NBER working paper puts AI-generated text at 29.4% of early-2026 US STEM dissertation abstracts
NBER researchers using the Pangram detector flagged AI-generated text in 29.4% of US STEM dissertation abstracts filed from January to May 2026. The unreviewed paper also finds that graduates with flagged abstracts who did their undergraduate degrees in the US were 17% less likely to work in academia.
Reality
- Evidence50
- Adoption30
- Hype gap+15
- Incentives
- Not scored
- Confidence50
David Autor's trial of 133 patent lawyers found AI improved everyone's drafts, but only seniors kept a skill gain once it was gone. Firms that expected the tool to train junior associates still have to pay for that training, on the evidence of a Google-run working paper.
Reality
- Evidence55
- Adoption
- Not scored
- Hype gap+30
- Incentives70
- Confidence55
buildOne report1 publisher AI built stronger judgment only in senior patent attorneys over a 90-day trial
David Autor and Tanya Rodchenko's trial of 133 patent attorneys found AI lifted drafting quality for everyone but built stronger judgment only in seniors. Juniors' judgment scores split toward both ends with no average gain, so better drafts no longer show a firm which juniors are learning.
Reality
- Evidence62
- Adoption
- Not scored
- Hype gap+8
- Incentives62
- Confidence55
Chicago Booth economists estimate U.S. tariffs moved earnings from services into tradable industries
Rodrigo Adao and three co-authors used 2017-19 data from all 50 states and D.C. to infer whose income U.S. tariff policy treated as most valuable. The answer sorts by industry far more than by geography.
Reality
- Evidence48
- Adoption
- Not scored
- Hype gap+12
- Incentives35
- Confidence52
Measuring GLP-1 users against their own earlier selves erases most of the employment and marriage gap
In a new NBER working paper, Robert Kaestner and Cuiping Schiman followed adults with diabetes through 12 years of federal survey data and found little measurable change in depression, distress, employment or marriage.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+10
- Incentives40
- Confidence50
Before hiring, sort the role's ten tasks into four buckets to decide whether to hire, automate or redesign
An Entrepreneur contributor column asks managers to list the ten recurring tasks behind a requisition and sort them into judgment, relationship, repetition and coordination work. Its supporting studies measured support agents and consultants.
Reality
- Evidence33
- Adoption28
- Hype gap+22
- Incentives38
- Confidence55
Top-1% industry AI researchers out-earn matched academics by $1.5 million a year
The NBER analysis matches specialisms before it compares pay, which is the right control, but the headline gap describes the tail of the industry distribution, and the academics Nature asked all stayed.
Reality
- Evidence45
- Adoption
- Not scored
- Hype gap+30
- Incentives40
- Confidence50
Intent to stay in academia fell from 66% to 44% among biomedical trainees surveyed in March 2025
An NBER working paper asked 916 doctoral students and postdocs how they felt now and how they felt half a year earlier. The gap is large, and it measures stated intent, which is not the same as measuring departures.
Reality
- Evidence56
- Adoption
- Not scored
- Hype gap+20
- Incentives68
- Confidence50
Tennessee middle schoolers gained three points with an AI tutor designed to withhold answers
The winning condition both withheld solutions and made students redo the skill three times running, so what a district can copy from this experiment is a specification for how a tutor should behave rather than a shopping list.
Reality
- Evidence55
- Adoption20
- Hype gap+12
- Incentives65
- Confidence45
SNAP soda bans cut purchases 12%. The exclusion list decides what that is worth
An NBER working paper puts the effect at about 34 fewer cans a year per recipient. The substitution argument economists made survives, but not in the form they made it.
Reality
- Evidence52
- Adoption70
- Hype gap+20
- Incentives60
- Confidence50