Wednesday, January 27, 2010

Rating Birds and Birders

In early December, my brother Pete sent a link to a Birder's World article about BirdsEye, his new iPhone app. Well, his along with a few collaborators. I Googled a bit and found that Kenn Kauffman, who wrote the guide, said in a Birder's World interview that

I got involved because one of the principal people doing this project is an old friend of mine. Pete Myers was one of the three guys running the project. He’s been a friend of mine since the mid-1980s. We were both working for the Academy of Natural Sciences in Philadelphia at the same time, and then he went on to National Audubon. There was a period when I was the associate editor of American Birds magazine, back when Audubon was publishing that, and Pete was the Vice President for Science at Audubon. So I got to go out in the field with him frequently. He’s one of these scary intelligent people. [Laughs] Sort of existing at a higher plane and, you know, just frighteningly intelligent but also a really nice guy, fun and down to earth....

I have not yet found what crimes Kenn Kaufman has committed that Pete must be using for blackmail, but I'm working on it. Meanwhile, Pete just sent a Macworld review (4.5 out of 5 mice), which is nice, and he says they're working on an extension of BirdsEye that will allow birders to submit data as well as using the existing database. And I was thinking about that, and urging him to make it all available as RDF, e.g. via Triplify, and also to try to make sure that failure is tracked as well as success...just generally filling the role of Aggravating Younger Brother. But seriously it would be really neat to be able to say "species S seems to be disappearing from location L" simply because usually-successful searchers clicked to say they were searching, clicked (or closed the app, or just left location L) to say they were giving up, but didn't click to indicate success. And I was driving to the DMV to return some license plates, and spent a little time thinking. So, here is what I would try to do.

Version 0: let's say that the competence of a birder is measured as sightings per hour, and the difficulty of a species (in a given location) is measured as hours per sighting. Each of these are roughly normally distributed, I hope, so we can convert them into IQ-style scores (multiply by 15/stdev, then add 100-mean; you now have a mean of 100, stdev of 15 for competence(birder) and for difficulty(species,loc), respectively.) It's effectively a normalized score in a game of hide-and-seek, where birders seek and birds hide.

Now you can say that sightings and searchers are not all the same. Version 1: a normalized sighting is ranked higher if it's a hard bird to spot, lower if it's easy; just multiply the base value by difficulty/100. A normalized hour of search is ranked higher if it's a competent birder, lower if it's me: multiply the base value by competence/100. Now we can say that the competence of a birder is measured as normalized sightings per actual hour (you get more credit for harder birds), and the difficulty of a species in a given location is measured as normalized hours per sighting (you get more credit for being sought by more skilled birders). Put this in an equation-solver, maybe hand-coded 'cos it's too big for a spreadsheet but probably there's a package out there that will iterate to a solution.

What sort of data is needed? For each search (that's while the app is running in a particular location) you have a searchID, and you remember who the user is, where it is, how long it is, and what species are being sought--and found. That's a basic event. The rest follows. Maybe.

This is indubitably a Bad Idea, but maybe a version 2 or 3 or more could have actual value.

Or then again, maybe not.

Update: (Jun 3) Notes on the Immense Importance of this, at Count your chickens (and robins and pigeons ...), urge researchers:

Dr Elizabeth Boakes, lead author of the study from the Division of Life Sciences at Imperial College London, said: "The lack of recent data on common species and areas of low biodiversity is extremely concerning -- we need people's help to record the birds they see, however commonplace, on bird-watching websites. We think this kind of citizen science will be key to future conservation research. "People may not think that they are helping much by recording the date they saw a pigeon in central London, say, but actually it could make a big difference as we do not know what threats species might encounter in the future. We also urge websites to standardise data entries, for example asking that sightings are directly plotted onto an online map -- it takes a long time to read through people's personal blogs!

Labels: , , ,


Blogger liza myers said...

Amazing! Your father would be proud of you both!

2:33 PM  

Post a Comment

<< Home