I Thought America Had Lost Its Empathy. So I Checked My Own Claim.

This is the claim with the famous statistic behind it, and the statistic is real. In 2011, University of Michigan researcher Sara Konrath and two colleagues combined 72 studies of 13,737 American college students who took the same empathy questionnaire between 1979 and 2009. The two core measures, empathic concern (feeling for someone in trouble) and perspective taking (imagining how things look to them), both fell. Most of the drop came after 2000.

If the story stopped there, I'd have been right. It doesn't stop there. Konrath went back and updated her own work, adding 126 newer samples and six unpublished datasets from her lab, for a total of 38,693 undergraduates. The update found empathy among young Americans has been rising since 2008, nearly back to the levels of the late 1970s.

Behavior tells a similar story. Gallup asks people in more than 140 countries whether, in the past month, they helped a stranger who needed it. Across the decade from 2009 to 2018, 72 percent of Americans said yes, and the U.S. ranked first in the world overall on the Charities Aid Foundation's giving index for that decade. It has slipped since — as low as 19th in 2021 — but it was back in the top ten in 2024.

I didn't write this claim, but it's the one the evidence kept pointing to. Americans are still kind to a stranger in a parking lot. They have become far less kind in how they think about the tens of millions of fellow citizens in the other party.

In 2022, Pew Research Center surveyed 6,174 American adults and asked whether people in the opposing party are more immoral than other Americans. Here is how the answers moved in six years:

The same survey found 72 percent of Republicans and 64 percent of Democrats calling the other side more dishonest than other Americans. Meanwhile each side rated itself more moral than before. In 1994, fewer than a quarter of either party viewed the other very unfavorably. By 2022 it was 62 percent of Republicans and 54 percent of Democrats.

That symmetry is the part working people should sit with. The folks on the other side of your family table are, by and large, paying the same rent, the same insurance premiums, and the same grocery bills you are. When each half of the working class sees the other half as immoral, that is a division someone else gets to cash in on.

The best measure of adult skills is a test called PIAAC, run by the OECD and, in the U.S., by the Department of Education's National Center for Education Statistics. It gives real reading, math and problem-solving tasks to adults ages 16 to 65.

The 2023 results were bad. Between 2017 and 2023, U.S. adults' average literacy score fell 12 points, from 271 to 258, and their numeracy score fell from 255 to 249. The gap between the strongest and weakest performers grew: people at the top held steady while everyone else slid. The U.S. landed near the middle of participating countries in literacy and below average in math.

Two findings keep this from being a simple “they dumbed us down” story. First, this isn't only an American problem: among the countries tested in both rounds of the OECD's adult skills survey, average literacy improved significantly in only two, Finland and Denmark, and held steady or fell everywhere else. Second, if American schools were the main cause, you'd expect the youngest adults to show the biggest drop. They didn't. The age group NCES reported with a drop in numeracy was adults 55 to 65, people who finished school decades ago.

This is the part of my sentence with the strongest evidence behind it. In 2021, a team of political scientists from Utah, Washington University, Princeton, Dartmouth and Exeter surveyed two nationally representative samples totaling 8,285 Americans. They showed people real and fake news headlines, then asked them to rate their own ability to tell the difference.

About three in four people overestimated their ability. On average, people placed themselves 22 percentiles higher than their actual performance. Roughly 90 percent said they were above average, which is arithmetically impossible.

The worst finding: the people who were worst at spotting fake news were the least aware of it. Using real browsing data, the researchers found overconfident people visited untrustworthy websites more often and said they were more willing to share false stories, especially ones that flattered their own politics.

Here is where my own sentence fails my own standard. This site does not attribute intent without a source, and I don't have one. None of the research above identifies a person, company or party that set out to erase empathy or lower test scores. Konrath's own work offers possible explanations, not culprits: the 2011 study floated several theories, from media to parenting to rising pressure to succeed, and the 2023 update points to changes in how people relate to one another.

What I can document is something narrower and, I think, more useful: the incentives.

In 2021, researchers at Cambridge and NYU analyzed 2.7 million Facebook and Twitter posts from news outlets and members of Congress. Posts about the political other side were shared about twice as often as posts about one's own side. Each additional word referring to the opposing party raised the odds of a post being shared by 67 percent. Contempt for the other side beat every other predictor they tested, including emotional language in general.

The same paper cites Wall Street Journal reporting (Jeff Horwitz and Deepa Seetharaman, May 26, 2020) that a Facebook research team warned the company in 2018 that its algorithms exploited people's attraction to divisiveness, and that executives declined to adopt the team's proposed fixes. If that reporting holds, it shows a company that knew and kept the business model. That is a serious finding. It still isn't a plan to erase American empathy.