Pollution

I’ve been volunteering recently. The organization with which I’ve been working focuses on local and/or lower-profile races: state house, state senate, attorney general, and the like. Joe Biden raised $380 million last month; Jamie Harrison, Democratic candidate for U.S. Senate in South Carolina, broke a fundraising record with a $57 million haul last quarter. The presidential and U.S. Senate races have gained so much attention that the marginal dollar, or marginal hour volunteered, seem wasted. By contrast, in state house elections, where, typically, fewer than 100,000 voters participate, the concept of “making a difference” seems less farfetched.

Ours is a remote/geographically-distributed organization, and, even before the pandemic, we would have been unlikely to walk the streets, registering voters or knocking on doors. (Given the questionable effectiveness of door knocking campaigns, at least on a dollar-spent-per-vote-gained basis, our decision to spurn traditional canvassing would probably have been correct regardless.) So we have turned political volunteering into an activity not too different from work itself these days: something that can be done, at least by a sliver of the population, from the comfort of one’s apartment.

Lately, we’ve been “textbanking”. As the name implies, we collect a list of phone numbers for individuals we want to target (more on this later), and we set up a bank of devices (in this case computers) from which we can mass-text these individuals. Already you might recognize that we’re venturing into ethically gray territory. Don’t worry, it gets worse.

Textbanking has largely superseded “phonebanking” as a way to reach low-turnout voters, and particularly young voters, because these people (myself included) hardly ever pick up the phone, especially to respond to an unknown number. But, everyone reads their texts. As one article explains, “There’s no form of digital communication that’s more intimate—or as frictionless. Companies in the space report open rates between 70 and 90 percent for campaign texts, far higher than that for emails or phone calls.”

Textbanking has been enabled by the rise of “peer-to-peer” platforms built explicitly for the purpose of bombarding thousands of people with unsolicited messages. My organization used “BlueLink Messaging”, but there are several other companies who have built basically identical products. Bernie Sanders, arguably the pioneer of this technique, used Hustle (built by a former Facebook engineer who wanted to do something more meaningful with his life, of course). When Hustle decided to work exclusively with left-of-center groups in 2017, conservatives built rival companies like “RumbleUp” and “Opn Sesame”.

BlueLink’s website describes the company’s mission with the same kind of flowery mumbo-jumbo that brings to mind Mark Zuckerberg’s declaration that Facebook’s goal is to “bring the world closer together”:

We support Progressive campaigns and organizations by creating long-term tech infrastructure.

We believe that Infrastructure should be reliable, secure, and low cost. We emphasize sustainability rather than profit.

We are focused on building an open integration platform. We believe that non-technical users should be able to automatically flow their data without having to download, modify, and upload files to the next system. We’re agnostic to which tools they decide to use.

We are a group of technologists and organizers who come from a variety of backgrounds including: campaigns, Google, NASA, Health tech, Textio, and more. We care passionately about data accuracy and security.

Sounds nice, right? Here’s how it actually works. We upload a list of phone numbers and names that are associated with demographic and “voter roll” information. (How this information is obtained could be the subject of a separate essay, but suffice it to say that there are companies that make a lot of money doing this.) Demographic attributes might include race, gender, or age. “Voter roll” information comprises both your registered party as well as elections in which you have voted over the past few years. From that data, we can construct a profile of you: whether you are likely to be in a “high turnout” or “low turnout” group; whether you are likely to be sympathetic to our candidate, etc. These two dimensions—your ability to be “mobilized” and “persuaded”—ensure that the campaign does not waste its time targeting the wrong voters.

(Perhaps more worryingly, it also means that a campaign can design custom messages for particular groups of voters. In other words, I can moderate my message on behalf of a candidate when conversing with a Republican-leaning voter, or I can swing in the opposite direction for a disaffected, left-leaning Independent. The advent of digital targeting and rise of mass data collection has brought ever more sophistication to these efforts; the ability to talk out of both sides of your mouth is astonishing. It is what enabled Donald Trump, in 2016, to compile lists of voters to “deter” from the polls, and to target them with negative advertising. Black voters, in particular, were shown Facebook advertisements that portrayed Hillary Clinton in a negative light, especially on racial issues. White voters, on the other hand, were targeted with completely different messaging. The application of the techniques of advertising to politics is apparently what is meant by “support[ing] Progressive campaigns” and “emphasizing sustainability rather than profit”.)

