Making More Sense

We must present all the data we collect in a meaningful way.
Published: November 18, 2013

According to conventional wisdom, we have five senses—sight, sound, touch, taste and smell. We also experience sensations, such as hot and cold, pain and the passage of time. It’s generally agreed that it is not only useful to sense more than one thing at a time, but it is better—even essential.

As we build the Internet of Things, we are creating a digital nervous system, with a lot of sensors providing cross-referencing, complementary information and confirmation. The sensor that gets the most attention is RFID, and for good reason. RFID tags enable computers to sense identity, and knowing what something is almost always is a prerequisite to being able to use other sensory information, such as temperature. I consider an RFID tag a “sensor” because it can detect something about the physical world remotely and by proxy.

Having a lot of information is good. Showing a lot of information is bad. The first instinct of the typical sensor systems designer is to create a monstrosity of an interface called a “dashboard”—a hard-to-read display with a dedicated chart or dial for every single piece of information.

But our brains do not contain series of dials showing us the amplitude of sound waves coming into our left and right ears, the spectral range of light entering each of our eyes, and values about what we are smelling, touching and tasting. We get one, seemingly complete and comprehensible representation of the world around us. Similarly, information collected from many sensors must converge into a single meaningful number, report, action or some combination of these three.

An example of how to display complex, multisensor data was presented in September, at the annual Ubiquitous Computing (UbiComp) conference, in Zurich. Matthew Kay, a Microsoft researcher, took the show by storm with his insightful paper on a mundane topic: how to improve household bathroom scales. These scales consist of a sensor and an interface. Step on and get a single number that indicates your weight. But weight fluctuates throughout the day.

If a scale could sense who was weighing in (not hard to do, unless several users are the same weight) and days and times, it could show trending, like stock market charts do, making daily fluctuations less of a focus by giving the user the information he or she really needs: Am I getting closer to my target weight?

Kay concludes that adding more sensors to scales but providing clearer information can impact a user’s weight and health. This approach, he notes, can be extended to other basic health sensors, such as blood pressure cuffs, glucose monitors and thermometers.

As we move deeper into the age of the Internet of Things, we need to learn how to make more sense. The rule is simple, if counterintuitive: more inputs, fewer outputs.

Kevin Ashton was cofounder and executive director of the Auto-ID Center. He is currently a general manager at electronics maker Belkin.