How can a small amount of information contributed by each of a large number of volunteers be combined to reveal valuable scientific results? Here’s an example of #citizenscientists, worldwide, providing a UC Berkeley lab with access to the motion-detection sensor (accelerometer) in their smartphone to create an earthquake-detection system:
Scientists at the Berkeley Seismological Laboratory created an Android phone app, MyShake, which runs in the background, even when you’re sleeping (but doesn’t use much power). The app can identify earthquake-like shaking vs. other kinds of movement (for example, human motion or the motion of a car). When the app determines that the motion is earthquake-like, information on location and magnitude is sent to the UC Berkeley central system.
If lots of phones near one another all detect earthquake-like shaking at the same time, then the UC Berkeley scientists can be confident that an earthquake really happened at that location. If, on the other hand, the information is coming from only a small number of phones, then there’s no way to be sure. The motion of a single phone could be misinterpreted as earthquake-like shaking when it’s really not and the simultaneous shaking of a few phones could be a coincidence. The more phones that detect “earthquake” at the same time and place, the more certain the system can be that it is a real earthquake.
This is because when a sample size is small for any data set, whether from smartphone motion sensors, baby nap-lengths, bird counts, or rainfall measures, anything that looks like an interesting result could be due solely to chance, random events or mistakes. Scientists use statistical analyses to infer how “sure” they can be about how to interpret a data set. Very large data sets give results for which scientists can have more confidence. Often, a small data set will not provide sufficiently reliable information to support any interpretation, not even a weak one.
While some countries, like the US, have seismological instruments for early earthquake detection, the smartphone data could enhance those systems. In many parts of the world that are earthquake-prone, there are no government-run detection systems, but there are lots of smartphones. In the absence of a small number of highly-accurate but expensive detectors, a huge number of less-accurate but cheap detectors, combined, can serve the public.
UC Berkeley scientists are creating an earthquake sensor with data provided by people all over the world. Each phone, on its own, provides information about the movement of one 10 square inch spot on the planet. When the information from hundreds or thousands of phones is combined, it becomes a powerful data set that provides valuable information about earthquakes.
NYU Baby Sleep Study scientists are trying to create an infant brain development sensor with data provided by thousands of parents. On its own, information about one baby’s naps and meals may be of interest or use to that baby’s caretakers. By combining every participant’s contribution, it will become a powerful data set that will provide valuable information that we’re convinced will enable doctors to offer better, scientifically-based advice that parents can use to manage their infants’ sleep and eating. We also aim to discover markers or early red flags for developmental disorders.