Measuring progress in health care AI performance relies heavily on question-answer tests, researchers argue, and not enough on evaluating real-world medical tasks.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/03/030325_ac_ai-health-apps_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/03/030325_ac_ai-health-apps_feat.jpg?fit=800%2C450&ssl=1">
Artificial Intelligence
Medical AI tools are growing, but are they being tested properly?
AI medical benchmark tests fall short because they don’t test efficiency on real tasks such as writing medical notes, experts say.
By Ananya
Bacterial vaginosis often returns after treatment. A new clinical trial suggests treating the woman plus her male partner may help prevent the syndrome from recurring.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/03/030525_AC_BV_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/03/030525_AC_BV_feat.jpg?fit=800%2C450&ssl=1"> 
A researcher examines edge damage on a 1.5-million-year-old bone tool found in Africa, part of the oldest known set of such implements.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/03/030425_bb_oldest-bone-tools_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/03/030425_bb_oldest-bone-tools_feat.jpg?fit=800%2C450&ssl=1"> 
Billions of bacteria, archaea, fungi and viruses live in the vagina. But relatively little is known about these microbes and how they affect overall health.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/021325_mp_vaginal-microbiome_feat.png?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/021325_mp_vaginal-microbiome_feat.png?fit=800%2C450&ssl=1"> 
Clapping hands spew out a jet of air, visualized here by baby powder. That jet helps explain the sweet sound of applause.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022525_ec_clapping_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022525_ec_clapping_feat.jpg?fit=800%2C450&ssl=1"> 
In CAR-T cell therapy, T cells, like the one illustrated (brown) attacking cancer cells (purple), are programmed to track specific proteins on a cancer cell and kill it.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022125_at_car-t-survivor_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022125_at_car-t-survivor_feat.jpg?fit=800%2C450&ssl=1"> 
As a participant in a study, Jon Nelson uses a remote to record his brain signals for scientists to study.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022805_TDE_ep4_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022805_TDE_ep4_feat.jpg?fit=800%2C450&ssl=1"> 
A large measles outbreak in Texas demonstrates how quickly the disease can spread when too few people are vaccinated. This warning sign greets patients at an Odessa, Texas, medical clinic.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022825_ts_vaccine_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022825_ts_vaccine_feat.jpg?fit=800%2C450&ssl=1"> 
A new device can replicate food flavors in virtual reality by squirting chemicals onto the tongue.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022825_sm_virtual-taste_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022825_sm_virtual-taste_feat.jpg?fit=800%2C450&ssl=1"> 
Shards of glass are what remain of a young Roman man’s brain. Tests show that parts of the brain turned to glass after reaching 510° Celsius.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022625_av_volcano-ash_feat-1.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022625_av_volcano-ash_feat-1.jpg?fit=800%2C450&ssl=1"> 
An investigation of a previously excavated site in West Africa finds that people inhabited that region’s rainforests, such as the one shown here, by around 150,000 years ago.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022625_bb_rainforest-humans_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/022625_bb_rainforest-humans_feat.jpg?fit=800%2C450&ssl=1"> 
In experiments, scientists have found that gut bacteria can have an effect on mice’s desire for sugar. Some companies claim probiotics can have the same effect in people.
" data-medium-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/030125_column_feat.jpg?fit=680%2C383&ssl=1" data-large-file="https://i0.wp.com/www.sciencenews.org/wp-content/uploads/2025/02/030125_column_feat.jpg?fit=800%2C450&ssl=1">