With majority control of the Senate at stake, it’s no surprise that the recent Georgia runoff elections were the most expensive political race in American history. Wherever campaign dollars flow freely, as they did in Georgia, accountability questions proliferate. Where did that money come from, where did it go, and what influence (if any) did those spending choices have on results?
Election law requires that campaign expenditures be made public. Even so, it often takes weeks or even months for analysts to sift through a dense forest of information and build usable datasets. Important decisions have to wait; only after a dataset has been given the green light can anyone even begin to search for associated outcomes and trends.
Many have wondered whether there is a legitimate place for artificial intelligence in harvesting public data repositories. Can machine-generated data really be trusted? And even if it can be, can it replace human analysts?
When Data is Public but Nearly Useless
Each TV and radio station is required to carbon-copy invoices for political advertisements to the Federal Communications Commission, which makes them public record. This sounds great on paper, but the FCC mandate is to merely publish the invoice documents — and each election comes with tens of thousands of invoices.
The invoices are important because of the dates, callsigns, and amounts listed on them page by page. The data allows analysts to build spending maps and scrutinize campaign behaviors, but the data needs to be aggregated in a spreadsheet first. A spreadsheet analysis would ordinarily mean going through this massive amount of invoices by hand.
A Better Use of Data
Of course, it’s not just the FCC receiving invoices — all businesses receive invoices when they trade with each other. Any company of size faces the same problem as the data analysts who would like to use the FCC data.
One automation software company that solves the problem of translating business documents to data for enterprises decided to unleash its automatic data capture on the FCC invoices for the Georgia runoff elections. The vendor Rossum, in partnership with analysts from the e.ventures fund, just published the resulting data set.
AI-Powered Reporting Follows the Money
Ever since the 1976 film “All the President’s Men,” many of us have internalized the admonition to “follow the money.” The practice remains a tried-and-true method for discovering unseen relationships and shedding light on patterns of activity and motivations that we might otherwise miss.
Data capture technology cannot read the minds of campaign managers — something for which we can all be thankful. But the new breed of automation based on artificial intelligence might at least enhance our ability to better see the tracks they’ve left.
Accelerating Data Collection
In addition to eliminating the time-consuming tedium of keystrokes and accelerating data collection, policy analyst Jordan Shapiro found that Rossum’s processing of the Georgia spending data produced a greater degree of granularity. This granularity, in turn, enabled her to better grasp the thinking that lay behind decisions made by the various campaigns.
Following the Spend
“As a political analyst, Rossum’s data about 2020 Georgia runoff election spending gave me the opportunity to get inside the heads of the campaigns to see which areas they thought were more or less competitive based on how much each candidate spent in that region,” says Shapiro. “Particularly helpful for my work was the ability to compare county-level spending patterns with a shift in vote share between November and January.”
Data Suggests a Shift from Red to Blue
Southern states have historically been fertile territory for Republicans seeking election, but that appears to be changing.
An NPR report from January 2020 found more Black Americans are moving south, and as they relocate, they are contributing to a shake-up in election results. In 2000, Rockdale County (southeast of Atlanta), for example, was a predominantly white area with a Black population of approximately 18%. Today, that same area has a 55% Black occupancy.
Changing Demographics
Shapiro notes that Georgia’s changing demography, heavy campaign spending by Democrats, increased voter mobilization at the grassroots level, and a tumultuous national political stage all played a hand in what many saw as upset victories. Political turmoil and civic unrest on the national scene rocked Republicans and set the stage for Democratic wins. Both races were reasonably close, with Ossoff winning his race by about 55,000 votes and Warnock winning his with around 93,000 votes.
Regardless, Rossum’s analysis found something even more important than spending-outcome correlations: reasons to default to AI-driven analysis.
Why AI-Enhanced Reporting Will Be Big
The mandatory availability of campaign spending records affords skeptics and naysayers an opportunity to fact-check any reporting that emerges after an election. Reports that have been compiled using AI-powered scanning techniques can be fact-checked just as easily as traditional reports constructed by workers furiously pounding away on keyboards.
As AI-powered processes continue to improve, confidence in this newer methodology is certain to grow as well. As that happens, voters can expect to see more accurate, useful information in the run-up to Election Day.
AI-driven reporting benefits democracy in at least four ways:
1. Enhanced Transparency of Public Data
Much data is the subject of public record, but too often, it is not readily available for analysis and therefore carries only a fraction of its potential. The problem with campaign spending records is that key dates and amounts are scattered through scanned documents instead of being aggregated in a spreadsheet that can be readily analyzed for new insights. This problem is common for many registries, FoIA data releases, and internal government operations.
2. Rapid Data Analysis
The speed advantage of AI-enabled reporting systems over traditional methods of computation will make relevant data available much earlier than in past election cycles. Earlier delivery of results could, in turn, open up new options for campaign managers to consider as constituents respond favorably or negatively to various messages.
3. Added Depth and Surfacing of Less-Obvious Correlations
As noted above, Shapiro was able to take Rossum reporting on election data and cross-reference it with migration data. In doing so, she discovered a trend in specific Georgia counties shifting from presumed Republican strongholds to surprise Democratic wins.
4. More Informed Policymaking and Decisions
Tightening the link between election results and policymaking can serve as a check on any politician’s temptation to drift away from the will of the people they serve. A democratically elected official can only disregard his or her mandate for so long. Faster access to accurate information allows voters more time to assess a politician’s voting record.
What’s Coming in 2022
With another election cycle coming entirely too soon, keep an eye out for new applications of AI. Both parties will use the technology to unearth patterns, analyze results, and suggest new political strategies. Policymakers will check proposals against their constituents’ latest voting trends.
Will 2022’s political environment be any less fraught than the 2020s? Maybe not — but it will be more data-driven.
Image credit: edgar colomba; pexels
