From McFlurrys to Favelas, how crowdsourcing is a numbers game

Two interesting stories this week highlight how crowdsourcing apps can really benefit the people that use them, providing that the numbers involved in the networks are high enough.

The first one is frivolous to all but people with a six-year-old daughter in tow who really wants a McFlurry. It also highlights the weakness in crowdsourcing in that it relies on high levels of participation. You know… a crowd.

As reported on The Verge, Ice Check is a crowd-sourced app for iOS that will effectively tell you if the ice cream machine at your local McDonald’s is downrightnow. Given that means both milkshakes *and* McFlurrys this is an important consideration, especially if, like me, you live equidistant from about four McDonalds, your daughter goes through phases of absolute besotment with the ghastly place*, but each is a 25-minute drive away at different compass points.

(Yes, the nearest Starbucks is 25 minutes away too – welcome to rural life.)

It’s a neat idea that, just like participatory democracy, unfortunately, needs a high level of involvement to work properly. There are great holes in the coverage map, but then it’s exactly the sort of app that needs neutral, self-interested parties to u it for the benefit of the community. McDonald’s will hardly get behind such an initiative itself: knowing that for every person that says ‘Oh, okay then, I’ll try your Coffee Lottery again’ (this being the concept that McDonald’s coffee exists in a binary state of being either quite nice or fecking hideous) there are others that will get in the car again and drive off.

And yes, the machines are that important to the user-base. All of which makes it all the more flummoxing that the WSJ’s digging around in the subject reveals the average McDonald’s ice-cream machine requires an 11-step, four-hour long cleaning process every day.

As The Verge writer puts it: “It’s so complicated the article says: “if someone ordered an ice cream while employees were in the process of cleaning the machines, they often just said it was down rather than reassembling it.” On top of McFlurry-related sadness, I now have McFlurry-related trust issues.”

And now, a more serious crowdsourcing application


Wired’s story, Rio’s favelas are crowdsourcing crime data to keep people safe, is both a lot less trivial and a lot more successful.

This has the crowd to make the sourcing work, 110,000 users of the Fogo Cruzado app providing realtime reports on incidents of violence in Rio de Janeiro in an attempt to keep people safe and away from the areas where trouble is occurring.

And it occurs a lot. The platform has registered over 7500 weapons discharges and, in its first year of operation, there were 1349 fatalities in the city and surrounds.

“When a notification arrives, the app’s algorithms cross-reference the information with specially developed filters and scripts before making it public – on both the interactive map and social networks,” writes Wired.

It’s an Amnesty-backed project, and it’s not the only one either:  Onde Tem Tiroteio (it translates as ‘where shootouts are’) checking incoming reports with a network of 7000 reliable sources on the ground.

Crowdsourcing and AI

The question is how to join the two: how to make sure that the understandable numbers seen in the Brazilian apps that can save lives are also reflected in the trivial ones that very much fall under the aegis of #firstworldproblems.

Some form of AI seems to be the most obvious answer, allowing people to sign up to crowdsourcing information resources but then freeing them from the responsibility of actually have to contribute to them themselves in the same breath.

Of course, AI might even be using a sledgehammer to crack a walnut in many cases. Give Ice Check access to enough other apps on a phone, and it could just flash up a prompt saying ‘Ice Cream machine working, yes or no’ whenever a phone GPS detected someone was in a McDonald’s and then the user-entered information would get carried to a central point via hashtagged tweet, text or whatever.

The key here lies in the prompting, which is classic nudge psychology. But once even that is automated by us in a set and forget manner, the numbers involved in these distributed information networks increase rapidly. From stuff that makes us warm and fuzzy to information that keeps us alive, the amount of data that we rely on on a day to day basis will probably soon be provided by the rest of us simply walking around with a phone in our pockets.

Now, what was that line about trust issues…?

*Ghastliness recently mitigated somewhat by the introduction of the Signature burger range.