The Political Bias In Algorithm Sorting (BIAS) Emails Act of 2022 (S. 4409) was introduced by Senator John Thune (R-SD) and has attracted over 20 Senate Republican cosponsors, including Senate Minority Leader Mitch McConnell (R-KY). A House companion bill has been introduced, H.R. 8160, by Representative Debbie Lesko (R-AZ) and also has over 20 Republican cosponsors, including House Leader Kevin McCarthy (R-CA) and members of House Republican leadership.
This legislation is predicated on a study conducted by the North Carolina State University in which they analyzed email services use of spam filtering algorithms (SFAs) and the interactions between SFAs and political candidates on the right and left. Based on this study, Republicans leading S. 4409 and H.R. 8160 believe Google and other email services are disproportionately and unfairly categorizing emails from conservative candidates as spam. However, the authors of the study have stated that their study is being misinterpreted.
Digital communication and fundraising is increasingly being used by political campaigns. Digital fundraising tools allow potential supporters to give money to their preferred candidate or committee from anywhere with relative ease. Email campaigns also allow candidates to capitalize on momentum, quickly communicate with supporters, and utilize small dollar donors to bolster campaigns.
Large fundraising platforms have become very important for political campaigns. For example, after the death of Justice Ruth Bader Ginsburg, ActBlue, the dominant fundraising platform for Democrats, broke a record and processed over $100,000 per minute. Meanwhile, in 2021, WinRed, the preferred conservative fundraising platform, processed over $559 million in 2021 for more than 3,200 campaigns and organizations.
If email services were filtering emails to spam solely based on the political affiliations, this would understandably be concerning for lawmakers, political candidates, and voters. Proponents of S. 4409 and H.R. 8160 claim passage of the bill is critical for the integrity of elections and ensuring voters receive information from candidates. Unfortunately, there are several concerning provisions in the legislation that could undermine consumer privacy and create burdensome and unnecessary requirements.
The North Carolina State University Study:
The study from North Carolina State University, “A Peek into the Political Biases in Email Spam Filtering Algorithms During US Election 2020,” sparked inspiration for H.R. 8160 and S. 4409. The study focuses on four questions:
- Q1: Do SFAs of email services exhibit aggregate political biases? How do these biases compare across email services?
- Q2: Do SFAs treat similar emails from senders with different political affiliations in the same way?
- Q3: Do the interactions of the users with their email accounts, such as reading emails, impact the political biases of SFAs?
- Q4: Do SFAs exhibit different political biases for recipients belonging to different demographic?
To address these questions, the study created 102 email accounts and subscribed to two presidential, 78 Senate, and 156 House candidates. The email services were Gmail, Outlook, and Yahoo. In their conclusion, the study states that SFAs exhibited political biases, with Gmail leaning towards the political left and Yahoo and Outlook leaning towards the political right.
Some Republicans point to this study as a smoking gun that email services are politically biased and could impact elections. This argument has been subject to some criticism from the authors of the study. Muhammad Shazad, an author of the study, told the Washington Post that their study did not come to the conclusion that Big Tech was tilting the scales towards left-wing candidates.
Instead, Shazad states that there is bias under certain circumstances, and it does not demonstrate that someone is deliberately trying to turn elections. He also notes that spam filters demonstrated political biases in “default behavior” with new accounts, but when user preferences were simulated, the biases in Gmail almost disappeared, but in Outlook and Yahoo it did not.
Email providers dispute the claims that their spam filters are politically biased. Google has stated that the study has “flaws” and has too small of a sample size. In response to these concerns, Google has asked the Federal Election Commission (FEC) to approve a pilot program. According to reports, this pilot program would allow authorized candidate committees, political party committees, and leadership political action committees registered with the FEC to be exempt from spam detection as long as they do not violate Gmail’s policies.
What Does The Political BIAS Emails Act Do?
The Political BIAS Emails attempts to address the concerns Republicans have shared in response to the study by prohibiting algorithms from labeling political emails as spam without a user taking action and instituting transparency requirements. This legislation would apply to all email services except those with fewer than 500 employees in the most recent six-month period and that averaged less than $5 billion in annual gross receipts in a 12-month period.
Specifically, this legislation would make it unlawful for an email service to use a filtering algorithm to apply a label to an email from a political campaign unless the owner or user of the account took action to apply a label. Essentially, this would prevent email services from marking political emails as spam without intentional action by the user. This provision would take place three months after the legislation was enacted.
