Six-Chart Sunday – Machine Politics (AI & elections)

6 Infographics + 1 Video (Candidates matter even in the age of AI)

What does AI mean for elections? Campaigns have always leveraged new innovations, from radio to TV to direct mail to the Internet, with traditionalists forever fretting the “end of democracy” as a result. Yet new technologies often just amplify & accelerate preexisting trends, for better and for worse. Will AI be different? Here are six potential impacts:

---

1. Super-Gerrymandering. Researchers were using high-performance computing and big data to discover new materials and analyze proteins before AI… but AI massively accelerated & improved such data-intensive efforts. Gerrymandering dates to 1812 in the U.S. Party strategists have been leveraging voter data and computing to draw advantageous maps for decades. AI promises gerrymandering on steroids — weapons of mass division in the 2025 redistricting wars — with powerful AI models able to (1) precisely-sift unprecedented amounts and unprecedentedly-personal data, (2) compare unlimited potential maps to optimize outcomes. But while AI-enabled cramming could reduce the paltry 20% of seats that are currently competitive, AI-drawn maps might also create more competitive seats by shifting voters out of safer seats (where risk-averse incumbent politicians often prefer them).

2. More Powerful Persuasion. In the business world, AI is making campaigns more targeted, efficient & effective. AI chat-bots are more persuasive than humans 64% of the time when provided with minimal demographic information. AI-powered campaigns show 131% increase in click-through rates and a 41% increase in overall engagement vs non-AI. Of course, campaign communications were already personalized to specific voters before AI models arrived. But AI-enabled messaging will leverage individual search histories in unprecedented ways, bringing cutting-edge brain science to the art of political persuasion.

3. Truth is in the AI of the Beholder, Further Marginalizing Mass Media. The era of “choose your own reality” and “fake political news” preceded AI models by at least 200 years. AI is not the reason trust in mass media has declined for most of the past half century, with voters increasingly turning to non-mainstream media sources to get their political information. But AI will surely make it harder to detect, and far easier to mass-produce, misinformation, disinformation & “pink slime” (partisan advocacy posing as legit media). In the age of AI, an increasingly-atomized electorate will choose among candidates based on increasingly-divergent perceived realities.

4. States Setting the Guardrails. American politics has long witnessed cycles of abuse & reform. Campaign excesses in the Gilded Age led to laws restricting corporate and union contributions and a Constitutional amendment providing direct election of Senators. Nixon Era abuses led to increased finance transparency and giving limits. Diffusion of radio, TV, cable and telephone platforms all led to campaign-specific regulations – e.g. the Fairness Doctrine, public interest standard, equal time rule and anti-fraud prohibitions. So far the federal government does not specifically-regulate use of AI in election campaigns, while 26 states that have passed laws to do just that.

5. Empowered Challengers. Our current electoral system is structurally skewed in favor of incumbents and against challengers. Why else do >90% of incumbents get reelected every year despite only 13% of voters approving Congress’ job performance? AI promises to level-up challengers against better-funded incumbent candidates, giving insurgent campaigns more firepower despite fewer resources (e.g. AI reduces costs of producing ad copy, writing speeches, developing policy planks, obtaining and analyzing data). Challengers are often likely to be more innovative and take more risks (e.g. McCain in 2000, Dean in 2004, Obama in 2008, Trump in 2016).

6. Government of the Model-Makers, for the Model-Makers & by the Model-Makers? American history is filled with examples of extraordinary individuals wielding significant political clout – think newspaper publishers such as William Randolph Hearst (“You furnish the pictures and I’ll furnish the war”) or mega-donors like Elon Musk (top spender in 2024 elections at $291M). Yet AI may have greater impact on elections than any single newspaper or super-PAC. If open-source models prevail, access to AI may prove transparent & democratized, but at the moment the top three large language models are proprietary and command 77% market share. Proprietary models are opaque and heavily-shaped by the preferences of the model makers, as seen by Google’s “woke Gemini” and Grok’s “MechaHitler” fiascos. Never have such a small number of individuals potentially-exercised such a large impact on what people see, hear, read, think and believe… with huge potential implications for how they vote.

CODA: This question is not new. In 1960 the Kennedy Campaign created a computer simulation of the 1960 election, in which they could test scenarios on an endlessly customizable virtual population. Newton Minnow, who became JFK’s FCC Chairman, declared “my own opinion is that such a thing (a) cannot work, (b) is immoral, (c) should be declared illegal… I shudder at the implication for public leadership of the notion . . . that a man shouldn’t say something until it is cleared with the machine.” That was 65 years ago.

VIDEO

While AI may transform elections, candidates will still matter: