In an experiment with 383 UC Berkeley students, those holding socially unpopular views self-censored more. Showing students how many peers had stayed silent corrected their beliefs about public opinion and made holders of unpopular views about 17% more likely to speak.
In 11.66 million Fiverr transactions, longer reply delays meant a lower chance of winning the contract, and even 5 to 10 minute delays dramatically reduced hiring; experiments with 3,603 people showed fast replies signal future responsiveness.
A study of 2,606 young adults in 28 countries and Hong Kong found that recalling a nostalgic memory increased social connectedness, self-continuity and meaning in life across cultures, while its triggers varied with national wealth and climate.
Gaze patterns across 700 photos identified individuals 52% to 63% of the time, against 1% to 2% by chance, and people whose gaze on social scenes was more distinctive reported fewer autistic traits.
Surveys of Chinese parents show the 2021 ban on for-profit private tutoring raised expected total fertility by 7 to 8%, mainly because parents saw less educational competition. Birth rates also fell less in cities hit harder by the ban.
In blind tests, raters found ChatGPT 3.5's jokes funnier than those of 63% to 87% of laypeople, and its satirical headlines scored 2.37 out of 6 against 2.49 for real Onion headlines, a gap too small to be significant.
Across four preregistered experiments with 770 participants, people rated sources as more credible when their statements had been repeated, and that boost carried over to new statements from the same sources.
Across about 1,000 languages, those that pack more information into each word were spoken faster, and their conversations covered a conceptually narrower range of topics, discussed in greater depth.
A review of brain imaging and whole-body measurements in mammals finds that solving cognitive tasks usually adds little or nothing to overall energy use, because keeping the resting brain running is already about as costly.
In an experiment with 206 law students, AI highlighting of legal complaints cut task time by 29.7% with accuracy unchanged, while AI summaries saved no time and combining both tools was slower than highlighting alone.