lab
Hany Farid, University of California, Berkeley
Dipayan Ghosh, Harvard Kennedy School
Frances Hill, University of Miami
Siwei Lyu, SUNY Buffalo
Mary McCord, Georgetown University
Tanu Mitra, University of Washington
Kevin Munger, Penn State University
Jonathan Nagler, New York University
Brendan Nyhan, Dartmouth College
Ciara Torres-Spelliscy, Stetson University
Joseph Uscinski, University of Miami
Claire Wardle, First Draft
Collecting Ads/Sponsored content
Crowdsourcing
The Herald collects data from readers in three major ways. All are being promoted on the Herald’s social media platforms in both English and Spanish. The submission platforms allow readers to submit screenshots of the ads and/or forwarded emails. Reader submissions are our highest priority to research and enter into the database.
1) A Cognito form embedded in online stories published by the Miami Herald and el Nuevo Herald.
2) A specially created email address: ElectionAdDecoder@miamiherald.com.
3) A Google voice number that can accept text and WhatsApp messages: 305-396-1481.
Self-entry
Reporters also collect political ads that they encounter during their normal Internet browsing or are delivered to them through the mail, as well as ads that gain attention through news coverage.
Facebook API
Reporters pull sponsored content targeting Floridians from the Facebook Ad Library.
Inclusion criteria
The Election Ad Decoder includes “Ads/sponsored content” seen by Floridians that meet one of the following criteria:
Any paid content about the 2020 election, which includes content that mentions candidates or parties involved in federal, state or local races, and is disseminated either via the mail or on a platform not owned by the group that paid for the ad/sponsored content.
Platform types include social media, email, text messages, Whatsapp message, newspapers, television, radio robocalls, and programmatic ads.
Any social media content that purports to be independent but is in fact attributable to foreign governments and addresses the election, even if the content is not promoted/boosted on social media.
For instance, if a group secretly linked to a foreign government posts a news article to Facebook without paying Facebook to promote it, the Herald considers that post an “ad/sponsored content” because the foreign government was ultimately responsible for its dissemination.
Content Flags
Herald reporters will analyze each ad and apply the following warning signs where necessary.
False content
False content meets one of the following criteria:
A verifiable claim (rather than an opinion) that is entirely untrue.
An ad/sponsored content that has been digitally altered to show something that is not real/didn’t happen.
Misleading Content
Misleading content flags are applied to ads where information/content is presented in such a misleading, decontextualized or manipulated manner that it clearly implies a false or untrue conclusion.
Misleading: a series of partial truths that strung together imply something that is untrue.
Decontextualized: when claims, statistics or partial quotes are provided without proper context in order to imply something that is untrue.
Deceptively Framed: An image that has been cropped or stripped of its context in a way that changes the meaning of the original image.
Mislabeled: An image that has been mislabeled as to the date, time or location it was taken in a way that is misleading.
Procedure for flagging misleading content
If the material appears to meet the criteria of misleading content, Herald reporters will send the ad/sponsored content to our panel of disinformation experts. Each expert previously agreed to the Herald’s definition.
The experts evaluate whether the “misleading” flag is appropriate based on our definition and a detailed reporting/analysis that we provide.
If the majority of respondents agree the ad/sponsored content is misleading, the ad is flagged as “misleading content.” If only one expert responds, their judgment will be applied. Herald reporters will never apply this flag without an expert opinion.
Mislabeled/decontextualized images
If an image has been mislabeled as to the date, time or location it was taken in a way that changed its original meaning, reporters send the image to experts to decide whether the mislabeling has fundamentally altered the meaning of the image.
Experts will decide if the “misleading content” flag is appropriate, based on our definition and a detailed reporting/analysis that we provide.
If the majority of respondents agree the ad/sponsored content is misleading, the reporter will flag it as “misleading content.” If only one expert responds, their judgment will be applied. Herald reporters will never apply this flag without an expert opinion.
Manipulated Image
Manipulated images are defined as meeting one of the following criteria:
An image that has been digitally altered in a way that changes the meaning of the original image.