Additionally, email service providers covered by this bill would be required to submit quarterly transparency reports which would be made publicly available. These reports would include the number of instances political campaign emails were flagged as spam, the number of instances political emails were flagged as spam by an algorithm without user direction, and the number and percentage of emails flagged as spam broken down by party affiliation (Republican and Democrat) during the previous quarter.
A political campaign would also be allowed to request a report from email providers once per week during election years and twice per month during non-election years. Email providers would be required to provide a political campaign with the number of instances the campaign’s emails were flagged as spam, the percentage of emails flagged as spam, the number of emails that reached the intended recipient email inbox, the percentage of emails that reached the intended inbox, and a descriptive summary of why the political campaign requesting the information did not reach the intended recipient primary inbox.
Finally, email providers would be required, upon request from a political campaign, to provide best practices on steps campaigns should take to increase the number of emails that reach a recipient’s primary inbox.
While requiring reports and calling for transparency would generally be considered to be a “light-touch approach”, there are several concerning issues with this legislation. It would negatively impact the privacy of users of popular email providers and create onerous requirements for both email providers and their users.
Privacy concerns: This legislation would require that email providers search through inboxes and spam folders to produce transparency reports. This would be a significant privacy concern. If Republicans are worried about an anti-conservative bias from tech companies, requiring that they search through your inbox to see if you’re receiving emails from the Republican National Committee or Republican candidates should be an area of concern. It’s unlikely users of email services would be pleased to know that Google, Yahoo, or other email providers are mandated to routinely comb through inboxes searching for political emails they recieve.
Burdensome reporting requirements: This legislation would allow political campaigns to request information such as the number of emails that were marked as spam, the number of instances an email reached a user’s inbox, a descriptive summary of the reasons an email was marked as spam, and other information. A political campaign could make this request once per week during election years, twice per month during non-election years, and once per week in the 12 months preceding a special election. These reports would have to be provided no later than four days after the email provider receives the request.
In a given election year, Americans could vote on 33 Senate seats, 435 House seats, and the White House. If only two or three candidates ran for each of these positions, which is likely on the very low side, that would mean over 1,000 requests could be made weekly for email providers to produce reports within four days. That would also not include reports that could be requested by national committees, state committees, joint fundraising committees, and other committees covered in this legislation. These would be an extremely onerous requirement.
Roadmap for bad actors: This legislation would also require that email providers provide email campaigns with best practices to avoid being flagged as spam. First, it’s unclear why this provision is necessary. Email marketing services, like Mailchimp, provide guidance on what best practices are available to help avoid spam filters, like removing excess code or avoiding link shorteners.
While spam filters are sometimes negatively referred to as a “black box,” this can be a positive feature. Over half of all global email traffic is spam, and malicious emails like phishing or malware can pose real threats to users. If a roadmap on how to avoid spam filters is made widely available, bad actors would also have a much easier time bypassing spam filtering algorithms. Spam filtering algorithms are complex and, unfortunately, not always accurate on marking emails as spam. However, by providing a guide on avoiding spam filters to political campaigns, this legislation could also risk giving the same guidelines to bad actors and create cybersecurity and data security issues.
User-friendly service: This legislation would make it unlawful for a filtering algorithm to apply a label to an email (such as mark it as spam) unless the owner or user of the account took action to apply the label. Do consumers want this? It can be common for a user to sign up for an email list, or simply supply a company or organization with their email address, and then receive numerous emails as a result. Rather than mark each of these emails as spam, many users may simply delete the message. The same applies to political emails.
A user may have filled out a survey online, unaware that they were adding their name to an email distribution list. While some may unsubscribe or mark as spam, others may simply hit delete or never open the messages. Algorithms recognizing this behavior and filtering messages that are not opened or routinely deleted without being opened as spam is generally a user-friendly feature. While spam filtering algorithms are not perfect, having users go through and mark each political email as spam may be an unwanted nuisance for many.
Filtering certain emails to spam solely based on political affiliation would be concerning, but the evidence to suggest this is happening is, at best, mixed. Unfortunately, there are several concerning aspects with the supposed cure as well. As lawmakers look to address speech concerns, they should be careful not to undermine consumers’ privacy.