InnovateUS 
AI and Elections: The Chatbot Campaign Trail Nobody's Regulating
Candidates are learning to shape chatbot responses by publishing AI-friendly content. One Missouri legislative candidate created a fact sheet specifically to improve how Google's AI summarized his campaign after discovering the major chatbots knew almost nothing about his platform. ("Politicians Are Trying to Change What Chatbots Say About Them," New York Times, July 19). Will the desire to appear above the fold on Google create incentives for candidates to put out better information?
At the same time, voters are increasingly turning to ChatGPT, Claude, and Gemini as political research assistants, to help them decide how to vote rather than conducting traditional web searches or reading voter guides. ("Who Should I Vote For? Voters Turn to AI Before Casting Their Ballots," New York Times, July 4, 2026.)
Congress is only beginning to grapple with the issue. A bipartisan pair of House members has asked DHS, DOJ, CISA, and the FEC to coordinate on the risks posed by AI chatbots to election integrity, underscoring that no clear oversight framework exists yet. ("Bipartisan lawmakers press agencies on AI election threats," The Hill, July 9).
State AI election laws are facing their first real test. Across the country, candidates are beginning to invoke new disclosure laws in disputes over AI-generated campaign content, including an Arizona lawsuit against a super PAC over allegedly unlabeled AI-generated images. ("State AI deepfake laws face first big test in 2026 midterm elections," Arizona Capitol Times, July 21).
While these stories suggest a new, largely unregulated layer of political communication, there are some overlooked solutions, like old-fashioned voting aid applications that millions of people use in other countries to get accurate information about candidates.
Which Brings Us to Information Literacy
We've spent years worrying about AI slop and misinformation: information abundance. Increasingly, though, we're confronting the opposite problem as well: information compression, in which thousands of sources are reduced to a single AI-generated answer that many people never click through.
Pew Research Center recently found that when an AI-generated summary appears above search results, click-through to the underlying sources falls by roughly half.
Reflecting on these changes for the New Jersey State Library's national conference on information literacy, I argue that information literacy must expand beyond teaching people how to evaluate sources. It now has to help individuals question AI-generated answers rather than simply accept them, help communities build knowledge together with AI, and help institutions use AI to listen at scale.
Read the full piece on Reboot: Information Literacy Is No Longer Just an Individual Skill
AI and Workers: Two Ways to Read the Same Layoffs
Amazon announced another 14,000 layoffs this week—its third major workforce reduction since October—even as it continues investing heavily in AI infrastructure. ("Amazon to lay off 14,000 corporate employees as spending on artificial intelligence accelerates," PBS NewsHour, July 21, 2026).
The easy headline is "AI replaces workers." But PwC's 2026 AI Jobs Barometer, based on more than one billion job postings across 27 countries, paints a more nuanced picture. The data suggest the labor market is increasingly rewarding occupations built around professional judgment while automating routine work. Jobs where expertise complements AI are growing faster and paying more than jobs where AI mainly substitutes for human labor. ("AI reshapes global labour market into two distinct paths, rewarding human skills," PwC, June 15, 2026).
Google's new AI & Economy ATLAS, drawn from 15 million Gemini interactions, finds AI touches 88% of US employment but handles just 21% of tasks per job, with under 10% attempting full automation — no evidence, Google's own researchers say, of "massive automation and displacement."
Thus, Amazon's cuts can be read as a management choice, not a technical inevitability. (One interesting note: government-service queries are nearly 20x over-represented in after-hours AI use, with people reaching for AI when the institutions are closed.) "Google’s AI & Economy ATLAS v1.0: Mapping Gemini Usage in the Economy," Google, July 23, 2026
Other evidence points in the same direction. Stanford's Digital Economy Lab, using ADP payroll data covering roughly one in six U.S. workers, finds that employment is declining among 22- to 25-year-olds in occupations most exposed to AI.
California has launched the first public dashboard to track AI-related displacement using unemployment insurance claims. And workers over 55 in AI-exposed occupations have been leaving the workforce at higher rates since ChatGPT's release, through a mix of layoffs and voluntary exits. (Fortune, June 27, 2026; California Policy Lab, June 2026; CNBC, July 13, 2026).
The policy conversation is beginning to catch up. Brookings' Xavier de Souza Briggs proposes a framework built around slowing harmful automation, steering adoption toward better outcomes, strengthening worker protections, and rethinking work itself. ("A framework of solutions for AI’s coming impacts on work and workers," Brookings, June 29)
Partnership on AI finds that organizations that involve workers build better AI systems. ("Seeking Worker Input in the Use of AI," Partnership on AI, June 29, 2026)
New America argues that reskilling alone is insufficient without giving workers meaningful influence over AI governance. ("Nothing About Us Without Us Can Actually Raise Us," New America, June 30, 2026).
Two counterpoints: Forbes argues that AI spending at companies such as Microsoft, Uber, and Nvidia is outpacing demonstrated productivity gains and that subsidized AI pricing may be masking the true economics. ("AI Costs More Than The People It Replaced," Forbes, July 2, 2026)
The Wall Street Journal, by contrast, points to demographic labor shortages and research by Daron Acemoglu suggesting that labor-saving technologies may become increasingly important as the workforce shrinks. ("Why AI Might Actually Help Solve the Next Labor Crisis," Wall Street Journal, July 13).
AI and Education: What Should Humans Learn When AI Does the Routine Work?
The evidence from the workforce points in one direction: AI increasingly rewards judgment while automating routine execution.
That raises an obvious educational question. If AI now performs much of the procedural mathematics students spend years mastering, what mathematical thinking should schools prioritize instead?
That question motivated my essay in Fast Company this week with Ted Dintersmith, author of Aftermath, when Cambridge, Massachusetts, required every eighth-grader to take Algebra I—a decision likely to result in more than 60 percent of students repeating the course.
The larger issue isn't Cambridge. It's that schools continue to organize mathematics around manual computation while giving comparatively little attention to statistics, probability, quantitative reasoning, and the judgment needed to interpret evidence, assess risk, or evaluate claims. Only about one-third of American adults have the numeracy skills needed to navigate common financial and health decisions, while more than 90 percent report some degree of math anxiety.
Our proposal is to use AI itself to engage thousands of students, parents, teachers, and employers in rethinking what quantitative reasoning should look like in an AI era—just as Charlotte-Mecklenburg Schools recently used AI-enabled engagement to hear from more than 10,000 families while developing its AI strategy.
Read the full piece (Fast Company, Jul 23, 2026 or on Reboot)