Researchers believe AI may be used to measure the amount of sleep time lost due to internet use.




 Everyone sleeps, but there aren't many techniques available to measure how much sleep is being consumed globally. We could investigate global shocks in almost real-time with the aid of AI and sleep. It's common to hear individuals mention being fatigued several times a day, but why? Sleep is essential for good health, but because it is so personal, there are few measures available to gauge how much sleep an average person is getting. Time logs, sleep questionnaires, sleep labs, or more recently, wearable technology is all used in the methods currently in use to monitor sleep. However, none of these strategies is prepared to address a global epidemic of sleep loss. 

Internet addresses reflect the daily cycle of human behavior as individuals move online and offline during the day: a trough in the morning, growing activity throughout the day to a peak in the evening, and then a significant decline overnight. However, no two cycles are the same: the day of the week counts (going out on a Friday night slows internet activity), staying at home orders definitely matters (we surf online earlier and longer), and in predominantly Muslim areas, activity falls during Ramadan prayer times are also noticeable.

Americans are questioned about their day's activities, including when they woke up and when they went to bed, in the American Time Use Survey (ATUS). The Monash study calculated when people went to sleep and woke up each year using survey data collected from 81 US cities over six years. The same calculation was then made using internet usage data. 

A machine-learning algorithm was then taught by the researchers to monitor changes in internet usage over a day about the typical waking and sleeping hours in each city. 

The system was accurate to within 20 minutes when asked to forecast the anticipated average sleep time for a city it had never encountered before. It was accurate to nine minutes when determining the typical morning wake-up time. 

A machine-learning algorithm was then taught by the researchers to monitor changes in internet usage over a day about the typical waking and sleeping hours in each city. The system was accurate to within 20 minutes when asked to forecast the anticipated average sleep time for a city it had never encountered before. It was accurate to nine minutes when determining the typical morning wake-up time. 

This implies that it is possible to estimate our global sleep patterns in almost real-time for any (internet-connected) city on the planet. These studies can be used for a wide variety of purposes, including impact mapping after natural catastrophes, recording internet outages linked to breaches of human rights, and even offering assessments of internet accessibility during the Russo-Ukrainian War.  

It will be interesting to see if this strategy can be used globally. American technology and sleeping patterns can be unusual. If so, a machine learning (ML) model that picks up on the internet and sleep association in the US will fail outside of its borders. Internet usage patterns are likely to be influenced by the technological mix in use; for example, a continent that prioritizes mobile use, like Africa, may have quite a different internet usage patterns than North America, which mainly relies on desktop computers.  

Both obstacles can be overcome by expanding the model's training pool, as is the case with many other difficulties in the application of AI to the health sciences. The more data from conventional sleep studies that researchers have, across various geographies, cultures, and technological contexts, the more confident they can be in any model prediction. 

Population health and sleep scientists may stand to benefit the most from the creation of a global sleep observatory (based on internet measurements).

Researchers may swarm to the area and use more specialized gear to further study if significant increases in internet usage reflect similar changes in sleep habits. Significant global shocks like pandemics and recessions can also be studied in near real-time for their effects on our sleep, prompting timely education about the value of sleep during stressful times, better technology and app design, and the right public health messaging around mental health and sleep.

Comments