Human beings as sensors provide raw data for scientific analyses

One third of the world’s population carries a smart phone throughout the day.  Smart phones carry a plethora of sensors including location, temperature, light, pressure and movement. Two billion human beings with smart phones covers most of the Earth’s land surface.  A recent study shows what is possible by analyzing anonymized data measured and recorded by smart phones.  Research has shown that smart phones are accurate pedometers.  In the study anonymized data from three quarter of a million smart phones was used to detect and map physical inactivity globally.  Physical inactivity measured by smart phones was related to the prevalence of obesity and to the walkability of cities among other things.  This study is one of the first examples of the type of scientific analyses that is possible using human beings with smart phones as sensors.  Combining and analyzing data from millions or even billions of smart phones using big data analytics will make possible a higher level of understanding of human activity, our built world and the natural world not just for professional scientists but also for citizen scientists.

Introduction

Most of the raw data that forms the basis of climate research comes from remote sensors on satellites, aircraft, balloons, or on-site such as the over 30,000 Argos buoys in the ocean.  Another potential source of raw data are the over 2 billion smart phones that over a third the world’s 6 billion population carry.

An advance in technology that has dramatically changed our world is the development of microelectromechanical systems (MEMS) in the 1980s and 1990s.  This development has made possible the mass production of micro inertial measurement units (IMU) comprised of a combination of accelerometers, gyroscopes, and magnetometers.   These have enabled the smart phone which contains a range of sensors, typically an accelerometer, thermometer, gyroscope, pressure sensor, humidity sensor, light sensor, location sensor (GPS),and a magnetometer in addition to communications allowing connections to the internet over wireless networks.

At the Sixth National GIS Symposium in Saudi Arabia in 2011,  Steve Liang of the University of Calgary Liang discussed  how the smart devices that we all carry can be used to create a sensor web that can be queried for non-private information that can be used for scientific analysis.  As an example anonymized data collected from smart phones is already being used for real-time traffic monitoring.

Steve Liang and his team at SensorUp are making inexpensive air quality sensors available to citizens in Calgary to enable to detect air pollution from wildfires in British Columbia, Canada.  The sensors measure the concentration of fine particles, 2.5 microns in size, which have been shown to have greater affect on human health than larger airborne particles.   Overnight, a distributed, shared, network of PM 2.5 sensors was deployed in St. Albert, with all sensors reporting to a collective map of readings. Citizens were able to monitor air quality in real-time as the B.C. wildfires spread.   In the future inexpensive sensors like these could be included on smart phones.  3D laser scanners have recently dropped dramatically in price and it is entirely conceivable that in the near future smart phones will include this capability

Data

A recent study shows just how powerful human beings with smart phones can be as a source of data.  The study, published in Nature, reports on how  anonymized data from three quarter of a million global smart phones was used to detect and map physical inactivity around the world.   The accelerometer included in smart phones enables a phones to be used as an electronic pedometer.   Recent research has demonstrated that smartphones provide accurate step counts and reliable activity estimates in both laboratory and free-living settings. In this study the dataset includes recordings of physical activity for free-living individuals from 111 countries and includes 68 million days of minute-by-minute step recordings from 717,527 anonymized users.

Analysis

The researchers winnowed down the countries for which data was recorded to 46 that had at least 1,000 users.  For these countries 90% of the users were from 32 high-income countries and 10% were from 14 middle to lower-middle income countries. The average user recorded 4,961 steps per day over an average span of 14 hours.  Extensive statistical analyses were used to correct for missing data and age and sex biases, and to ensure that the conclusions were reliable for both high- and middle-income countries.

The researchers compared national activity levels derived from smart phones with national statistics on obesity.  They found that a computed quantity which they call activity inequality was a better predictor of obesity prevalence than the average number of steps for the country.  Activity inequality is computed in the same way that wealth inequality is computed and reflects not just the average level of activity but also the range of the distribution.  A country with many people with low activity levels as well as many people with high activity levels is predicted to have higher obesity prevalence than one with nearly everyone reporting medium activity levels.  For example, the USA and Mexico have similar average daily steps (4,774 versus 4,692), but the USA exhibits larger activity inequality (a wider distribution) compared to Mexico and hence is expected to have greater obesity prevalence.

The researchers also compared activity levels with walkability scores for cities around the world.  Walkability is a measure of how easy it is to get around a city on foot.  For example, European cities are generally more walkable than U.S. cities.  It was found that higher walkability scores are associated with lower activity inequality.   For example, San Francisco, San Jose, and Fremont are Californian cities that are close geographically to one another.  The smart phone data reveals that activity inequality is lowest in San Francisco, the city with the highest walkability score.  The analysis also reveals that in more walkable cities activity is higher on weekdays during morning and evening commute times and at lunch time and on weekends during the afternoon. This is interpreted as indicating that walkable environments increase physical activity during both work and leisure time.

Conclusion

This study is an outstanding example of the type of scientific analyses that is possible using anonymized data measured and recorded by sensors on a smart phone.  One third of the world’s population carries a smart phone throughout the day.  Human beings with smart phones are moving sensors.  Currently most smart phones are able to report location, temperature, pressure, speed and acceleration, orientation, light intensity, and capture digital imagery.  In the near future smart phones will include air quality sensors, 3D laser scanners, radar and other sensors making it possible to monitor air quality at a much more granular level and by citizens rather than agencies.  Smart phones will be able to capture real-time dynamic 3D maps of urban environments and natural events such as earthquakes and volcanic eruptions.  Smart phones are and will have a strong democratizing impact.  Combining and analyzing data from millions or even billions of smart phones using big data analytics will make possible a higher level of understanding of human activity, our built world and the natural world not just for professional scientists but also for citizen scientists.

Source:

Large-scale physical activity data reveal worldwide activity inequality, Tim Althoff, Rok Sosi, Jennifer L. Hicks,Ab by C. King, Scott L. Delp & Jure Leskovec, Naturem 547, 336–339 (20 July 2017) doi:10.1038/nature23018