Regulation has been slow to adapt. There is apparently a rule that text messaging cannot be fully automated: one cannot run a script that sends a message to a list of phone numbers, or use a bot to reply automatically. However, as long as there is a human pushing a button to send the text, the process transforms, almost magically, into something above-board. So I spent several hours last Sunday pushing that button hundreds of times, the names of my interlocutors flying by so quickly that I could not read them all.

I joked to a friend of mine that all textbanking requires is a working index finger and half a brain. Even the latter might be an overstatement. We were given effectively an algorithm to follow for making conversation. If the voter was undecided, we had a “canned response” ready to ask the voter about the issues they cared about. If the voter was committed to our candidate, another canned response prompted him or her to share their voting plan with us. And so on. We were instructed to send messages in batches of 100 before checking back for responses, the idea being that the response rate is so low (5-10%) that any time spent flipping back and forth from the message composition screen to the inbox is wasted.

Here are some of the responses I received:

“I’m not Scott STOP”

“Hi you need to update your records. This is not Robert’s phone and never has been in his name. Please do not contact me. Also why are you located in Utah if you’re for Texas? Not cool.”

“She has my vote!”

“Don’t text again”

“Porn” (probably my favorite, in response to, “is there an issue you’d be interested in hearing about”)

“Yes, we even know her personally.”

“I will be voting straight Republican. The democrats have shown who they are at their core, with the Russia hoax, impeachment, not moving forward with a new PPP bill to help the American people for no other reason than politics, and now trying to invoke the 25th Amendment three weeks before Election Day? I’ve always been a democrat, and I’m now ashamed to admit that. I will NEVER vote blue again in my lifetime.”

“We just did [vote]”

“Yes, I would be interested in a yard sign [too]”

“No”

“Fuck off you creep. I’ve said stop already!”

In the end, the opt-out rate (the people telling us “STOP”) was around 3-5%, while the response rate was perhaps double that. And this is considered to be a relatively successful text campaign.

It’s worth stopping at this point and asking: who wants this system? Judging from anecdotal evidence, voters, even sympathetic ones, seem to be overwhelmed with the volume of messages and disgusted by the creepy nature by which their numbers were obtained. Voters who slipped through the targeting filters (Republicans targeted by Democrats, leftists targeted by centrists, etc.) are doubly angry: both that they received an unsolicited message and that it is one that they do not wish to hear. The 5-10% of recipients who do respond are arguably making the problem worse by encouraging campaigns to persist. (Not that I’m blaming them; everyone wants someone sympathetic to talk to about important issues, especially in these atomized, socially distanced times.) And, as election day nears, the trickle of messages will turn into a torrent. Already, in the last week, I’ve received messages from Joe Biden, my U.S. Senate candidate, Planned Parenthood, the West Virginia U.S. Senate candidate (who probably got my number through the Bernie campaign), a candidate for my hometown school district’s school board (I used to be registered there), and two different organizations affiliated with my state’s Democratic party.

Each little organization is doing what it thinks is best. It is true that knocking on doors is better than nothing; that calling phones is better than nothing; that texting and volunteering and giving money and spending time are better than nothing. But the problem is that each of these actions makes sense in isolation but contributes collectively to a situation that makes people miserable. It is the epitome of the “tragedy of the commons”. As someone who has been on both sides of textbanking—the sender and the recipient—I can understand both the appeal of wanting to reach thousands of voters, and the desire to not want to be appealed to. But technology and the incentives embedded in the system give primacy to the former perspective above the latter. Free speech has become pollution, and the government has abdicated its responsibility to regulate it.

It is perhaps also worth thinking about the well-meaning people who volunteer or work for organizations that they see as doing good. I would love to help win an election, or flip a state house, or disseminate democratic socialism, or shift the Overton window. But why is it that this effort inevitably gets coopted; that the former “Google, NASA, Health tech” employees who want to effect political change and bolster progressivism end up building, well, this? Why is that I am polluting the very waters I am drinking from? I am afraid that this is one of my essays where I don’t have a pat, glib conclusion. To do something beneficial in the world, for the world, is a yearning that most of my liberal friends share, particularly the ones who have made it rich not doing those things since graduating. But the idea that this is as simple as accepting a large pay cut to work for a “progressive”, “mission-driven” organization seems highly dubious to me. This is not to say that we should continue collecting large checks for our current bullshit jobs, but instead that the distance between Facebook and BlueLink, or between my job and my volunteering work, is not as vast as it appears.

Advertisement