Procedure for flagging manipulated images
If the reporter suspects an image had been digitally altered but cannot make that determination themselves, they will send the ad/sponsored content to our panel of manipulated image experts.
Experts evaluate whether the image has been digitally altered in a way that changes its original meaning. The experts will decide whether the “manipulated image” flag is appropriate based on our definition and a detailed reporting/analysis that we provide.
If the majority of respondents agree that the ad/sponsored content meets our definition of “manipulated image,” the reporter will flag the ad/sponsored as “manipulated image.” If only one expert responds, their judgment will be applied. Herald reporters will never apply this flag without an expert opinion.
Hateful Speech
“Hateful speech” is content that expresses hate against a group of people or against an individual as a member of a group of people not because of their political ideas or activities but based on their religion, ethnicity, nationality, race, color, descent, gender, sexual orientation, immigration status, disability or other immutable characteristics. Such speech can include calls for discrimination or incitement to violence against others on the basis of identity.
Such speech must go beyond the merely pejorative...i.e., “[People of a certain nationality] are boring” does not qualify as hateful speech.
Procedure for flagging hateful speech
If the material appears to constitute hateful speech, Herald reporters will send the ad/sponsored content to our hate speech experts.
Experts evaluate whether the ad should be flagged as hateful speech. Each expert previously agreed to the Herald’s definition. The experts will be asked to decide whether the “hateful speech” flag is appropriate based on our definition and a detailed reporting/analysis that we provide.
If the majority of respondents agree the ad/sponsored content is misleading, the reporter will flag it as “hateful speech.” If only one expert responds, their judgment will be applied. Herald reporters will never apply this flag without an expert opinion.
Promotes Violence
The following definition was informed by the policies against promoting violence used by Facebook and Twitter, two of the United States’ largest social media platforms.
The Herald considers any statement that meets at least one of the following criteria to be promoting violence:
1) Includes an explicit or implicit call to violence.
2) Uses images of weapons in a way that could inspire violence.
3) Praises instances of violence by others.
4) Promotes disinformation about a person or a group of people that implies they are a threat to the safety of others and could make them the target of violent reprisals.
5) Promotes or encourages the activities of private militias and other unlawful armed groups.
Procedure for flagging content that promotes violence
If the material appears to be promoting violence, Herald reporters will send the ad/sponsored content to our violent speech experts.
Experts evaluate whether the ad should be flagged as promoting violence. Each expert previously agreed to the Herald’s definition.
If the majority of respondents agree the ad/sponsored content is misleading, the reporter will flag it as “promotes violence.” If only one expert responds, their judgment will be applied. Herald reporters will never apply this flag without an expert opinion.
Foreign Interference
In order to be flagged as “foreign interference,” the content must meet at least one of the following criteria:
Be posted on social media by a group definitively linked to the secret efforts of a foreign government. That link must be established either through an investigation by law enforcement, the U.S. Congress, a social media company or reputable researchers/journalists.
Be posted on social media by a news outlet classified as a “foreign mission” by the U.S. government PROVIDED that the content is an attempt to amplify other, more clandestine foreign interference efforts, such as RT’s amplification of WikiLeaks in 2016.
NOTE: Herald reporters adapted a definition of foreign interference used by the Atlantic Council’s Digital Forensic Research Lab, which states that “foreign interference is activity by a foreign actor that is ‘coercive, corrupting, deceptive, or clandestine’ in nature, distinguishing it from the more benign phenomenon of foreign influence.” Additionally, foreign interference does not always rely on disinformation. In fact, it often relies on legitimate news stories being shared by groups that are “sock puppets” of foreign governments.
Procedure for flagging foreign interference
If the material appears to constitute foreign interference, reporters will send the ad/sponsored content to our foreign interference experts.
Experts evaluate whether the ad should be flagged as foreign interference. Each expert previously agreed to the Herald’s definition. The experts will be asked to decide whether the “foreign interference” flag is appropriate based on our definition and a detailed reporting/analysis that we provide.
