This post is part of my liveblogged account of a conference. Two disclaimers: Liveblogging is hard, and I often get things wrong. If I did, please feel free to correct me via email or in the comments and I’ll make changes when appropriate. Second, the opinions expressed in these sorts of posts are those of the speakers, rather than mine.
Ted Vicky is trying to answer the question, “Can Twitter make you fit?” He’s had a long path to becoming a PhD researcher in Galway, Ireland. For eleven years, he ran the online fitness center at the White House, under Presidents Bush, Clinton and W Bush. With his background in exercise physiology, he’s taken on questions of “connected health”, asking how social media might be helping people combat obesity and other chronic health issues. He points out that in only one country in the world – Japan – is obesity decreasing.
He’s collecting data from mobile fitness applications – Runkeeper, MyFitnessPal, DailyMile, Nike’s system, Edomondo – by looking at their social media presence. While these tools have a social media component, it’s easiest to collect data when people share their information on Twitter. He’s gotten over a million tweets, and is starting to understand patterns, both by measuring what tools people are using and sending data into analytics engines like Klout. Perhaps unsurprisingly, the people most using these technologies thus far are categorized by Klout as “explorers”
And Vicky is starting to put together a model to understand what people talk about when they tweet about fitness – reports on activity, “blarney”, and conversations between participants. There’s a closely analyzed set of 36 days of data representing 234k tweets and 57k unique users. He hopes we’re going to learn why people share their exercise information, what reinforcement and motivation we get from each other’s behavior and how this information could turn fitness into a broader social trend.
Vipul Gupta is building sensors at Sun/Oracle research. The Sun SPOTs is an Arduino-like sensor platform, built around a more powerful 32 bit ARM CPU that can run Java natively, rather than a clunky variant of C. There are 20,000 devices out there, collecting data using a variety of sensors, including a tilt sensors and radio. They’re being used by researchers, hobbyists and the educational market.
Gupta notes that mobile phones have lots of sensors, including GPS. He wrote an app for an Android phone that tracked his movements as he drove his car, posting his location every 30 seconds to sensor.network. A simple output is a Google map mashup of where he’s been. This is potentially sensitive data, so Gupta has written tools that can blur the data to allow it to be published, smearing position so it’s less personally revealing. You could also obscure data in terms of time series – perhaps you don’t want anyone to know you’re out of town while you’re traveling. You might be more comfortable publishing that data after the fact. You could also release averaged data – your electricity consumption can reveal when you wake and sleep, but an averaged figure might help your building manager understand load needs at different times of day.
Gupta is now experimenting with adding additional sensors to his devices – air quality, ambient noise, radiological sensing. In the long run, he suspects we’ll not be connected to dedicated sensor devices, but to mobile phones, as they subsume the roles of other computing platforms in our lives.
Dave Marvit with Fujitsu believes that we’re moving from a world where companies create a vertically integrated ecosystem around sensing, to one where we have sensing, analysis and service as separate components offered by separate actors. His experiment with the Sprout platform are designed to show what this new ecosystem might look like.
The platform links an ARM processor running Linux with an Apache webserver with 5 miniUSB ports for sensors. He tells us that it will eventually fold into a mobile phone, just like everything else. In the meantime, it’s an excellent tool for integrating multiple sensor streams. To track stress, he’s analyzing data from a fingertip pulse monitor and an accelerometer. The former is a pretty noisy signal – if you move, the data gets messy pretty quickly. But accelerometer data can tell you about the movement and let you throw out that data. He’s used this system to watch himself playing speed chess and is able to correlate moments of stress to particularly stressful moments within the game.
This ability to integrate multiple sensor streams could lead to better measures of ambulatory blood pressure or sleep apnea monitoring. The goal in the long run is to build a platform that supports may modular metasensors and their synchronization and interaction.