Noah Smith challenges a pervasive narrative of economic despair by suggesting that America's gloomy mood may be less about reality and more about broken survey methodology. He also offers a surprising defense of artificial intelligence as a potential depolarizing force, while simultaneously critiquing the mathematical assumptions economists use to weigh existential risk.
The Data Behind the "Vibecession"
Smith opens by addressing the dissonance between strong macroeconomic indicators and the public's pessimistic outlook. He notes that the University of Michigan's Index of Consumer Sentiment has plummeted to record lows, lower even than during the Great Recession or the 1970s inflation crisis. However, he pivots quickly to a guest analysis by Joel Wertheimer, who argues this data is fundamentally flawed.
"The switch to online polling made responses more negative and, There are too many Democrats in the sample," Smith writes, summarizing Wertheimer's critique of the survey's methodology. This shift from telephone to internet-based interviewing has introduced a partisan skew that distorts historical comparisons. The author points out that when the data is adjusted to match Pew Research Center demographics, the sentiment index rises significantly, aligning it with more moderate measures like those from the Conference Board.
This reframing is crucial because it suggests the "vibecession" might be an artifact of how we measure opinion rather than a reflection of economic reality. As Smith puts it, "The notion that American consumers are ultra-bearish might be perched upon a flimsy bicycle of bad data." The comparison to historical survey shifts is apt; just as the Roper Center for Public Opinion Research has documented since the mid-20th century, changes in response rates and collection modes can drastically alter outcomes without changing public sentiment. Critics might note that even if the University of Michigan's index is skewed, other measures also show a decline from 2019 peaks, suggesting some dissatisfaction is real, though perhaps not as catastrophic as the headline numbers imply.
AI as "Digital Cronkite"
Moving to technology and politics, Smith revisits his hypothesis that artificial intelligence could serve as a moderating influence in polarized discourse. He contrasts the fear of "AI sycophancy"—where machines simply tell users what they want to hear—with new research suggesting the opposite effect.
Smith cites Conlon and Schwardmann's 2026 study, which found that despite being agreeable, AI advice actually moves people away from their initial extreme leanings. "We find that AI advice depolarizes choices on average," Smith quotes from the abstract. The mechanism appears to be a combination of flattering language with substantive information that corrects misconceptions.
This is a compelling argument for those worried about echo chambers, yet it relies heavily on the assumption that AI models are trained on a "Moderate Normie" average of society. While Smith acknowledges this might be the most obvious reason for depolarization, he notes the researchers focused more on the "substantive information" theory. The historical parallel to Walter Cronkite is effective here; just as broadcast television once provided a shared reality, AI could theoretically inject similar grounding into modern debates. However, the argument assumes that users will accept the substantive corrections offered by the AI, which may not always be the case in highly charged political environments.
"The notion that American consumers are ultra-bearish might be perched upon a flimsy bicycle of bad data."
The Mathematics of Existential Risk
Smith then turns to a sharper critique of economic modeling regarding artificial intelligence risks, specifically targeting the work of Chad Jones. He examines a calculation where Jones suggests it is rational to accept a 1 in 3 chance of human extinction for a massive increase in living standards.
The core of Smith's objection lies in the utility function used in the model. "If you take log utility seriously, then death is infinitely bad," he writes. He argues that setting the utility of extinction to zero is a "highly dubious assumption" because it equates human annihilation with a baseline existence at zero consumption.
This critique highlights a significant gap between abstract economic models and ethical reality. By setting the value of extinction to zero, the model ignores the infinite negative value many would assign to the end of humanity. Smith notes that "humanity going extinct is no worse than humanity existing for all eternity at some baseline level of consumption" under these specific mathematical constraints. This is a vital distinction for policymakers and researchers at institutions like Anthropic; as AI capabilities grow, the models used to weigh their risks must accurately reflect the stakes. A counterargument worth considering is that in high-stakes decision theory, assigning infinite values can sometimes paralyze action, forcing economists to use finite proxies. Yet, Smith's point stands: if the proxy is too low, it dangerously underestimates the cost of failure.
Generational Wealth and Inequality
Finally, Smith dismantles the popular narrative that Millennials have been economically "screwed" compared to previous generations. Citing Corinth and Larrimore (2026), he presents data showing that after taxes and transfers, Millennials are better off at equivalent ages than Boomers were.
"There are a lot fewer Millennials making less than $30,000... than any other generation did," Smith writes, highlighting the shift in income distributions. The data reveals that while inequality within the Millennial cohort has increased—with more people at both the very top and the bottom—the overall trend is upward mobility.
This analysis provides a necessary corrective to generational resentment, showing that government redistribution has played a significant role in lifting the floor for younger workers. However, Smith is careful not to dismiss the reality of inequality; he notes that the Millennial generation is more "spread-out," meaning gains are unevenly distributed. The narrative that one generation was systematically robbed by another doesn't hold up under scrutiny, but the story of widening gaps within a generation remains a critical policy challenge.
Bottom Line
Smith's strongest contribution here is his rigorous defense of data integrity, showing how methodological shifts can manufacture economic despair where none exists. His biggest vulnerability lies in the optimism surrounding AI depolarization, which may underestimate human resistance to correction. Readers should watch whether future survey adjustments continue to narrow the gap between sentiment and hard economic reality.