Friday, September 23, 2011

Searching for Citizen Science Bounties

Photo Courtesy: AMagill
Two weeks ago we had a great discussion after my Finalizing a Definition of Citizen Science post where I first mentioned the concept of bounty projects that could benefit from distributed computing. I also briefly touched on the concept when asking Is Distributed Computing Really Citizen Science?  But I haven't had a chance to describe it further until now.

What I'm thinking of is very similar to, but distinct from, some of the innovation awards that groups like the Ansari XPrize and NASA Centennial challenges have offered.  Those are all focused on achieving a specific technical goal or development of a process.  And they are also quite large endeavors with precise rules on how the goal should be accomplished.  They may be somewhat loose rules, but they are still somewhat prescriptive in how to accomplish the goal. 

The bounties I'm thinking of are different by focusing on discovery, not creation.  They reward accomplishment of a specific tangible goal but are usually smaller in nature or more precise.  The example is an award to people who can find examples of a particular rare bird, sight the first flower of a certain type to bud in spring, or discover an asteroid that will pass within a certain distance of Earth.  So it rewards a very specific discovery that is not a technical feat in and of itself, though building of tools to aid the discovery (such as building the proper telescopes of automating bird call identifications) may involve significant technical work.

Historically people have used these types of bounty prizes to accomplish scientific goals but they haven't been popular recently.   The most recent example of this type of "pure" bounty prize I've seen are the Electronic Frontier Foundation Cooperative Computing Awards for finding the largest prime numbers.  The group offered prizes of up to $250,000 to the first person to discover a certain type of prime number.  There were no rules on how to do it, or what should be done with the number.  All the group required was someone to prove that the number was a Mersenne Prime and of the certain length.  This type of bounty-hunting is well-suited for distributed computing approaches that could crunch huge sets of numbers by brute force until the sought-after number was found.  And that's exactly what the GIMPS distributed computing project did to win the two most recent EFF bounties.

As a side-note they are still looking to win more prizes...read all about it in my GIMPS blog posting.

There are also other uses for bounties that do not involve distributed computing but which could also benefit greatly from the approach.  One example is understanding protein folding...there are millions of potential solutions but only one correct one, so why not offer a reward to the person discovering the shape of certain important proteins from their component molecular structure?  In other words, researchers would provide the components of a key Malaria protein and offer $1,000 to the first person that identifies it's shape.  Or provide the shape of an important AIDS protein and provide $5,000 to the person discovering a structure that will fit around the protein. In other fields I can even imagine prizes to the first person that spots a comet directly crossing Earth's orbit.  The possibilities are endless.

On a similar note, there is another concept that I also put into the bounties category.  These again focus on meeting a simple, tangible goal, but are used to reward progress or effort on a per-unit basis.  Again, nothing is being created here.  Instead we are rewarding someone who performs a scientific analysis task ten times, or collects three samples of a certain specimen, devotes 100 hours of computing time, identifies 200 uncharted Mars craters, or tracks the pollution in ten different streams.

Looking around I've seen a few examples of this so far but nothing major, and nothing active right now.  The closest I've found is the Cosmology@Home, a distributed computing project trying to model the current universe from various hypothetical starting points.  To encourage participation researchers offered a prize to the person whose computer model came closest to reality by a certain date; there was not a monetary prize but the winner would be mentioned in scientific articles about the work.  The most widely-known version may be Amazon.com's Mechanical Turk project.  Although not necessarily Citizen Science, it did provide bounties on a piecework basis for crowdsourced activities.  So people could be paid for writing ten web reviews, or transcribing a certain number of podcsts.  The private sector QMULUS Cloud Computing Platform also used this approach to encourage participation in an actual Citizen Science appplication.  Each month the company gave away gift certificates and free merchandise through a raffle to users of the system.  As a commercial entity they could afford to invest in these give-aways but there's no reason non-profit organizations couldn't do the same thing.   There could also be many variations in the raffle system...an entry for every work unit performed, or for per person using the system per month, or per participant in general.  There are many possibilities that could fit depending on the nature of the particular project.

So these are my initial thoughts on bounties.  Not much has happened in this area recently but I believe there's a strong potential for it to really take off.

But what are your thoughts?  Is this a legitimate subset of Prize Projects in general or have I made an artificial distinction?  Are there other projects currently utilizing bounties that I've missed? Are there existing Citizen Science projects this idea could be adapted to?  There's much more to write but I'd love to hear your thoughts in the comments below first.

No comments:

Post a Comment