Jill Lepore’s artificial state starts with a blunt claim: on major social platforms, there are now “far more bots than humans.”
That is the warning behind Lepore’s new book, The Rise and Fall of the Artificial State, which comes out on August 25, and her interview with Decoder, according to The Verge. Lepore, a Harvard professor and New Yorker staff writer, is not arguing that every use of AI is evil. Her argument is sharper: when institutions replace human judgment with measurement, sorting, prediction, and automation, democracy starts to behave less like self-government and more like a machine for extracting signals.
That is why the backlash matters. Students booing AI CEOs at graduation ceremonies, residents fighting data centers, and faculty asking for real debate over campus AI policy are not just vibes. In Lepore’s telling, they are signs that people still expect to be asked before machines and corporations redesign civic life.
“What we’re at risk of is losing the liberal constitutional democracy and having it become supplanted by what I describe as the artificial state.”
Why students booing AI leaders fits Jill Lepore’s artificial state argument
The visible trigger is cultural. College students are booing AI executives at graduation ceremonies. Communities are showing up to oppose data centers. Professors and students are asking why AI tools are being rolled into institutions before the affected people have had a serious say.
Lepore reads that resistance as a democratic reflex, not anti-tech panic. The target is not software in the abstract. It is the pattern of imposing systems that measure, sort, and replace human activity while calling the result inevitable.
That distinction matters right now because AI is moving beyond chat windows and productivity tools. In the Verge interview, Lepore and Nilay Patel discuss its role in elections, education, public discourse, platform moderation, data infrastructure, and the daily routines through which people experience power.
Silicon Valley’s answer often sounds like: build first, let democracy set guardrails later. Lepore’s answer is darker. If the systems being built weaken deliberation, privatize public debate, and turn citizens into data points, democracy may not be strong enough to clean up after them.
XOOMAR analysis: this is the useful frame for readers. The fight is not “AI versus no AI.” It is consent versus rollout, deliberation versus branding, and politics versus optimization.
What Jill Lepore means by the artificial state replacing liberal democracy
Lepore defines the artificial state as a successor to the liberal nation-state, where core functions of public life migrate to private multinational corporations.
Her clearest example is the public square. People learn about elections, news, rights, duties, and public controversies on corporate platforms. Those platforms are not designed primarily for democratic deliberation. Lepore says their priority is “the monetization of attention and of outrage.”
That does not mean the formal state disappears. Courts, legislatures, agencies, schools, and local governments still exist. But governance changes when platforms, algorithms, bots, and corporate policies decide what gets amplified, suppressed, monetized, automated, or turned into data.
This is the difference between government and governance. Government is the visible machinery of the state. Governance is how power actually reaches people.
A private platform’s moderation rules can shape public debate. A recommendation system can steer political attention. An AI tool can draft a campaign message, select a recipient, and measure the response. Lepore’s point is that when these systems become the operating layer of civic life, democracy keeps the shell while losing muscle.
For related XOOMAR coverage on how AI power is already colliding with platform control and legal accountability, see Apple Lawsuit Threatens OpenAI Hardware at Worst Time and Nudify Apps Drag Apple and Google Into Legal Crosshairs.
How polling, data mining, and microtargeting made politics feel less human
Lepore does not start the story with ChatGPT. She starts with measurement.
The arc runs from census-taking to scientific polling to computer-aided targeting. A key turn comes in the 1930s, when George Gallup starts the American Institute of Public Opinion research in 1935. Polling made it possible to understand voters without talking to all of them directly.
That sounds efficient. It also removes friction that democracy needs.
Lepore contrasts polling and data mining with older campaign work: volunteers attending PTA meetings, baseball games, basketball games, and library events. Those conversations were not just inefficient inputs. They were “the civic glue that holds a civil society together.”
