There are many possibilities when it comes to the potential from researching mobile and location data. As Professor Cecilia Mascolo said: “Devices are in the pockets of people and they reach people that weren’t reached before, in regions and on scales that we haven’t seen.”
Technological advances that were depicted as science fiction on television series like Star Trek in the 1960s are now commonplace in the 2010s, such as translation devices, communication earpieces and touchscreen interfaces. Sensors are now able to collect mobile data that can tell a lot about us. These include accelerometers, light sensors, cameras and microphones, with Mascolo adding, “this is powerful and scary at the same time.”
"Using GPS, we can have a global view of our cities"
One of the powerful uses is the ability to look at the way cities function and how populations use spaces both outdoor and indoor. Mascolo said: “Spatially, because we are using GPS, we can reach very low granularity. But also, because the service is used globally, we have global view of our cities.”
To illustrate this, the professor displayed a visualisation of city spatial data in New York over a period of time. The researchers categorised different locations as arts and entertainment, colleges, or shops with circles of different sizes and colours. The venue was depicted by the colour and the number of people using that venue determined the size. She also said that this kind of data can used to study the growth of cities to uncover how and why they grow the way they do.
The analysis of two picture maps of London taken three years apart in 2011 and 2014 meant that researchers were able to chart the increase in the number of venues in the east of the city. This was due to regeneration resulting from the 2012 Olympics.
Detailed location data on movement and direction can be extremely useful indoors, too. “Using a variety of devices that use radios and one that is able to understand the angle of interactions mean that there are different aspects that can be studied,” she said.
One group of researchers looked at the way in which a set of employees interacted with each other over a three-month period to see the amount of time spent face-to-face and in small groups. The same set of employees were subsequently studied in a different building, again over 90-day period. The researchers were able to create a visualisation that showed greater interaction between the employees in the second building.
“We only knew they were meeting more. We then studied the result with the building structure. They had introduced more meeting places and a bigger cafeteria, so there is an effect. The architects were studying this interaction and the results,” she said.
Mascolo said there is also a way to use mobile data to provide indicators and diagnosis of mental ill health. Microphones can be used to understand the emotions of a user through an analysis of their voice. The MIT researcher behind this demonstrated how this worked on an episode of the Big Bang Theory (Season ten, episode five for those who would like to see it for themselves).
Studies also proved that radio receivers could broadcast waves that reflect off a person’s body which could then show the user’s heartrate and their rate of breathing which is linked to their emotions. “This is starting to be used in care homes to detect how a patient is doing with just waves s passive sensing,” she said.
Despite these opportunities, Mascolo recognises that there are significant limitations, one being the lack of sensitivity of some sensors. She gave the example of her sleep tracker, which one night recorded two periods of zero activity or movement. “Looking at this data makes me even doubt if the sensor is working at all.”
“A human will be able to distinguish movements. Accelerometers might not be able to.”
She also said that sensors can have trouble differentiating between types of human movement. “If you see a person walking quite fast and then a person climbing up a set of stairs, you as a human will be able to distinguish them. There are often cases where accelerometers might not be able to.”
Mascolo said one reason for this may be that the sensors we are using are cheap ones that are put in phones or other devices. “The accelerometer that is in your device to understand the rotation of your device. That was the pure function and now we are using it for all sorts of things.”
The professor added that another limitation is the difficulty of anonymising location and mobility data. “We know that location data, no matter how many times you anonymise it, is not anonymisable to a good extent. We are so unique in our patterns that that only having one or two points indicate that your trace can be deanonymised very easily," she said. Mascolo likened the sacrifice of location privacy to a negative side-effect of a drug treatment.
Mascolo is also sceptical of the notion that all data should go into the cloud. She pointed out that there are places in the world that rely on 3G and 4G connectivity and the uptake of wifi has been slow. In these regions, “to send data and get replies from the cloud would not work in the same way,” she said. She and colleagues have done analysis proving that by sending data to the cloud instead of a GPU, you are actually worse off in terms of energy efficiency.
For her part, Mascolo said that she is working on informing the public to understand that a lot of analysis can be brought down to the device and so does not have to be done in the cloud. She concluded by suggesting that a balance needs to be struck between analysis in the cloud and analysis on the edge or locally in the device.
Professor Cecilia Mascolo was speaking at the Women in Data Science conference hosted by the Alan Turing Institute.