Is A.I. the Seer That Politicians Have Been Craving?
WHY YOU SHOULD CARE
Using artificial intelligence and machine learning, politicos find new ways to reach voters amid lockdowns.
- Using artificial intelligence and machine learning, political upstarts are winning, forcing Democrats and Republicans to turn to the technology at a time when physical campaigning is tougher than ever.
- Candidates are using these tools both to target potential voters and identify those on the fence.
When Julie Emerson was facing reelection in her Louisiana state House race in 2019, the 32-year-old wasn’t about to take any chances. She had won her first election by just 247 votes four years earlier. And as the first Republican to represent her Lafayette district since Reconstruction, she knew better than to take any votes for granted.
So Emerson let her campaign be a guinea pig when Numinar, a Washington-based startup catering to conservatives, promised to use artificial intelligence to help her win. Emerson and Will Long, the company’s young founder and whiz kid from Oklahoma, worked through the kinks in the AI-based platform. For Emerson, the result was a landslide 70 percent victory. For the rest of the GOP sphere, it was a powerful case study — of how AI could suddenly transform elections.
There is no way to do any of this without A.I.
Daniel Scarvalone, Bully Pulpit Interactive
That realization is spreading across the political spectrum, as Democrats and Republicans alike uncover how AI and machine learning can help them win elections. The right has companies like Numinar and AdVictory. The left has newer additions such as Call Time and OutFox A.I., which are using advanced tech to streamline fundraising and better scale targeted ads. Left-leaning organizations like PredictWise and Grow Progress, which are a few years older, are also using AI to uncover granular details about how messaging campaigns can best influence voters based on their foundational beliefs.
“If you can make informed predictions about which values matter most to people, you can gain an advantage in reaching out to them,” says Daniel Scarvalone, senior director of research and data for the Democratic digital consulting firm Bully Pulpit Interactive. “There is no way to do any of this without AI.”
Even neutral companies are joining the political realm, such as Resonate, which uses AI to create actionable audiences for both corporate and candidate media targeting. The Google Max Lift program, a machine learning mechanism, can be retooled to help candidates optimize ad targeting based on what the algorithm learns from surveying target voters. The result: smarter and, most importantly, more scalable campaigns, at a time when coronavirus lockdowns have made it harder to reach voters than ever before.
“Because it’s scalable, you can come up with voter scores all down and up the ballot, to places where nobody is going to do a custom-scoring voter project,” Long says.
AI has been used in elections in the past, most infamously by Cambridge Analytica, which used it for targeted advertising programs to help the Trump 2016 and Brexit campaigns. The key difference, though, experts say, is that these efforts aren’t breaking election laws or invading voters’ privacy. The main value that new artificial intelligence programs add is highly customizable voter files that can be pinpointed to the type of local and statewide elections that usually don’t interest larger, national political strategy firms.
Consider Emerson’s Statehouse race. Numinar was able to project a 57-43 victory for the incumbent, based on scoring the projected partisan leanings of 30,000 voters — matching her pollster’s results in the process. Then, the AI model went a step further by highlighting 1,739 undecided voters who could be swayed. That gave Emerson a list of new targets, whether through phone calls, texts, emails or door knocking, leading in part to her blowout 70-30 win.
Such technology can also be used to tweak turnout models, as it did in a Virginia state Senate race for Geary Higgins, whose team originally projected around 33 percent turnout based on historical trends. Numinar predicted closer to 45 percent and made suggestions based on that shift, ultimately ending up just 2 percentage points off. When construction company CEO Bill Lee was running for governor in Tennessee as an outsider and political newcomer, Google’s Max Lift helped the candidate go from less than 30 percent name recognition to winning the crowded GOP primary.
As both Republicans and Democrats invest in digital campaign tech, from the Startup Caucus on the right to Higher Ground Labs on the left, such machine-learning-focused companies are gaining new prominence. While these forms of AI are relatively innocuous, there are other, nefarious uses of the technology — in India, for example, deep fakes notoriously used in porn were deployed to sway state elections earlier this year.
Navigating this space is also not easy for politicos who are for the most part notoriously slow to adopt forward-thinking tech. And some warn that, without a human eye, all the data reading in the world can’t be smartly applied. “How do you build the guardrails for the tools that constrain them with political intuition?” Scarvalone says. “You have to work on enough campaigns to understand the decisions you are trying to inform.”
Admittedly, campaigns can already learn most of those things, but they’ll have to pay big bucks to do it by hiring large national marketing agencies like Grassroots Targeting or WPA Intelligence (for such work, prices can start in the five-digit range and quickly grow into six figures). While it may lack the personal touch that human strategists can provide, most candidates running bootstrap races for everything from dogcatcher to state senator will likely have an easier time affording AI.
And that may prove more crucial in 2020 than ever, considering how the stock market crash and job losses have dampened donations. Especially now, why have a human do what a machine can do?