A few days after the Haiti earthquake, Ushahidi‘s Brian Herbert set up a dedicated website to crowdsource the translation and geo-location of text messages from Haitian Kreyol to English. This allowed thousands of volunteers from across the globe to help out in the disaster response. We need something similar for crisis mapping Pakistan but Mechanical Turk style.
I coined the term “turksourcing” a while back to mean crowdsourcing applied to micro-tasks. See this previous blog post for a quick introduction. A colleague from Pakistan recently launched this Crowdmap and short code to map flood related incidents. What I’d really like to see happen now is the development of a Turksourcing plugin for this and any other crisis mapping initiatives in Pakistan.
The idea would be to set up a simple website where incoming text messages could be pushed to for tagging and geo-location. Volunteers would use their email address and a password to access the platform. Once they login, they simply select an incoming SMS which they tag based on pre-set categories like those displayed on the Crowdmap for Pakistan. Volunteers would also map the location of the incident being reported. They would then press submit and move on to the next text message.
Each SMS would have to be validated by 3 or 5 volunteers before being officially mapped. This means that a given text message is only mapped if 3+ volunteers have each assigned the SMS the same tag(s) and approximate location. This is to ensure the quality of the data. If a given user consistently mis-tags/geo-locates incoming text messages, their contributions could be automatically ignored. (As opposed to barring them from the system which would prompt them to try and game it some other way).
Volunteers could also be awarded points for each correctly tagged and geo-located SMS. A public scoreboard could be displayed with the rank of most prolific volunteers to create further incentives to help out by rewarding turksourcing efforts. This introduces a gaming component to crisis mapping as I blogged about here. Colleagues of mine with Revision Labs in Seattle have termedĀ this “Playsourcing”.
The map below represents the location of volunteers who helped out with the Kreyol text messages in January. There’s no reason why we can’t rally volunteers around the world to do the same for the 20 million affected Pakistanis.
I have touched base with friends at Stanford, Crowdflower and with CrisisCommons and hope someone will be able to develop a quick turksourcing plugin for crisis mapping Pakistan and future disasters. Please do get in touch if you have bandwidth to take this on or help out. My email address is patrick at irevolution dot net.
Patrick,
As the ‘turksourcing’ tools come online, one community to possibly look into is the large digital Pakistani diaspora that is currently very active among Google MapMaker users.
Here is the MapMaker Pakistan Facebook page: http://www.facebook.com/GMapsPK
Here is a story from the Lat/long blog of a recent collaboration with the community for an earlier Pakistan disaster: http://google-latlong.blogspot.com/2010/06/map-makers-respond-to-pakistan.html
And here is the Pakistan-specific Google MapMaker discussion forum: http://sites.google.com/site/mapyourworldcommunity/map-your-world/discuss?place=forum/mapping-pakistan
Regards,
Max
Thanks very much, Max, just shared your comment with colleagues in Pakistan
Pingback: New emergency, old challenges | Diary of a Crisis Mapper
The concept of a “Tasking server” is one of the biggest opportunities for energizing an unprecedented community of participants in crisis mapping.
But I have a critique of the name — I think we should stop calling it “Turking.”
Four reasons:
1. It is a trademark of Amazon.
2. The implicitly Orientalist perspective in the origins of the name [1]. This is evidenced in the notion that “The Turk” is an exotic and magical automaton that solves problems for us, the intelligent and fully human “requestors.”
3. By extension, “Turking” implies a system which separates “us” from “the workers in the box” — and this is so damaging to the spirit of crowdfeeding.
4. It is highly problematic to associate nationality with a machine — that is, what if it were called “The Mechanical Indian” or the “Mechanical Californian,” or “Mechanical Nigerian.”
Granted, the name “Turking” will make immediate sense to those who are familiar with the Amazon service of the same name — but I believe that Amazon Mechanical Turk is one of the very worst precedents for collaborative volunteer micro tasking. The Amazon service is designed to be as exploitative as a perfectly fluid market will allow — and correspondingly is widely criticized for its unfair labor conditions (ie, no bargaining, no minimum wage, lax protections). These are quite contrary to our community, which has activism and collective “organizing” at its very core.
[1] On the (fascinating!) history of the name see: http://en.wikipedia.org/wiki/Amazon_Mechanical_Turk and http://en.wikipedia.org/wiki/Wolfgang_von_Kempelen and http://www.eapoe.org/works/essays/maelzel.htm
Very well put, Chris, fully agree. How about micro-sourcing or just keep crowdsourcing?
If I get a vote- it’s for Mechanical Californian.
I totally agree with Chris on this one- actually, it’s funny I’ve been thinking WTF this whole time with regard to Amazon’s “Mechanical Turk” – yes I get that it’s a clever esoteric reference but WTF mate!
-Sean
Pingback: How to cope with very large volumes of crowdsourced reports? Add more crowd! – The Ushahidi Blog