Thursday, June 5, 2014

Strava Resisting the Urge to Be Evil

Don't be evil
Don't be evil used to be Google's slogan.  You don't hear that much anymore.

Aside from the occasional charge of inspiring death in Cap'n Katusha wannabes (going for inter-stoplight downhill segments!), it'd be hard to think of Strava ever having to make a slogan of such urgent moralizing. They didn't really blink when a family started a lawsuit against them, accusing the company of inspiring the death of their Strava-obsessed son.

To anyone who uses the app, it's clear that the app wasn't at fault.  The guy who died was just, to put it mildly, misguided.

While not exactly evil-skirting on the scale of Google, Strava still collects a ton of data, much of it quite personal.  Strava has all the data you've uploaded in your 200 attempts to top Captain Katusha's downhill stoplight-running KOM through the whirling knives factory.  They have MABRA down cold.

This is probably no big deal, although it might be in the future.  As we're discovering with social media generally, the user provides the data (e.g., giving kadooos to Cap'n Katusha for breaking the sound barrier) for one reason and the company uses the data for a very different purpose.  Suppose, for instance, the police want to make sure the 7am stays the hell off Clara Barton. They could simply use Strava's live tracking application to make sure doesn't happen, or to ticket riders after the fact.

Come to think of it, nothing would prevent police from using Strava to ticket cyclists who don't stop at stop signs at Hains, or of putting cyclists at the scene of crimes.  I can think of dozens of other slightly scary invasions of privacy.  Not currently a big deal, but certainly antithetical to the spirit of the sport: liberation.

Do be good
Personal data collection is mostly innocuous when it's just Strava knowing where we ride; it's frightening when it's Apple helping the NSA turn on your iPhone and listen in on your conversation.

How useful can Strava's data be?  The company already makes heatmaps that overlay user data maps.  These heatmaps can be of a single user or of multiple users.  This one is of all Strava users in Washington DC:
As you can see, Hains and Beach are the beating heart and intestine of Strava-using cyclists.

The problem is that Strava users are only a small portion of and, I'd argue, not representative of all cyclists.  If we're looking for an accurate map of bike riding in DC, Strava's heatmaps of user activity are not representative of the entire cycling population of DC.  That is, a heatmap of Capital Bikeshare usage would be very different (there would be little activity on Beach Drive, for instance).  While Capital Bikeshare may seem insignificant for those focused solely on the bike as a means to crush the dreams of freds, its 1,300 trips per day (calculated using data from here) are significant from the point of view of urban planners and probably surpass the total trips of Strava users in the DC area.

Bikeshare data since it is both complete and vast, can yield beautiful animations (this one of London, unfortunately):

Still, Strava may have some statistical tools that allow them to adjust the data (e.g., by pulling out commuting Strava entries by time and regularity) to more closely match the data of the cycling population as a whole.  In that case, the data would be more useful.

Recently they announced a service called Strava Metro for use by cities and urban planners.  Presumably, it adjusts the data to allow cities to examine the routes and behaviors of its commuters.

Such a service could especially be useful because we don't have it now.  City data is collected on particular days or weeks, on a limited scale.  Some data comes from observation; others are from direct surveys.  To its credit, it isn't skewed (as Greater Greater Washington notes) by Strava-philiacs (i.e., spandex wearing, joyriding, Cap'n Katusha acolytes).  It is, however, limited by its scope: it is an extremely small sample size, and it is geographically incomplete.  The city puts a couple interns out on the curb for an hour or two with a clicker counting cyclists out on the 15th Street cycle track.  In no way can that match the size and scope of Strava's data.

Still, the data cities have used till now has been useful.  It has led to the development of cycling tracks on 15th Street and on L.  It also allows us to see where we should ride and where we should avoid.  MoveDC has published results of a survey on perceived safety of streets in DC.  They use a metric, what they call Bicycle Level of Service (BLOS), that rates perceptions of safety: streets cyclists see as safer are colored cooler colors (purple, blue) and more dangerous streets colored yellow and red.  They produced this map:

While cyclist perceptions of safety are important, it's also important to figure out actual safety stats: where is it that accidents occur?

A map (unfortunately, of Portland) along with a heatmap showing number of cyclists would be more useful, since it would show cyclists where to avoid and it would show city planners where to introduce safety measures.

What's the purpose of all this, though?  We get all this data showing where we ride and where we get hit--what do we get out of it?

Maybe we get better roads.  Maybe we inspire more people to ride instead of drive.

Thus far, few Washingtonians have given up their daily drive, and even fewer have done so in exchange for a daily bike commute.  Bike commuting growth in DC has been among the highest in the country (since 1990, DC ranks 3rd in the country in % increase in bike commuting (455%).  Yet, this is a drop in the bucket in terms of total commuters:
We stick to our cars, on average, 67 hours per year of delays--the most delays of any commuters in the country:
Driving to work already sucks.  We just need to make riding to work more attractive.

That is mostly a matter of getting enough leverage to improve city infrastructure, and to improve perceived safety of bikes.  As I wrote previously, would-be bike commuters cite safety as the primary reason they don't ride to work.  If Strava can contribute to making city streets safer for cyclists, then it would certainly be at least one not evil use of personal data collection.

That's the aim of Strava's president and co-founder, who describes Strava Metro's purpose this way: “If cyclists were cars, even in just the numbers that cyclists have now, and they were using the infrastructure that is generally provided to bikes, people would be up in arms about how bad the infrastructure is. So we think Metro is important because it shows just how much commuting is being done by bike and pedestrians. Even before the question of infrastructure, it’s about advocacy and awareness.”

How about that? Strava may turn out to be far more than a tool of obsessive losers bent on KOMs.

No comments: