In a lab study of 18 adults reading tongue twisters before and after drinking, a machine learning model using voice spectrograms identified intoxication above 0.08% breath alcohol with 98% accuracy, 98% sensitivity and 97% specificity.

Across five studies with 2,896 people, victims who chose not to broadcast their mistreatment were rated as better moral partners. In one study they scored 6.95 out of 9, against 4.70 for victims who complained widely.

Homes near new Florida noise barriers rose 6.8% in value within 100 metres. Scaled up, traffic noise costs the US $110 billion, falls hardest on low-income tracts, and would justify a $974 tax per vehicle.

Using an invasive beetle that killed Toronto’s ash trees, economists found one extra tree raises nearby property prices by 0.45%, and the hardest-hit areas lost 7 points of canopy and 7% of home value.

In a six-week trial with 486 undergraduates at three US colleges, students given an AI coaching app reported more positive emotion, belonging and resilience and less loneliness than a waitlist group. Depression, anxiety and stress did not change.

Personality traits that an AI extracted from the LinkedIn photos of 96,909 MBA graduates predicted school rank, pay and job tenure. The top fifth of profiles earned 4.2% (men) and 4.7% (women) more than the bottom fifth after controls, a gap similar to the Black-White pay gap.

Among 47,094 tour participants at a selective Northeastern college, hot tours cut applications by 10.1% and rainy tours by 8.3%, yet the weather did not change the quality or type of college students eventually attended.

Five experiments, including a large field experiment, found that apologies for service failures backfire when customers were unaware of the failure, lowering satisfaction, trust, recommendations and repeat business because the apology draws attention to the problem.