Social media and gaming, not total screen time, went along with preferring the immediate option
Among 101 iPhone users, social media and gaming use went along with choosing a smaller immediate reward. Total screen time did not.
AI and technology
Saying a technology works leaves many questions open: for which task, using which version, compared with what, and under what conditions? These readings return to the studies behind those claims.
What to look for in these studies
An experiment with a model, application or device describes that evaluation at that time. It does not show that every later system behaves the same way. Pay attention to the tasks, participants, measures and possible costs left outside the test.
- Which system or version was evaluated?
- Which task and comparison were used?
- What outcome was measured, and what was outside the evaluation?
Explore the research
Models that store summaries reproduced several human memory errors
Using chess, text, and drawings, the models filled in missing details with what they had learned. This was a simulation, with no new participants.
Experts and machines recognized faces similarly but made different errors
Forensic experts, super-recognizers, and neural networks performed similarly on average, but each group made different mistakes.
A small joystick on the screen distinguished five textures
One piece with a magnet and coil pushes and vibrates. Ten people distinguished five textures in about 97% of trials.
Warned in time, people manually checked more when the drone failed
With a simulated drone that failed more often, people given a verbal warning manually checked 63% of the time; without a warning, 22%.
Faces instead of distorted letters: people were correct about 94% of the time
In 42,000 responses in a 2014 study, people were correct about 94%, but blending faces heavily with the background reduced human accuracy.
A 3D-printed orthosis with materials costing $15 to $20
Among three people with spinal cord injuries, one task showed more blocks transferred in a minute. Other tasks had mixed results.
A phone distinguished glows that faded at different rates
Bands created by the camera reading the image row by row made it possible to separate materials that stopped glowing quickly from those that persisted.
When they heard a real recording alongside it, people were better at spotting the fake voice
Listening to each clip alone, 529 people identified 73% of fake voices. Hearing paired versions of the same sentence raised that figure to about 86%.
A €700 camera powered by a solar panel recorded visiting insects
Five cameras recorded insects in German meadows. A computer classified hoverflies well, but beetles and true bugs less well.
After reading on a bright screen, participants reported more visual fatigue
Twelve people read for over an hour on three media. They blinked less on the bright screen, and fatigue rose clearly only there.
Stories with AI ideas: higher ratings, but more alike
Among 293 writers and 600 evaluators, the improvement was greater for those who started with lower scores on a creativity test.
A more human-sounding voice went along with greater stated willingness to use the assistant
Biometric authentication also went along with greater stated willingness. The combination had the highest average, with no detectable interaction in usage intention.
Resistance to an assistant was linked to risk more than to intrusiveness
Among 419 people in China, perceived intrusiveness was not directly linked to resistance. A relationship appeared through perceived risk.
After helping a robot, people responded faster near its robotic hand
Thirty people helped a robot cut soap before an attention task. Among 22 others who did not collaborate beforehand, the pattern did not appear.
Only a minority rejected risky permissions on their phones
Some 26% rejected at least one privacy permission and 33% at least one security permission. For privacy, people who deliberated more rejected more often.
While searching for assistive-technology funding, ten people got stuck
They completed 120 of 164 tasks in a Canadian funding database. After revisions, the second round rated it easier to use.
Broadband: more internet access and use, with no clear improvement in skills
In rural Chinese households, broadband coincided with greater access and use. With technical training, there was a favorable association with skills.
Labeling an image as AI-made did little to change willingness to share it
People who saw labels believed a misleading post less. Deception warnings also lowered some intentions to engage.
Being able to adjust the algorithm made it more popular, but slightly less accurate
When predicting students’ math performance, human oversight increased algorithm choice by 7 percentage points and slightly increased error.
Some content changed, but opinions showed no consistent change
Nearly 9,000 US adults used a YouTube-like platform with altered recommendations, and researchers measured their political attitudes.
A visible, alarming notice: six in ten rejected cookies
About 22% rejected cookies with the comparison notice. Greater visibility and a stronger threat each separately went along with more rejections.
Fewer employees opened the fake link after information or practice
At a Dutch ministry, about one-third visited the test link without intervention. Informed or trained groups were 7 to 9 points lower.
Fewer megabits, similar internet use, and more frustration
In rural British Columbia communities, people reporting less than 50 Mbps had fewer devices and watched less streaming content.
The more useful an app seemed, the more people wanted to install it
With fictional apps, choosing permissions went along with greater willingness to install the weather app, but not the messaging or puzzle apps.
Older adults tried home sensors: few chose a camera
In the second phase, 19 of 22 older adults chose the smart speaker. Three of 22 chose the IP camera.
Transparency and control shaped trust in home sensors
Sixteen people in a Bristol project described what mattered to their trust. Some had trouble recalling or understanding the consent information.
Mouse, touchpad, or touchscreen: none won every Word task
Of 56 tasks, the mouse was fastest in 27, the touchscreen in 23, and the touchpad in six. The mouse had fewer errors in 32.
Password managers: convenient for users, concerning for nonusers
Of 248 US adults, 137 used a manager and 111 did not. Users emphasized usefulness; others worried about concentrating all their passwords in one place.
Feeling in control of data and trust were linked through understanding
Among 68 Brazilian students who tested a simulated smart building, control and trust were linked through privacy awareness, not directly.
