Monday, January 18, 2010

playing the verification game

Recently I've been given an exciting present; the PAD, aka Positive Activities database. In line with guidelines, access is via a web interface, and (as presented) it had obvious issues. A quick round of user testing later, we're sorting out most of it. It looks better, there are better links, the langauge is more friendly and other improvements are dropping in as we work down the list.

However, one problem remains, arguably the most thorny -- and definitely the most important. Verifying, extending and improving the data. This isn't just about checking accuracy (although that's important) . It's about capturing in the database all of the possible searches that might lead to a particular activity. It's about presenting an image of existing activities that the young people taking part in them will recognise. It's about increasing search returns without drowning the searcher in irrelevant data. Here are some examples:
  • If an activity describes itself as "street dance" will a young person looking for breaking and beats find it?
  • If someone describes their dance nights as "fun" how can we get that verified by young people taking part in the activity?
  • If someone wants to do salsa, how can we strip out Tap and Urban?

The answers come in, sometimes technical, sometimes a phone call, sometimes as old fashioned as requesting annual reports. But there's still too much filter between the people providing the information and the people who need to use it; we need some way of making this process dynamic, user-led and participative.

At around this time I watch a 51-minute video of Luis Von Ahn speaking about human computation. Well, I say watch; obviously I mean that it was running in a corner while I was doing something else. Keep it visible, though; there's some fun stuff in the slides. He's talking about persuading humans to provide data verification through playing games -- from captcha cracking farms to his own project, an image-tagging game called ESP.

I find myself thinking of Farmville and its ilk, all of those fiendishly popular and addictive sims-esque games on Facebook. A frequent complaint is that playing them serves no useful purpose. Could you kill the sense of time wasted by presenting people instead with a stylised leisure map of their own town? Ask them to fill in information, and reward them for information also entered by other people?* If they play accurately enough, give them the usual set of toys, treats and games to add into their dream home town; grass to roll across the streets, unicorns to graze in the park, penguins to populate the shopping centres.

Develop it for long enough, and it might even become a consultation tool for town planning.

It seemed like a fun idea, so I took it to my gamer deep throat -- currently working on making video recognition work for a celebrity exercise coaching game. He instantly started talking about Second Life and the various world-builder/politics games that small companies make and sell to governments, which was a problem -- small audience, very technically focussed/escapist. Not my real world ordinary leisure experts. Also, he was bang on the money when he said he wasn't sure it would appeal. Why would you waste time building an echo, a photograph, of the real world again in the computer? Even the four-dimensionality would pall. It's already yours. You already live there.

So it's back to running the verification game the old-fashioned way, at least for now. A shame though, I kind of fancied playing games and being useful!

* Players in the GMAP games are awarded points when they enter the same information as a randomly-assigned partner --either one playing at the same time as them, a verification play, or a saved game.