If the majority of respondents agree the ad/sponsored content constitutes a foreign interference attempt, the reporter will flag it as “foreign interference.” If only one expert responds, their judgment will be applied. Herald reporters will never apply this flag without an expert opinion.
Collects Personal Data
Ads/sponsored content or associated websites are flagged for collecting personal data if any of the following conditions apply:
The ad/sponsored content or website requests and stores the personal data of users such as a cell phone or email address, whether or not the collection of such data is stated in the privacy policy or terms of service.
Those responsible for the ad/sponsored content or website acknowledge in response to Herald questions that data collection is occurring.
Flagging Images
The Herald sought not to amplify disinformation, hateful speech and violence even as it brought such content to the attention of readers.
In cases where reporters identified ads/sponsored content where those flags, the images were published with a watermark clearly identifying the ad as either “disinformation,” “hateful speech” or “promotes violence.” The watermark makes it impossible for the content to be shared on social media but still allows readers to view it.
Ad labels
In addition to flags, Herald reporters will apply labels that offer at-a-glance information about the most important takeaways from the analysis.
Provides inaccurate information about the voting process
Applied when an ad about voting is flagged “false content,” “misleading content,” or “manipulated image.”
When an ad is labeled in this way, voters should not follow the instructions on the ad because it contains incorrect or misleading information about how, where, or when to vote AND/OR how, where, or when to get a ballot, AND/OR who can vote. The label is based on Herald reporting and reporting from other media outlets.
Provides inaccurate information about donations
Applied when an ad is flagged “false content” or “misleading content” based on what it claims about donations.
Ads labeled in this way contain incorrect or misleading information about who can donate, how a person can donate, AND/OR where the money will go. The label is based on Herald reporting and reporting from other media outlets.
Might contain inaccuracies
This label applies to pre-filled voter registration and mail-by-ballot form requests where at least one reader has discovered the forms contained incorrect personal identifying information.
Provides accurate information about donations
Applied when Herald reporters determine that an ad about donations is accurate.
Ads labeled in this way contain accurate information about who can donate, how a person can donate, AND/OR where the money will go. The label is based on Herald reporting and reporting from other media outlets.
Donation solicitation from a group that is yet to report spending money
Applied to ads when a group soliciting voters for donations has not yet reported spending money on political activities (i.e., excluding fundraising, salaries and other administrative expenses), according to the most recent filings from the Federal Election Commission.
Political Orientation Analysis
Herald reporters determined the political orientation of PACs spending money on ads and labeled them in the following ways:
Supports candidates and committees from both parties
Applied to any PAC where less than 60 percent of political spending goes to candidates and committees of a particular party in the current election cycle, based on a Herald analysis of OpenSecrets data.
If the PAC has not spent money in support of candidates, reporters examine its public messaging, mission statement and activities to determine its political orientation.
Primarily supports Democratic Party candidates and committees
Any PAC where more than 60 percent but less than 100 percent of political spending has gone to Democratic candidates and committees in the current election cycle, based on a Herald analysis of OpenSecrets data.
Primarily supports Republican Party candidates and committees
Any PAC where more than 60 percent but less than 100 percent of political spending has gone to Republican candidates and committees during the current election cycle, based on a Herald analysis of OpenSecrets data.
Exclusively supports Democratic Party candidates and committees
Any PAC where 100 percent of political spending has gone to Democratic Party candidates and committees during the current election cycle, based on a Herald analysis of OpenSecrets data.
If the PAC has not spent money in support of candidates, reporters examine its public messaging, mission statement and activities to determine its political orientation.
Exclusively supports Republican Party candidates and committees
Any PAC where 100 percent of political spending has gone to Republican Party candidates and committees during the current election cycle, based on a Herald analysis of OpenSecrets data.
If the PAC has not spent money in support of candidates, reporters examine its public messaging, mission statement and activities to determine its political orientation.
For questions or comments about our data or methodology, please contact sblaskey@miamiherald.com and nnehamas@miamiherald.com.
This project was edited by Miami Herald Investigations Editor Casey Frank with visual input from McClatchy Director of Design Eduardo Alvarez.
This story was originally published October 8, 2020 at 2:15 PM.