A useful comparison:
| Political method | What it captures | What it loses |
|---|---|---|
| Door-to-door organizing | Conversation, trust, local context | Speed and scale |
| Polling and analytics | Preferences, probabilities, segments | Human exchange |
| AI microtargeting | Message fit, timing, susceptibility | Shared public debate |
The 1980s matter in Lepore’s history because personal computing and Silicon Valley culture made targeting more scalable. Ads and political campaigns could become cheaper, faster, and more precise.
The trade-off is brutal. A volunteer talking with parents at a basketball game has to listen. A model predicting which message might move one voter does not.
How bots and AI campaign tools push elections toward AI versus AI
Lepore’s most immediate election warning is simple: voters may ask chatbots how to vote while campaigns use AI to craft and deliver the messages those voters receive.
In the interview, she describes coverage of the 2026 election in the US and elsewhere, where people report asking ChatGPT, “How should I vote in the primary?” At the same time, campaign messages may be crafted by artificial intelligence, with delivery decisions also made by artificial intelligence.
Her phrase is stark: “AI versus AI.”
This is bigger than misinformation. The concern is not only that a false image or fake claim circulates. It is that public opinion becomes something to be measured, modeled, nudged, and harvested at scale.
Targeted advertising is the bridge. The same infrastructure that can sell a foam mattress can sort people by identity, anxiety, resentment, or susceptibility to a candidate’s pitch. The source discussion also points to legal and ethical problems around ad systems that sort people into sensitive categories tied to race, gender, housing, and other protected areas.
Lepore’s deeper claim is that this changes the citizen’s role. The person is no longer mainly a participant in a shared public argument. The person becomes an addressable unit.
That is politics flattened into a dashboard.
The Yale AI branding anecdote shows why institutions are losing trust
The most concrete mini case in the interview comes from Yale.
Lepore describes giving lectures there that became the basis for the book. She says a colleague told her that the first meeting of an AI task force, which faculty expected to address pedagogy and proper use, instead focused on how to brand AI at Yale.
The line that lands is not technical. It is institutional.
Students and faculty, in Lepore’s telling, wanted a real discussion. What should AI be used for? Where should it be off limits? Could the university pause for two weeks and hold campus-wide discussions involving staff, faculty, and students?
The students’ reaction, she says, was: “Let’s do that. Why can’t we do that?”
Instead, Lepore says Yale branded its relationship with OpenAI and made it a corporate agreement. The point is not just Yale. It is the pattern: institutions adopt AI tools, form partnerships, and present the result as progress while the people inside the institution ask why consultation came after commitment.
That connects directly to local data center fights raised in the interview. Residents attend meetings, object to projects, and still see officials approve them in the name of jobs or growth. The democratic ritual happens. The decision feels pre-made.
Can AI backlash force guardrails before the artificial state hardens?
Lepore is skeptical when tech leaders say democracy will restrain them.
Her objection is not that democratic process is useless. It is that Silicon Valley often uses “democracy” to mean feedback collection, voting in an app, or turning people into data points. That is not deliberative democracy.
Real democratic process is slower and messier. It needs debate, rules, representation, accountability, and people in a room who can challenge one another.
That is also where Lepore finds hope. Not in a better branding campaign. In students booing. In public forums. In local fights. In faculty refusing to treat AI adoption as inevitable. In officials facing political consequences when they ignore communities.
The practical implication is clear. Watch less for AI promises and more for consent mechanisms. Who gets consulted before deployment? Can communities say no? Do institutions publish rules before signing deals? Are citizens treated as participants or as inputs?
AI backlash is vital because it proves people still expect public life to be governed by humans, not quietly redesigned by machines, platforms, and corporate partnerships. The next test is whether that backlash becomes process, policy, and power before Lepore’s artificial state hardens into default.
Impact Analysis
- Lepore frames AI backlash as a democratic response to institutions adopting automation without public consent.
- The debate is shifting from whether AI tools are useful to who controls their role in civic life.
- Resistance from students, residents, and faculty signals broader concern about replacing human judgment with automated systems.
Originally published on XOOMAR. For more news and analysis, visit XOOMAR.
Top comments (0)